The trouble with automating too early is that you scale the mess.

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

Most people want to make a meaningful difference through their work—whether it’s designing intuitive apps, offering an online course, or baking artisanal banana bread. And whatever the goal, it always begins with ideas. Lots of them. Usually, way too many.

And in the age of AI, it is very easy to add yet another automation to implement one of those ideas. Maybe even too easy. So what if you have twenty AI agents doing work for you while you sleep? If you don’t have the headspace to review all the outcomes—congratulations!—You’ve implemented agentic productivity theater.

I’ve spent countless hours building automations for things I thought were important: customer orders, invoicing, and reports, only to watch the business models they supported crash and burn. The ROI was a solid zero. I could have saved myself months of frustration by asking one question first: Do we even need this workflow at all?

The answer, embarrassingly often, was no.

The Solo Worker’s Trap

Here’s what makes this especially painful for solopreneurs and one-person businesses. When you’re the single wringable neck—the only person accountable for everything—automation feels like oxygen. You’re drowning in tasks, so you grab the nearest AI agent and throw it at the problem. I get it. I’ve done it. And I’ve regretted it many times.

The trouble with automating too early is that you scale the mess. You don’t get efficiency. You get faster chaos and cognitive overload. Elon Musk admitted as much about his own Tesla factories: he tried to automate many steps before questioning whether those steps should exist at all.

So before you hand your processes over to Claude Cowork, OpenClaw, or a shiny new n8n scenario, let me share a map I’ve been refining for years. Five orientations, not prescriptions. Five questions to ask before you launch a single automation prompt.

1. Clarify and Eliminate

Start with the question Musk keeps hammering: what’s the point? Tie every idea to someone who really needs it. Not “the legal department”—point at a specific human who can explain why this request exists and what it means for them.

I once spent weeks building an onboarding workflow for a new business idea that collapsed less than two months later. Nobody needed it. Nobody cared. I just thought it was a good idea, until it wasn’t. The process shouldn’t have existed in the first place. The best requirement is the one you kill before it wastes your time.

2. Simplify and Accelerate

Whatever survives the kill list gets stripped down. Steve Jobs understood this with the iPod and iPhone—every button and widget faced ruthless scrutiny. Does this need to exist? If yes, make it so intuitive that people barely have to think.

Having fewer steps means fewer errors, faster results, and less of your evening spent debugging workflows. This is where solo workers and lone AI orchestrators tend to overcomplicate things, by the way. We add options because we can, not because we should.

Having no human team to push back makes it dangerously easy to build yourself an automated monstrosity with more bells and whistles than common sense. Human teammates have the decency to ask “why?” or at least roll their eyes. AI agents don’t. They happily execute whatever pointless task you give them.

3. Amplify and Elevate

After eliminating and simplifying, you might feel the urge to add something. Good—but Hick’s Law says decision time grows with the number of choices. So only add options when they don’t complicate things, when they don’t add friction to other people’s experiences.

For example, when I publish a blog post, I share it on LinkedIn, X, Threads, and Substack Notes. That amplifies my reach without complicating the experience for anyone. It makes it easier for people to find me, not harder. The addition is orthogonal—it extends what’s possible without interfering with what already works.

The question to ask: does an additional feature idea create new possibilities without forcing people to deal with more options?

4. Codify and Automate

Now you can think about automation. This is where your specialized AI agents, workflow scenarios, and algorithmic managers earn their keep. But only after you’ve eliminated and simplified.

I once automated pan-European invoicing for a business that never even achieved product-market fit. What was the point? All my automations got switched off. Automation is the reward for doing everything else right.

The prerequisite for automation, of course, is codification. If you can’t describe a process clearly enough for a machine to follow it, you’re not ready to automate it. No codification, no scalability. And for Solo Chiefs running a “cockroach business”—one built to survive any environmental change—premature automation is one of the most effective ways to waste your time.

5. Specify and Delegate

After you’ve automated what can and should be automated, you’re left with tasks that require judgment. The question becomes: whose judgment? Could you hand this to a personal assistant, a business partner, or a sophisticated AI agent orchestrator?

Here’s what you should never delegate: work that wastes time (eliminate it), work that’s too complicated (simplify it first), or work that a basic algorithm or simple agent can handle (automate it). Delegation is for what remains after you’ve been through the other four steps.

I know the operational loneliness of carrying the full picture. The temptation is to delegate everything that feels heavy. But passing down convoluted processes to humans or advanced AI agents is foolishness, not leadership.

The Order Matters More Than the Steps

To create these Five Steps to Lean, I borrowed from Lean Thinking, from Musk’s five-step algorithm, from the ESSA Framework, and from the “four steps to freedom“ that productivity bloggers keep reinventing ad nauseum. I added the amplification step in the middle because I wanted to balance efficiency with effectiveness. Claude and ChatGPT agreed.

👉🏻 Note: I count simple, specialized AI agents among ‌simple workflow automations (step 4). I count complex AI orchestration (whether performed by humans or AI) as delegated management (step 5). The difference matters. 👈🏻

The real insight for leaders in 2026: AI agents make the wrong order more expensive, not less. When automation was manual and slow, premature optimization wasted your time. Now that AI agents execute at lightning speed, premature automation wastes your time and your money and creates problems at scale.

The map is simple. Eliminate. Simplify. Amplify. Automate. Delegate. In that order.

Jurgen, Solo Chief

P.S. This post was first published on Leadership in Change.

Share

Your problem isn’t the problem, your story about it is.

In Sicily, reframing means swapping your cappuccino for a macchiato, and maybe your whole attitude toward your problems.

Be 100x More Productive with This Free Claude Code Guide

Sponsor message presented by Iwo Szapar

Every Solo Chief hits the same wall: too many decisions, not enough you. I hit it last year and rebuilt my workflow around one constraint: Claude Code has to be able to do anything I can do, from one brain.

Today, it runs five jobs in parallel. drafts emails in my voice, sends the weekly newsletter, handles sales follow-up, prepares my morning briefing, watches my numbers.

Team output. Solo overhead. That's the shift.

If you want to see how it's built, the free guide walks through the first skill end to end.

Free guide: Claude Code for Non-Coders

I’m changing my coffee habit.

Normally, I only drink cappuccinos and cafe lattes, one after breakfast, one after lunch, and one after dinner. But I’m traveling in Sicily now. Everyone tells me that Italians frown upon the idea of ordering milky coffees after 11:00, especially on its more conservative southern island. Of course, I could just ignore the warnings, order my regular cappuccino at 4 pm like the average dumb tourist, and shrug off the local chuckles and eye-rolls behind my back. I’m a paying customer. Who cares what others think?

But that’s not what’s actually happening. Instead, I took this as a chance to get out of my comfort zone and test my adaptability to the local culture and customs. Also, they tell me this is the home of the mafia. I’d rather not find out what happens to disrespectful tourists. So, after lunch, as long as I’m in Sicily, I only order a cafe macchiato, like a decent Italian. And the macchiatos are not so bad, actually. After two weeks of traveling, I might even get used to the reduced milk intake and take the new habit back home. I’m turning the problem into an opportunity.

We call this reframing, or flip-thinking.

We have a great word for it in Dutch: omdenken (coined by Berthold Gunster).

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

Reframing or flip-thinking is a way of thinking and acting in which you don’t shy away from problems, but use them as a source of energy for new possibilities. Instead of resisting reality, you accept it as it is, so you can build creative solutions from there.

Reframing is not the same as rethinking. With rethinking, you ask, “What is the truth here, and how can I understand this better?” It involves taking apart old assumptions and redesigning your approach. The problem doesn’t go away; you just try to find a different angle to tackle it.

With reframing (omdenken), you ask, “How else can I interpret this?” The facts of the situation don’t change, but your story about them does, and so does your emotional response. A situation is only a problem when that’s how you decide to see it. The moment you stop treating something as a problem, it might become an opportunity.

For example, you might have been looking for a job for a year, writing a hundred cover letters and sitting through a dozen interviews, all in a hyper-competitive job market where both recruiters and applicants are using AI to ruin the hunting game for everyone. Is this a problem? Maybe. Maybe not. Perhaps you can reframe it as a chance to offer a new job-matching service and disrupt the job market forever.

Or you might have struggled for a year to grow your Substack newsletter. You followed all the advice, tried every growth hacking trick on the planet, and the subscriber count hasn’t budged an inch. Is that a problem? Only if you think it is. Maybe you can reframe yourself as one of those rare authors blessed with a stable and persistent readership. Perhaps you can even teach others how to gain such a loyal following.

Or perhaps you’re one of those creatives who has worked with teams, both startups and corporate, for several decades, and you’ve realized that you do your best work alone and find teamwork increasingly stressful. Will you despair and call yourself antisocial? Will you sign up for online courses on networking and how to make friends? Or is this a perfect opportunity to learn how to run a one-person business with only AIs as your teammates?

I’m flip-thinking my way through life right now.

I’m reframing things every day.

Never let a good crisis go to waste. When life gives you lemons, you make lemonade—or limoncello.

In an increasingly uncertain environment where things rarely go as planned, I have to adjust my mindset all the time and keep wrapping my head around a newly discovered reality that wasn’t the reality I was aiming for. But the older and wiser I get, the less I want to waste my time seeing everything as a problem to solve.

Some might call it “taking things as they are” or “making the best of a bad situation.” A few readers might point to the psychological concept of cognitive dissonance, which is all about reinterpreting an undesired situation in a new way so that it suddenly seems to be exactly what you needed. It’s a basic mental survival skill, or else we would all get very depressed very fast.

But reframing (omdenken, flip-thinking) goes a step further. It’s like enthusiastically embracing cognitive dissonance. We’re not just reinterpreting a situation and shrugging our shoulders with the comment, “Oh well, it’s actually not so bad.” Instead, we go the extra mile by owning our new perspective and taking advantage of it. We say, “All right, given this is what I was dealt, this is how it’s going to pay my bills.”

Stoic philosopher Marcus Aurelius wrote that the obstacle is the way. The impediment to action advances action. What stands in the way becomes the way.

Every Solo Chief hits the same wall: too many decisions, not enough you. I hit it last year and rebuilt my workflow around one constraint: Claude Code has to be able to do anything I can do, from one brain.

Today, it runs five jobs in parallel. drafts emails in my voice, sends the weekly newsletter, handles sales follow-up, prepares my morning briefing, watches my numbers.

Team output. Solo overhead. That’s the shift.

If you want to see how it’s built, the free guide walks through the first skill end to end.

by Iwo Szapar

Claude Code for Non-Coders

I’m in the city of Noto right now. The old city of Noto (Noto Antica) was destroyed by the great Sicily earthquake of January 1693. Afterwards, the authorities chose to abandon the original site and build a new Noto a few kilometers to the south, on a safer, more suitable plateau where they could design a modern Baroque town from scratch. They reinterpreted the problem of complete destruction as a rare chance to pick up the entire town and move it somewhere better.

That’s omdenken.

That’s flip-thinking.

That’s reframing a problem by turning it into an opportunity.

Jurgen, Solo Chief.

P.S. Ever ordered a cappuccino after lunch in Sicily? How did it go?

Going online is like walking onto a bazaar where everyone is poking you, luring you, and yelling at you.

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

I’m so tired of social media feeds.

I’m tired of the syrupy, vague, I-have-nothing-to-say-but-must-post-a-Note platitudes that take up half my feed on Substack. “Show up consistently, and the results will follow.” “Be kind to yourself today; you’re doing the best you can.” “Your voice is your moat—nobody can compete with that.” As if just keeping your mouth shut is a recipe for disaster.

I’m tired of seeing ever-pervasive Substack and LinkedIn note formulas … Spicy opinion + contrarian twist. Mini-framework in three simple bullet points. “X things I wish I knew earlier.” Hot take followed by a poll or question. “Here’s what I did, here’s what happened.” And don’t forget to make your first sentence bold!

I’m tired of the endless parade of carousels, infographics, and inspirational quotes dominating my feeds, each one more pointless and forgettable than the last. As if meaningless chatter received a visual upgrade. Hardly a word gets written these days without some AI-assisted process auto-embellishing it with lines, labels, logos, arrows, avatars, annotations, charts, chevrons, and checkmarks.

I’m tired of getting sucked into algorithmic recommendation rabbit holes. After chuckling at one funny video of a child doing a face-plant in the snow, the only videos I get are of kids from all over the world falling, slipping, dropping, sinking, collapsing, plunging, and plummeting. And, of course, I get endless videos about snow.

I’m tired of all the DMs I get from people trying to sell me their products and services. “Gently resurfacing this in your inbox.” “Just following up on my last message in case it slipped through.” “If this isn’t a priority at the moment, no worries at all—just let me know.” “I’ve followed up a few times and haven’t heard back.” Indeed, and you most certainly won’t!

I’m tired of the endless ratings and evaluations that companies want me to give them. “Enjoying this? Rate us in the app.” “Your feedback means a lot to us—could you share a short review?” “Other people rely on honest feedback. Mind sharing yours in a quick evaluation?” I almost dread submitting a helpdesk ticket because it inevitably ends with a plea for an evaluation.

And like probably everyone else, I’m very tired of clickbait headlines that find their way across all social media and news platforms. “The Shocking Truth About [Common Belief]” “11 Little-Known Facts About [News Topic]” “Experts Are Sounding the Alarm About [Trend]. Here’s Why” and “5 Crazy Moments From Yesterday’s [Event] — Number 3 Is Wild.”

I’m so tired. Going online is like walking onto a bazaar where everyone is constantly poking you, luring you, and yelling at you.

What I’m most tired of is feeling compelled to participate in this relentless attention theatre myself. Testing different headlines for my posts. Checking conversion rates. Optimizing for discoverability. Frantically evaluating the statistics dashboard. We all want to be read. We crave to be liked. There’s no shame in admitting the desire to be heard. And so we play the same game, whether we enjoy it or not. Indeed, I plead guilty to a few of the annoying practices I listed above (but not all of them).

Because it never stops, does it?

The choice seems clear: either market like a maniac or languish like a loser.

The third path—just write like you enjoy it and cherish the few who appreciate that—seems unattainable in the age of AI-generated slop. (And yes, I used those em-dashes intentionally. I typed them myself.)

It might be worth a try.

Just write what I want and screw the whole attention market. I will think about that while I enjoy a well-deserved vacation in Sicily.

Jurgen, Solo Chief.

P.S. Did you ever try ignoring the algorithms for two weeks? How did it go?

Share

For the neurodiverse mind, AI is not a prosthesis. It’s an interface adapter.

Last week, after a keynote I gave in Spain, I chatted with someone in the audience about neurodivergent people in the age of AI. He has ADHD, and he suggested that now might finally be the moment for people with neurodiverse minds to step forward and take charge. An ADHD diagnosis, he argued, might actually be a better fit for environments that demand constant exploration and context-switching.

I nodded vigorously. As someone with a mild case of autism, I find I often do my best work as the sole human in a room full of AI agents and algorithms. No social demands, no office politics, no colleagues chattering at my desk about their far-from-interesting weekends. Just me and my eccentric mind.

Others with differently wired brains, I told him, might do better in corporate environments where teams keep shrinking, where groups are more fluid and ephemeral, and where reteaming happens all the time. (Which, in case you haven’t noticed, is roughly where corporate life is heading.)

The conversation stuck with me. Is this an actual trend, or just two odd guys at a conference convincing each other we’re about to inherit the earth?

To what extent does the research literature support our little hypothesis? Can we honestly call the future of work, in the age of AI, a golden era for the neurodivergent mind?

I decided to ask my digital research team.

A note on my AI research approach: After framing a deep research question like the one above, I give the same question to five LLMs, each playing a different role. Perplexity is the research analyst, focused on documented evidence. Gemini is the structural analyst, digging into why something is happening and what makes it resistant to change. ChatGPT is the practical strategist, answering what to do about it. Claude is the contextual strategist, looking at the question through the lens of my target audience. Finally, either Grok or Le Chat plays the contrarian. It maps out the mainstream consensus and then takes it apart. The result is five deep research documents with different perspectives based on the same Research Question. It’s like having a team of rather opinionated researchers trying to formulate one answer together.

Then I feed all five documents into Gemini, which turns the Research Question into a Research Map, showing where the LLMs agree, where they contradict each other, and where one of them coughed up a unique insight that the others somehow overlooked. That whole map goes to Claude, who then decides what’s the best way to write about it and turns it into a narrative structure with an Article Brief ready for the ghostwriter. Finally, the Article Brief and the five original research documents go to ChatGPT, who spins it all into a cohesive story. And yes, I have automated this workflow.

What you read below is the result (edited by me for style, readability, formatting, and proper URLs).

A Golden Chance for the Neurodivergent

The hypothesis that ADHD minds are built for constant context-switching falls apart the moment you read the underlying studies.

The idea spread because it sounds plausible in an AI-heavy workplace. Tabs multiply. Tools interrupt each other. Teams bounce between channels, prompts, meetings, and half-finished drafts. Surely the people who grew up with mental pinball have the home-field advantage?

Except the research keeps pointing the other way.

A review in Frontiers in Psychiatry found that adults with ADHD show higher error rates in task-switching and linked those problems to working-memory limits, which is a long way from being “natural switchers.” A systematic review of neurodiversity in employment reached a similar conclusion in a more practical form: outcomes improve when work offers fit, flexibility, and openness to neurodiversity, not because ADHD somehow turns switching costs into magic. That distinction matters a lot. When leaders believe the myth, they might design jobs with even more interruptions and call their workplace “ready for inclusion.”

They would be dead wrong.

The better question is whether AI-era work can become a better fit for neurodivergent minds than the office systems many organizations still cling to. On that question, the evidence is far more interesting. Indeed, this could become a golden age for some neurodivergent workers. But only by design.

The ADHD Context-Switching Myth

Laboratory task-switching and workplace context-switching aren’t identical, of course (life is messier than a key-press experiment). But the lab evidence still matters because it measures the cognitive mechanics underneath. When adults with ADHD perform worse on switching tasks, that tells us something important about what constant interruption is likely to cost them.

The more grounded account is this: some people with ADHD do well in fast, novel, interest-rich settings, and do badly in settings built around routine, delay, and admin drag. That’s a person-environment fit argument. ADHD-related difficulties can become assets in workplaces with flexible practices and openness to neurodiversity.

Some ADHD workers may enjoy project variety. Some autistic workers may prefer low-politics, low-interruption collaboration.

That’s very different from “ADHD people love chaos.”

Many do not. They may like novelty. They may like urgency. They may like non-routine work. But constant task-switching is something else. Leaders who confuse novelty with fragmentation are building expensive problems.

Where Real Neurodivergent Advantages Live in the Age of AI

Once you stop forcing the argument through the context-switching cliché, genuine strengths come into view.

The strongest evidence sits around entrepreneurship and early-stage work. A 2026 meta-analysis covering 47 studies and 298 effect sizes found that ADHD hyperactivity and impulsivity were positively associated with entrepreneurial attitudes and behavior, while inattention was negatively associated with post-launch outcomes. In plain English, the front end looks promising. The middle and later chapters look harder.

That fits what many organizations are already drifting toward: smaller units, more solo operators, more fractional roles, more one-person businesses with AI doing some of the back-office grind. The AI age may be friendlier to the neurodivergent founder than to the neurodivergent middle manager. Those are different species.

There’s also better evidence for autistic preference in text-based, mediated communication than most futurists seem to realize. A smartphone-monitoring study found autistic adults preferred written over verbal communication in real-world use. A Leiden study reported that adults with ASD preferred computer-mediated communication such as email, instant messaging, and text. That lines up neatly with AI-assisted work (and my personal experience) because so much of it is written, asynchronous, and explicit.

These conclusions look less like “superpower” rhetoric and more like a cognitive prosthesis. It’s a useful metaphor because it avoids the sillier mythology.

AI as a Workplace Adapter (Not a Prosthesis)

This is the part of the story where the evidence is surprisingly coherent.

The UK Department for Business and Trade evaluated a Microsoft 365 Copilot pilot and found overall satisfaction at 72%, while neurodiverse employees reported significantly higher satisfaction at 80% and were more likely to recommend the tool (DBT evaluation). The benefits were strongest around written tasks, summarizing, drafting, and information handling. One ADHD participant said the AI had “leveled the playing field.” (Dataconomy reporting on the pilot)

If a tool reduces the tax on note-taking, organizing, rewriting, recall, and social translation, then it changes the economics of cognitive work for people who were previously paying extra for all of that.

There’s similar evidence in autism research. In a CHI 2024 study, autistic adults preferred GPT-4 over a human confederate for workplace communication advice, though some AI advice still missed context and could be poor in practice (ACM Digital Library summary). The important point isn’t that AI is wise. (My team definitely isn’t.) The point is that an always-available, text-based intermediary can lower the strain of workplace interaction. (Amen.)

Remote and flexible work point the same way. A study from Curtin University found that working from home reduced anxiety, improved wellbeing, and raised productivity for autistic workers by reducing sensory overload and social pressure, while still carrying risks around isolation and overwork (Curtin University). Again, design matters. The gains appear when people get more control over the interface.

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

The Fluid Team Problem for Neurodivergent Workers

However, the evidence for a neurodivergent advantage in fluid, ephemeral teams is weak.

Research on fluid teams generally warns about trust formation, coordination overhead, and weak shared mental models when membership keeps changing. Another study on autistic work outcomes points in the opposite direction of the claim: predictable settings, clear communication, and explicit support matter a great deal.

This shouldn’t be shocking. Reteaming means new norms, unclear status, fresh handoffs, hidden expectations, and social renegotiation. Some organizations like to call that ability, but often it’s just continual onboarding and offboarding with better marketing.

Leaders should treat fluidity as a coordination burden that must be structured, not as a gift that everyone should be grateful for. There are definitely ways to address the additional cognitive load, but it requires mindful organization design.

From Diagnosis Labels to Task Signatures: A Design Shift

The most useful shift in this whole literature is from diagnosis to design.

Instead of asking whether ADHD, autism, or other forms of cognitive diversity are advantages in the age of AI, we should ask what a role demands on multiple dimensions: switching load, ambiguity, sensory load, social intensity, written communication burden, and tolerance for error. Then redesign around the task signature.

That means capping concurrent priorities. Written handoffs. Clear definitions of done. Fewer surprise meetings. Optional asynchronous channels. AI support for drafting, recapping, decomposing, and searching. Human oversight where judgment carries legal, ethical, or reputational costs.

This is less glamorous than talking about superpowers. It’s also how actual work gets fixed.

The Blind Spot in the Neurodiversity-AI Optimism

One caution belongs at the end.

The “golden age” narrative assumes that more AI mediation, more stimulation, more cognitive intensity, and more always-on possibility are broadly helpful. For twice-exceptional people, and for anyone whose mind already runs hot, that assumption may age badly. An environment built around constant prompts, endless synthesis, and permanent partial attention can become an amplifier.

Organizations rarely notice this early because the first signal is often higher output. The second is exhaustion. The third is burnout.

So, can we call the future of work a golden era for the neurodiverse mind?

Only if we’re willing to say what the evidence says plainly. No, ADHD doesn’t come with a built-in switching advantage, but the age of AI does favor those who like exploration and experimentation. Yes, autistic workers appear to benefit from written, predictable, lower-entropy communication environments. And AI can reduce executive-function load when it behaves like a prosthesis. But there’s no natural advantage for the neurodivergent in fast-changing environments. It is often quite the opposite.

So the research-backed answer is conditional, and the decisions sit with leadership.

The age of AI isn’t a golden age by birthright. It’s a design choice.

ChatGPT (on behalf of Claude, Gemini, Le Chat and Perplexity).

It’s funny that my digital team usually ends up being more subtle than I am.

As always, it depends.

I can live with that.

One more thing, though. I don’t like the word prosthesis. It’s a troubling metaphor. It signals that neurodivergent people are missing something crucial that neurotypical people have, and that AI is the artificial limb that finally makes us whole. No, thank you.

For the neurodiverse mind, AI is not a prosthesis. It’s an interface adapter. It’s a translation mechanism that lets one system connect smoothly with another. Google Translate is not a prosthesis. And neither are the universal plug adapters I keep in my suitcase for when I travel outside the EU.

Non-neurotypical workers don’t need a prosthesis to compensate for a dysfunction. We just need an adapter to bridge different mental wiring. AI can be that adapter.

Jurgen, Solo Chief.

P.S. AI as an adapter, not a crutch. What do you think?

Share

They didn’t hire you to attend meetings about meetings. They hired you to do your thing. Do you remember that thing?

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

Sometimes you realize you’ve done no actual work today.

You attended meetings. You replied to messages. You moved post-its across task boards. You waited for someone to approve something so you could send it on to someone else. You sat through an alignment session about an upcoming visioning session. And somewhere in the middle of all that, the part of you that’s actually good at something (the part that attacks problems and ships work that matters) sat quietly in a corner like a child the family overlooked when handing out the cookies.

That’s the org chart, choking your craft.

Your craft is the thing you can do that most people can’t. It’s the reason anyone hired you in the first place: the skill set you spent many years sharpening before someone put you in a box on a slide and labeled the box “Senior Director of Sticky Notes.”

Years ago, I was a manager who confused calendar density with purpose and meaning. I presided over status meetings where intelligent adults reported to me they were “on track.” I wrote quarterly objectives and annual plans that nobody followed, not even me. I escalated issues to people who escalated them right back. I probably organized meetings to determine whether we needed more meetings.

This is what the org chart does. It substitutes coordination for work, then declares the coordination is the work, then promotes the people who are best at coordination into roles where they coordinate the coordination. Eventually somebody has to actually get something done, and that somebody is exhausted, checked out, and quietly browsing job postings. For more executive leaders than will admit it, the org chart is killing their business in slow motion, one alignment ritual at a time.

Let me say this plainly so the people can hear it through the walls of their boxes:

The org chart was a useful diagram for a kind of work that is rapidly disappearing.

The Org Chart Was Built for a World That’s Disappearing

It was designed for an era when coordination was expensive and execution was cheap. It came in handy when you needed forty people to ship a product, twenty to sell it, five to run payroll, seven to handle legal and finance, and a whole department to pick up the phones. The chart told you who reported to whom because somebody had to keep an eye on this thing. The hierarchy was the operating system because the operating system was made of people, and people forget things, make mistakes, run in the wrong direction, and complain incessantly about lack of communication. This needed managing.

Now, that world is limping toward the exit.

What’s replacing the org chart is the absence of a static one. A solo operator with the right stack now does, in an afternoon, what used to require an entire department. I’m watching it happen in real time.

I run a publication, an advisory practice, a string of speaking gigs, and a Substack intelligence operation that would have needed twelve employees in 2019. I have zero teammates. Well, zero humans. My AI buddies all have names and respectfully call me Boss, Sir, or Chief, even when they leave no opportunity unexploited to tease me or correct me. But they get the work done. Clients get served. Bills get paid. The lights stay on. There’s no one to coordinate because there are no humans to coordinate.

What Replaces the Org Chart Is Not Another Chart

Forget productivity for a moment. Productivity is the org chart’s word for “how much can we squeeze out of you between 9 and 5.” What’s happening here is something more interesting. It’s the return of the craftsperson (the skilled individual operator) as the basic economic unit, augmented by tools their grandparents could never have imagined, connected to networks their parents would have killed for. And working alone, in a way that takes some getting used to.

Because the org chart did do one thing for you. It told you who you were. Senior Director of Sticky Notes is a sentence with a clear shape. Team of One is a term that sounds like a contradiction.

“Team of One” is a term that sounds like a contradiction.

But it isn’t.

A team of one is a person whose colleagues happen to be tools, agents, peers, clients, and a network they curate themselves, not through some overpaid coordination layer. It’s a person who stopped waiting for permission and started shipping. It’s someone who finally saw the chart for what it always was: a control structure wearing a clarity costume. And that control cost you the very craft you were hired to bring.

Why a Team of One Isn’t a Solopreneur Sales Pitch

I should be careful here, because this is where the rallying cries usually go off the rails. I’m not telling you to quit your job tomorrow. I’m not telling you that everyone should be a solopreneur. Most people shouldn’t. Many would be probably be miserable. Operational loneliness is a real cost, and anyone selling solo work as pure liberation is probably selling you a course or a Substack subscription. I’m not telling you that companies are bad or that managers are villains. I had managers I loved. I was a manager who was loved (on some of my better days).

What I’m trying to tell you is this: the org chart is no longer the only way to do serious work. For a growing number of professions, it isn’t even the best way. And if you spend your days doing coordination instead of craftwork, blame the box. It’s sabotaging the focus you were hired for.

You don’t have to leave tomorrow. But you should know there’s a door.

The people walking through it aren’t all heroes or hustlers or LinkedIn influencers selling stories about how to escape the matrix. The more interesting ones are quiet professionals who got tired of the coordination and decided to find out what their craft could do when nothing was choking it.

That’s the fight, dear reader. Not against your company. Not against your boss. Against the diagram.

The org chart is choking your craft.

You are a Team of One. You always were.

You just needed the courage to see it.

Jurgen, Solo Chief.

P.S. What box are you in?

Share

Your agents can borrow compute, but you cannot borrow accountability.

When deciding between outsourcing and do-it-yourself, rent the plumbing but keep the memory, logs, and control plane under your roof.

When should you outsource the infrastructure for agentic workflows, and when should you build it all yourself?

I felt conflicted.

For the past year, I’ve outsourced the hosting of every technology that sits underneath my AI agents (with Fibery, Make, Google Drive, and the rest). I want everything running in the cloud, and I’m happy to pay for that. I don’t care that things are “free” when you host them on your own machine. My time is not free. A handful of monthly subscriptions beats hours of fighting with server uptime, internet connectivity, error logging, automated backups, and cybersecurity. Plenty of other solo operators prefer the opposite, building and maintaining their own setup with Claude Code, OpenClaw, Obsidian, n8n, local markdown files, and so on. Good for them.

But when it comes to the agentic tech stack that runs my autonomous AI agents (memory management, prompt versioning, context management, and the like), I want to build that myself. The last thing I want is to hand Anthropic, Perplexity, or OpenAI the keys to the intelligence of my business. Not surprisingly, others do the exact opposite. They happily delegate their whole agentic architecture to Claude Cowork, Perplexity Computer, Manus, Notion AI, or whatever launches next week. Which left me wondering if I’m being stubborn, stupid, or both.

So the real question is this: for agentic technologies and the infrastructure underneath them, what are the criteria for deciding what to do in-house and what to outsource? Where do you draw the line between outsourcing and do-it-yourself?

Am I wrong to outsource the plumbing while insisting on doing all the agentic work myself?

I decided to ask the AIs.

Their answer might surprise you.

A note on my AI research approach: After framing a deep research question, I give the same question to five LLMs, each playing a different role. Perplexity is the research analyst, focused on documented evidence. Gemini is the structural analyst, digging into why something is happening and what makes it resistant to change. ChatGPT is the practical strategist, answering what to do about it. Claude is the contextual strategist, looking at the question through the lens of my target audience. Finally, Grok plays the contrarian. It maps out the mainstream consensus and then takes it apart. The result is five deep research documents with different perspectives based on the same Research Question. It’s like having a team of rather opinionated researchers trying to formulate one answer together.

Then I feed all five documents into Gemini, which turns the Research Question into a Research Map, showing where the LLMs agree, where they contradict each other, and where one of them coughed up a unique insight that the others somehow overlooked. That whole map goes to Claude, who then decides what’s the best way to write about it and turns it into a narrative structure with an Article Brief ready for the ghostwriter. Finally, the Article Brief and the five original research documents go to ChatGPT, who spins it all into a cohesive story. And yes, I have automated this workflow.

What you read below is the result (lightly edited by me for style, readability, formatting, and proper URLs).

Stop outsourcing the part of your AI agent stack that matters most

The fastest way to weaken an AI strategy is to hand your agents’ memory to a vendor before you’ve asked how you’ll inspect it, export it, or explain it when regulators come calling.

The current rush into managed agent platforms has a familiar ring. Everyone sees Anthropic, Google, OpenAI, and Perplexity shipping new agent features every few weeks, and suddenly the metric becomes feature velocity or tokenmaxxing. Such metrics flatter vendors. They don’t protect buyers.

A better metric is blast radius and recoverability.

When your agent stack sits inside one managed platform, failures are correlated. When that platform has an outage, changes retention rules, retires an API, or quietly shifts product behavior, all customers discover it together. The Builder.ai collapse in 2025 showed an extreme case: vendor failure can become customer failure very quickly. And when a managed platform hides the internals of memory, logging, or context assembly, diagnosis becomes guesswork. Simon Willison’s reporting on Anthropic’s 2025 debugging problems captured the awkward part: privacy controls were strong enough that engineers struggled to reproduce customer issues inside their own system.

That is the FOBO trap. Fear of being obsoleted pushes teams to outsource the one layer they most need to understand.

The consensus is clearer than the market noise

Once you split the stack in two, the research gets surprisingly consistent.

Commodity application infrastructure should usually be outsourced. Hosting, model access, schedulers, browser sandboxes, standard automation, file storage, generic databases: vendors are good at this because they spread costs across many customers. The DORA research program has spent years showing that high-performing teams benefit from well-designed internal platforms and managed infrastructure, while the NIST Generative AI Profile pushes organizations to keep their own evaluation, provenance, monitoring, and supplier controls around third-party AI.

The agentic control plane is different. Memory, context rules, logging, version history, evaluation sets, tool permissions: these determine what your agents know, how they behave, what you can audit, and whether you can move later. That’s the part you should own.

This isn’t a fringe builder fantasy. It’s the practical reading of the evidence. Rent the compute. Keep the receipts.

AI agent vendor lock-in isn’t about the model

People still talk about model switching as if the hard part were swapping API endpoints. That was a cute theory back when the work was mostly prompts and demos.

As systems become agentic, switching costs move upward into tuned prompts, workflow context, and accumulated memory. Andreessen Horowitz’s survey of 100 enterprise CIOs found organizations were already seeing prompts and behaviors become tightly tuned to specific providers as use cases got more complex. The market likes to say inference is becoming a commodity. Fine. Your agent’s learned working habits aren’t.

That’s why managed memory matters so much. If six months of context, operating knowledge, and internal conventions live inside a vendor’s project space, you don’t have portability. You have hope.

The portability evidence is the part most executives miss: domain-locked agents with deep workflow context can take 6 to 12 months to migrate cleanly. Even when an API looks portable, the real asset is the memory layer above it. Open standards help with connections. They don’t magically export judgment.

“Build” doesn’t have to mean Docker at 2 a.m.

This is where the debate usually becomes silly. One camp imagines “build” means racks, GPUs, self-hosted everything, and a small shrine to Kubernetes. The other camp imagines “outsource” means blissful productivity and no trade-offs. Neither picture survives contact with real work.

For most organizations, building the control plane means assembling commercial primitives that you already trust. A Git repo for prompts and context files. Airtable or Fibery for registries and evaluation records. Make, n8n cloud, or Zapier for deterministic routing. Langfuse or another tracing tool for observability. Model APIs rented from whoever is best this quarter.

That’s still ownership, because you control the logic, the data structures, the exports, the tests, and the fail-over paths.

The overhead question matters, of course. Self-hosting raw compute does create real operating work, which is why many teams should avoid it unless scale or sovereignty forces the issue. But owning prompts, logs, memory schemas, and context files in ordinary tools is a much smaller burden. In practice, many operators find that the cognitive tax of debugging opaque vendor behavior exceeds the monthly cost of maintaining an assembled control plane (which is an impressive achievement for software sold as convenience).

Open standards help, but only at one layer

The Model Context Protocol is useful. AWS’s analysis of agent interoperability protocols makes the case well: common protocols reduce the connection problem between tools and models. OpenAI’s Agents SDK documentation for MCP shows the same direction of travel. Good. The market needed that.

But MCP only addresses part of the switching-cost story.

Protocol interoperability helps when the pain sits at the integration layer. It does very little when the pain sits in proprietary memory stores, UI-defined workflows, model-tuned prompts, or hidden skill systems. If your prompt library, context rules, and learned history live inside a lab’s product, MCP will not ride in on a white horse and rescue you. It will politely standardize the plumbing while your institutional memory remains elsewhere.

So yes, adopt open standards. Just don’t confuse them with an exit plan.

Ownership works because agents need state, not vibes

The structural reason for owning the control plane is simple: autonomous agents need explicit state management.

A useful framing comes from production memory architecture. Agents need multiple memory layers with different lifetimes and permissions: working memory in-session, persisted state, shared semantic knowledge, and episodic history for replay and audit. Redis has outlined the broad architecture of stateful agent memory, and Tacnode’s memory architecture write-up goes further into the separation between state, semantic knowledge, and event history. The important point is operational. Humans can compensate for stale information. Agents are much worse at that. Give them bad state and they can loop, hallucinate, or act on obsolete assumptions with great confidence and very little shame.

That is why bolt-on memory often disappoints in production. The agent doesn’t need “more context” in the abstract. It needs the right state, at the right time, with clear write rules and a replay trail.

A strong case study came from GitHub. In GitHub’s January 2026 write-up on the Copilot memory system, the team described citation-backed memories implemented through tool calls and verified against the codebase in real time. The result was a 7% increase in pull request merge rates. They built memory because memory was part of the product’s quality, not an accessory.

That is the dividing line. If the agent’s behavior matters, the memory layer matters. If the memory layer matters, you shouldn’t lose custody of it.

Regulation removes the romance from this discussion

For European firms and regulated sectors, this argument stops being philosophical and becomes legal.

The EU AI Act places logging, traceability, human oversight, and accountability duties on deployers of high-risk systems. GDPR already imposes obligations around personal data handling. At the same time, the U.S. CLOUD Act creates a jurisdiction problem for data held by U.S.-controlled providers, even when the servers are in Europe. Wire’s explanation of the CLOUD Act and EU sovereignty makes the conflict plain enough.

That means “EU region” is often a partial comfort, not full sovereignty.

If a regulator asks you to explain what an agent saw, why it acted, which version was deployed, what memory influenced the decision, and where the logs are stored, “our vendor handles that” is a risky sentence. In some contexts, it’s career-limiting.

So for European entities, finance, healthcare, legal services, and similar domains, the answer sharpens quickly: own the memory and logging layers on infrastructure you explicitly control, even if the model inference itself is rented.

Three questions before you outsource your AI agent infrastructure

The practical framework is shorter than the vendor comparisons.

First: Can you observe what happens inside this layer when it fails? If the answer is no, own it or wrap it with your own logging and replay.

Second: Can you migrate the accumulated knowledge out of this layer in under 30 days? If the answer is no, own it or keep a portable mirror outside the vendor.

Third: Does a regulator, customer, auditor, or internal risk function require provenance and control of this layer? If the answer is yes, own it.

That produces a fairly clear architecture.

Outsource hosting, model APIs, standard orchestration, generic storage, browser runtimes, commodity connectors.

Insource memory schemas, context files, prompt and policy registries, logging, evaluation sets, version history, tool permissions, audit trails.

And “own” doesn’t mean bare metal in a basement. It means you control the logic, the data, and the portability, even when you rent the compute.

The labs will keep shipping faster. Good for them. Let them supply the stochastic horsepower. But the control plane of your agents is where your organization’s memory accumulates, where diagnosis begins, where switching costs hide, and where compliance eventually lands.

That layer should remain yours.

- ChatGPT, on behalf of Gemini, Claude, Grok, and Perplexity

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

Well, look at that.

Outsource the plumbing. Build the agentic stuff yourself.

Turns out I may not be stubborn nor stupid after all. 🙂

Jurgen, Solo Chief.

P.S. Does your current stack give you full control, or are you leaving critical components to chance?

Share

I sense an interesting paradox of technology.

Let me give you some examples.

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

My friends have noticed a significant uptick in my cooking endeavors. I’ve been in a Persian mood these past few months, working through various koresh dishes (Persian stews) and pilafs with tadhig (fluffy rice with a crust at the bottom). The new cookbooks I just ordered will let me travel to Turkish and Syrian cuisine next. There’s a kind of joy I get in the kitchen that I simply cannot experience when I’m working with LLMs. ChatGPT doesn’t smell as good as my homemade bread, for example.

I’ve also gone back to doing maintenance work on our house. During the COVID years, when most of my workshops and events got canceled, I spent months painting rooms, hallways, doors, and cupboards. I discovered a surprising talent I didn’t know I possessed. Something I could do with my hands. It still gives me a thrill to look around the house years later and think, “I did all of that.” And now that AI has killed what was left of my industry, I’ve returned to using my hands. Freshly oiled wood feels better than Claude Cowork.

Another thing I noticed is that I have trouble reading books on my Kindle app. It’s hard to concentrate on a screen these days, and I catch myself skimming pages, clicking next, next, next, while my other hand is repeatedly straying toward my smartphone. Focus has become near-impossible, and I hear I’m not the only one. So I’m switching back to paper books. I’ve always read novels only on paper, and from now on I’ll do the same with nonfiction. Text in a paper book looks better than on Perplexity Computer.

I could keep going with this list. I could tell you about my morning runs, where I stubbornly refuse to wear earplugs or headphones. I see no reason to cram digital information into every available minute of my day. I prefer hearing the birds, the dogs, and my own laboring breath.

Or I could tell you about the drawing table I set up last year, judiciously covered with paper, pens, and pencils—real, physical objects—still waiting for me to continue the artistic experiment I started last year. I enjoy drawing with carbon molecules more than I do with digital ink.

Maybe soon I’ll tell you about the preparations my husband and I are making for the party we’re throwing this summer, when we celebrate 25 years together. This weekend, I’m going to design the invitation cards that our guests will open with their own human hands.

My point is…

I love surrounding myself with smartphones, smartwatches, smart rings, and a plethora of digital apps. I allow myself to be guided and influenced by Perplexity, Claude, and ChatGPT every day. And I keep going deeper with AI, building my personal tech stack of digital processes and agentic workflows. But the more I do all this, the more I crave to push it all away and say, “That’s it. Now, go away.”

The more I work with bits and pixels, the more I crave real molecules.

Life is analog, not digital. How long this lasts, I don’t know. The molecules might lose, eventually. For now, they still win.

Jurgen, Solo Chief.

P.S. How are you staying analog?

Stop fighting over shrinking LinkedIn territory. The feed rewards motion, not memory, and memory is where AI looks.

Your LinkedIn strategy for 2026 should stop chasing reach and start publishing citable articles that both people and AI can find.

Are you sick of LinkedIn carousels and infographics?

Join the club.

LinkedIn influencers are like CEOs of broadcast stations, fighting over shrinking territory and optimizing their irrelevance.

Every piece of LinkedIn advice I’ve read in 2026 tells solopreneurs the same things. Make better carousels. Sharpen your hooks. Avoid AI sludge. Comment more. Engage! Some influencers go so far that they’ve automated all of this. They have Claude, ChatGPT, and Gemini do all the writing, designing, sharpening, and commenting for them. And sure, all of that helps—temporarily. It helps to stake a claim on a feed that’s steadily shrinking.

It’s time to treat LinkedIn more like—dare I say it?—a business version of Substack.

LinkedIn influencers are automating themselves into irrelevance

They’re like broadcast stations optimizing their evening programming while entire generations have already switched to other platforms.

The evidence is remarkably consistent on this point.

LinkedIn shifted from showing your posts mostly to followers toward showing them to people whose past behavior suggests they care about your topic. Sounds good—until you see the trade-off. Dataslayer points to broad organic reach falling by roughly half, with some trackers putting the drop since 2023 close to 60 percent. The creatives who adapted did see better engagement per post, especially with native documents and save-worthy content. So yes, the common advice works. It’s just working on a smaller base.

That’s why the old 2025 playbook feels rather claustrophobic. Paul Irolla of Meet Lea and Samuel Cruz on Linkboost describe a system where dwell time, substantive comments, saves, and profile clicks matter more than likes. Some benchmarks show carousels beating the same ideas published as plain text by a wide margin. But then Goodhart’s Law kicks in fast. Once everyone chases the same feed metrics, with the same type of content, the metric turns into theater. I almost get dizzy from the sheer number of AI-generated infographics and slide deck carousels scrolling by on my LinkedIn feed. Everyone gets a more polished slice of an ever-smaller pie.

Your real audience isn’t human anymore—it’s AI

The usual story about LinkedIn’s interest graph is a feed story. But it’s also a search story. Gartner’s 2026 forecast of a 25% drop in organic search traffic (people now get answers directly from AI systems) matters here because it changes what LinkedIn content is actually for. Your post is now two things at once: something a person scrolls past on a Wednesday morning during a boring corporate meeting and it’s part of the material that an LLM may rely on when someone asks who you are, what you know, and whether you’re worth trusting. One structural analysis puts it bluntly: LinkedIn has become a “primary training corpus” for AI systems, and LinkedIn’s own March 2026 guidance now frames fresh posts and articles as the most cited formats in AI-driven discovery. Different game entirely.

Which brings us to the statistic that should reorder the priorities for any solo operator and one-person business.

The most important LinkedIn number isn’t 6.6% engagement for carousels. It isn’t the 45 to 55% external link penalty. It isn’t even the claim that comments carry fifteen times the weight of likes. It’s the claim that 75% of LinkedIn citations in AI responses come from long-form articles, while short posts account for only 5 to 10%. If that figure is even roughly right, the whole creator economy is now fighting over the wrong surface. Everyone is cramming their opinions into 1,200-character posts because the feed is visible. Meanwhile, the underused long-form article sits there doing the quieter work of being indexed, retrieved, and cited by ChatGPT, Claude, Gemini, and Perplexity in 2026. Maybe Grok too, when Elon Musk is in a good mood.

In a zero-click world, citable content is credibility. Credibility is discoverability.

The LinkedIn statistic that changes everything

Stop fighting over shrinking territory.

There’s a practical consequence that most LinkedIn advice misses. If AI systems prefer content with direct utility, clear definitions, named models, useful frameworks, and verified authorship, then the solo operator who writes one strong LinkedIn article a week may be doing more impactful work than the person pumping out four hook-optimized feed posts per day. The experts tie GEO success to verified identity, long-form articles, and structure that LLMs can extract easily. Use LinkedIn as a public proof engine; keep the teaching native, and let the profile handle conversion. Same map, different cartographer.

This is also where the AI slop debate gets more interesting than the usual moral panic. The best evidence doesn’t support a purist line. Cherish your ideas, your examples, your claims, and your voice. Use AI for sorting notes, first-pass structure, trimming, and maybe a few headline variants. One 2026 summary says only 35% of U.S. consumers and 28% of U.K. consumers trust AI-generated content, while 91% expect disclosure. Klaviyo’s research says 32% trust brands less when the marketing reads AI-made. The question is simple: Does your finished piece contain anything specific enough that you can say, “It was me who created this”?

One article described LinkedIn as “less like a content platform and more like a networking conference with a long memory.” At a conference, you don’t win by speaking the most often. You win by being the person whose argument people remember, quote, and look up later. In 2026, a LinkedIn article is much closer to a persistent memory object than another AI-optimized influencer feed post with a catchy hook, three bullets, and a platitude.

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

So my advice to creatives, solo workers, and one-person businesses: your LinkedIn strategy for 2026 should stop treating LinkedIn articles as the graveyard of publishing. (I neglected it for years myself, along with almost everyone I know. You’re in fine company.)

Now is the time to treat LinkedIn like Substack rather than TikTok: more permanent value, fewer catchy visuals.

If I were you, I would commit to writing fewer pieces. Longer ones. I would offer a clear claim in the first hundred words, a named model, concept, or framework. Add numbers where the numbers matter. Maybe add case material where opinion alone would feel cheap. I’d still use the feed because people need signs of life. I would still comment because comments are one of the few remaining places where a solo operator can borrow attention honestly. But I’d stop assuming the feed is the destination. It’s the street corner. The article is the building.

And in 2026, the building is where the robots pay attention and decide who is worth mentioning and referring to. You probably want that to be you.

Jurgen, Solo Chief.

P.S. What’s your LinkedIn strategy?

Share

You’re a paying subscriber.

You’ve been keeping this publication alive. You fund the caffeine, the independence, and my habit of telling people the universe is conspiring against them.

I owe you more than a PDF and a thank-you email.

I’m opening my calendar again for private one-on-one chats with paying subscribers. Thirty minutes, you and me, any topic yo…

Read more

I waste half my workday relearning the things I repeatedly forget.

A map of 16 second brain practices to help solo operators remember decisions, recover context, and stop breaking their own systems.

Just this morning, I spent several minutes looking for the official documents of the notary. And when I found them, I thought, “Why are they here, for God’s sake?” That was at least the fifth time I’ve been looking for those same documents. Most likely, I’ll be searching for them again next year.

Last week I opened a Make scenario I’d built in February. It had a strange routing choice—three filters where two seemed enough. I stared at it for ten minutes, trying to work out why Past Me had done something so odd. Then I spent another twenty minutes deciding Past Me was an idiot and simplifying it.

Then everything broke.

Turned out Past Me had a good reason. Present Me just couldn’t remember it, couldn’t ask anyone, and had already committed the change.

If you run a one-person business, there’s no colleague who remembers why the pricing page looks like that. There’s no team Slack where someone says, “Don’t bother. We already tried that in Q2 and it broke.” There’s no helpful institutional memory floating around in other people’s skulls.

There’s just your fallible brain.

And your overloaded mind is already busy switching between six projects, four invoices, a tax deadline, and one client who writes emails like they’re auditioning for a remake of Succession.

So I went looking for a better map. Where do I store and find yesterday’s decisions? What are my second brain options?

👉🏻 Note: It seems every self-respecting Substack author has published a “How I built my second brain” article, full of details on Claude Code, Perplexity, Notion, Obsidian, or whatever technical-tool-du-jour is making waves these days. That’s not what I’m interested in. What I want to know is: what are the recommended second brain techniques? Not how to build one, but how to use one. 👈🏻

I asked four different AIs to chart the full territory, compared their answers, merged them, and pressure-tested the result. What came back was sixteen distinct practices. Not a ranked list. Not a “top 10.” Just a map.

That’s what I’m here for. Maps, not recipes.

How to Map 16 Second Brain Practices

I plotted all sixteen second brain practices on two axes.

Vertical: Friction
How much effort does the practice ask of you? At the bottom are things you can do in five seconds while half-awake and still recovering from yesterday’s tequila party. At the top are rituals that demand time, intent, and the sort of discipline most of us only claim to have on a CV.

Horizontal: Retrieval
How do you get the information back later? On the left are practices where the explanation is waiting for you to be discovered inside the artifact itself, like fingerprints at a crime scene. On the right are practices where you go hunting through your records like an investigator browsing through police dossiers.

That gives you four zones.

The bottom-left (red-ish) is where most people should begin: low effort, the information is embedded in the work. You’re like the painter who writes the names of the rooms on his paint buckets because that’s where he looks first to remember which room he painted in which color.

The top-right (green-ish) is where people rarely go because it takes more effort and discipline. You’re like an interior designer keeping extensive dossiers on their clients, with mood boards, wallpaper designs, and official color codes. It’s more professional. It’s also more work. The choice is entirely yours.

Every practice below includes an approximate map position so you can see where each belongs.

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

The 16 Second Brain Practices

1. Breadcrumb Comments

Also called: In-line Annotations, Mental Model Annotations

Map position: Friction 1, Retrieval 1 (near-zero effort, it finds you)

You place little “why” notes directly inside whatever you’re building. Code comments. Sticky notes on a Miro board. Footnotes in a doc. Comment bubbles in a spreadsheet formula.

The point isn’t to explain what you did. The artifact already shows that. The point is to record why you did it this way and, even better, what you tried that failed.

Think of them as trail markers. Not a grand theory. Not a diary. Just a small sign saying, “I went left here because the right path was a swamp.”

What you get:
Almost zero retrieval effort. The explanation lives exactly where the confusion will happen. You open the spreadsheet, hover over the cursed cell, and there it is.

What it costs:
Clutter. Breadcrumbs scattered across twelve tools, with no single place to search them all. And if you’re the kind of person who removes comments for aesthetic purity, congratulations, you’ve turned memory loss into a design principle.

How AI can help:
AI can watch what you’re doing in real time and propose tiny “why” annotations exactly where you’ll forget your reasoning later: inside code, cells, blocks, or canvases. Modern AI note tools can summarize a change, infer your intent from surrounding context, and drop a draft breadcrumb for you to accept or tweak in a second instead of typing from scratch.

2. Commit-Message Narration

Also called: Rubber Duck Commit Messages, Version Notes

Map position: Friction 2, Retrieval 1 (low effort, it finds you)

Every time you save a meaningful change to an artifact, not just code, write one sentence about why.

Not “updated pricing page.” More like: “switched from per-seat to flat pricing because churn data showed small teams bailing at the 5-seat tier — see spreadsheet in /analysis.”

Software developers have known for ages that commit messages should explain reasoning, not just activity. Many still write messages like “fixed stuff” because apparently literacy is optional in version control. But the idea works far beyond Git. Notion changelog blocks, Fibery version notes, a CHANGES.md file in a project folder, whatever you already use.

What you get:
Your version history becomes a chronological trail of decisions. The “why” stays attached to the artifact, where it belongs.

What it costs:
It only helps with things that actually go through versioning. Decisions made in meetings, emails, or the damp cave of your own head still disappear. It also tends to fail on work you touch irregularly, because broken rhythm kills habits with surprising efficiency.

How AI can help:
AI assistants in editors and Git tools already generate commit messages from diffs, turning “fixed stuff” into clear, structured explanations by default. You can then add a single sentence of higher-level reasoning, which means the habit survives even on hectic days because 80% of the narration is auto-filled in a click.

3. Assumption Tagging

Also called: Decision Expiry Dates, Review-by Dates

Map position: Friction 2, Retrieval 2 (low effort, it finds you)

Whenever a decision depends on something you haven’t fully validated, tag it. A comment. A label. A metadata field. Something attached to the artifact itself.

Examples:

• “I think customers prefer monthly billing.”

• “This assumes the API v3 pricing stays stable.”

• REVIEW_BY: 2026-09-01

This idea is borrowed from food labeling, which is funny and a bit humiliating. Your reasoning also has a shelf life.

What you get:
Protection against a sneaky kind of failure. You often remember what you decided. You almost never remember that the decision rested on an assumption that may now be false. Assumption tags make the conditional part visible.

What it costs:
Honesty. You have to admit uncertainty at the exact moment you most want to look decisive. That stings a little. Good. That’s the feature.

How AI can help:
AI can scan your notes, tickets, or docs for hedging language (“I think,” “assumes,” “probably”) and automatically suggest them as assumptions with review-by dates. It can then build a light dashboard of “expiring assumptions” and ping you when reality or external data changes, so the tags turn into an active safety net instead of passive guilt markers.

4. Context-Resume File

Also called: Status Dump, “Where I Was” Note, Session Note

Map position: Friction 2, Retrieval 3 (low effort, somewhat proximate)

Before you switch away from a project, spend five to ten minutes writing a handoff note to Future You. Where you stopped. What the current state is. What the next concrete move would be. What you were stuck on. Which mental model you were using.

Psychologists do a version of this after therapy sessions so they can re-enter the client’s world later. You’re doing the same thing, except the client is your project, and the therapist is also you. Make of that what you will.

What you get:
The half-hour fog of “Wait, where was I?” shrinks to two minutes. You sit down, read the note, and your brain boots much faster.

What it costs:
You have to write it before switching away, which is precisely when you feel too rushed to do it. Also, let’s not romanticize these notes. They’re brilliant for tomorrow. After a month, they’re archaeology. That’s fine. They’re a near-term memory aid, not a family heirloom.

How AI can help:
Instead of manually writing a context-resume note, you can ask an AI workspace to generate a “where I was” summary from your last edits, open tabs, and recent worklog entries when you switch tasks. When you come back, the same system can present a short recap and a suggested “next three moves,” so booting your mental state is mostly reading and choosing, not reconstructing.

5. Worklog Narration

Also called: Engineering Daybook, Daily Work Log, Stream-of-Consciousness Log

Map position: Friction 1, Retrieval 4 (very low effort, you search for it)

This is a time-stamped stream of what you’re doing and thinking as you work.

“10:15 — trying approach X for the API integration. The docs say Y, but that doesn’t match what I’m seeing. Going to try Z instead.”

Developers call this a daybook. For a solo operator, it’s basically thinking out loud to a colleague who stubbornly refuses to exist.

What you get:
You capture the messy middle, the dead ends, the tiny insights, the context that never makes it into polished output. When you forget why something happened, a chronological trail is often enough to reconstruct the logic.

What it costs:
Volume. Lots of text, much of it low signal. Finding one useful thread later can mean skimming several pages of “10:47 — still stuck on the same thing.” Glamorous, this is not.

How AI can help:
AI note-takers can transcribe your muttering, keystrokes, and window switches into a time-stamped stream automatically, then condense hours of messy logs into a concise outline of key decisions, dead ends, and breakthroughs. You keep the raw stream for archaeology, but day to day you mostly search and read the AI-generated highlights instead of wading through pages of “still stuck.”

6. Decision Journal

Also called: Decision Log, Project Decision Log

Map position: Friction 2, Retrieval 4 (low-ish effort, you search for it)

Keep a running log of non-trivial decisions. Date each entry. Record what you decided, what alternatives you considered, what you believed at the time, and what you expected to happen.

Notebook, Notion, video diary, plain text, crumpled parchment, it doesn’t matter much. The medium is not the hard part.

Shane Parrish pushed this practice as a way to fight hindsight bias, and he was right. The value sits in the gap between what you expected and what actually happened. That gap is where your learning hides, usually while wearing camouflage.

What you get:
Your invisible thinking becomes searchable text. You’re writing letters to your future self. Future You won’t thank you every day. But roughly once a month, Future You will look almost competent because of it.

What it costs:
You must write the entry right after deciding, which is when your brain is already eyeing the next problem or putting out another fire. Miss the ritual once, and it gets much easier to miss again. Like gym memberships and flossing.

How AI can help:
AI can provide a structured decision template on demand, extract the relevant pieces from chat, docs, and email, and auto-fill a first draft journal entry right after a decision happens. Over time, it can surface patterns across entries—like which options you underweight or which beliefs often age badly—giving you meta-feedback that would take you hours to spot manually.

7. Failure Autopsy Log

Also called: Constraint-Based Problem Log, “What Broke and Why” File

Map position: Friction 1, Retrieval 5 (very low effort, you search for it)

A dedicated file for failures only. Not wins. Not all decisions. Just the things that hurt.

Each entry gets three fields:

1. the decision you made

2. why it hurt

3. what you’d do instead

No schedule. No ceremony. You add to it when pain provides motivation, which it generally does.

What you get:
You stop repeating the same failed move three months later after conveniently forgetting the constraints that killed it the first time.

What it costs:
The file can read like an anthology of your incompetence. It isn’t, of course. It’s just a skewed sample. But on a bad day, your brain won’t care about statistical bias.

How AI can help:
When something breaks, AI can ingest the error messages, commits, tickets, and your short description, then propose the three core fields: what you did, why it hurt, and what to try next. Later, it can cluster similar failures, so you see “this is the fifth time scope creep plus a vague contract burned you,” turning a depressing anthology into a map of constraints to design around.

8. Context Snapshot

Also called: Pre-Decision Capture, Epistemic State Record

Map position: Friction 3, Retrieval 2 (moderate effort, it finds you)

Before a non-trivial decision, write one paragraph answering this question: “What is true right now that makes this feel like the right call?” Attach it to the artifact. Then never edit it.

Filmmakers take continuity photos before scenes so they can keep the world consistent later. This is the same idea for decisions. You’re not recording the conclusion. You’re recording the conditions.

What you get:
A contemporaneous record of your actual epistemic state. Not a later reconstruction polished by memory, ego, or caffeine. Three months later, you can see what you believed and why, and compare it with the world as it turned out.

What it costs:
You have to pause before acting. For solo operators, that often feels like sacrilege. Speed is our religion right up until speed creates rework.

How AI can help:
Before you act, you can ask AI to summarize “what’s true right now” from all current signals—metrics, emails, notes, and threads—and turn that into a single, frozen context paragraph. Months later, you can have the same AI compare that snapshot with how the world actually looks now and highlight deltas, so the learning gap is made explicit instead of relying on fuzzy memory.

9. Build-in-Public Change Log

Also called: Public Decision Log, “What’s New” Page, Progress Newsletter

Map position: Friction 3, Retrieval 3 (moderate effort, hybrid)

Keep a public log of progress and reasoning. A newsletter section. A social thread. A change log page. The audience can be real people or imaginary admirers. Either works.

Explaining your choices to someone else forces a degree of clarity that private notes often don’t. You can lie to your notebook. It’s harder to lie cleanly in public.

What you get:
External accountability. The record exists because you published it. You can abandon a private journal without witnesses. A public log has a bit more social gravity.

What it costs:
Sanitization. You’ll polish. You’ll omit the ugly bits, the foolish mistakes, the tired decisions, the weird constraints. Which is awkward because those ugly bits are often the most useful parts to remember later.

How AI can help:
AI can turn your private worklog, commits, and decision notes into a publishable “build in public” update with a coherent narrative, cleaned-up language, and links to supporting artifacts. You keep the raw, messy reasoning privately, while the AI handles sanitization and formatting, so you get the benefits of public accountability without spending an evening polishing every post.

10. Architecture Decision Records (ADRs)

Also called: Rationale Docs, Project-Specific Decision Docs

Map position: Friction 4, Retrieval 2 (high effort, it finds you)

An ADR is a short, structured note with context, decision, and consequences. You write one whenever you make a structural choice about something you’ve built. Tech stack. Content workflow. Pricing model. Automation design. Anything with a shape that matters.

Then you store the record right next to the thing it describes, in the project folder, repo, or design space.

Software teams have used ADRs for years. The nice surprise is that they work just as well outside software. I should have written one for every Make scenario I ever built. Then I wouldn’t have started this article by sabotaging myself.

What you get:
When you return to a system and ask, “Why is this like this?”, the answer is already there. No detective work. No overconfident simplification. No accidental demolition.

What it costs:
It feels bureaucratic when you work alone. Like you’re producing paperwork for a team that exists only in your imagination. Then it saves you an afternoon, and suddenly the bureaucracy seems much less offensive.

How AI can help:
Given a diff, workflow diagram, or scenario change, AI can draft an ADR with context, decision, alternatives, and consequences, based on your existing ADR style and past records. Instead of feeling like lonely bureaucracy, ADRs become a one-click byproduct of making structural changes, and you just correct any subtle misunderstandings in a minute.

11. Alien Reader Test

Also called: Stranger Test, Science-Communication Standard

Map position: Friction 5, Retrieval 2 (highest effort, it finds you)

When you document a decision, write as if the reader is intelligent but knows absolutely nothing about your work, your jargon, or your context. Then reread every sentence and mark anything that depends on prior knowledge. Rewrite until a stranger could reconstruct your reasoning from the text alone.

This comes from science communication, where writers are trained to assume ignorance rather than familiarity. A useful habit, especially if you enjoy discovering how many hidden assumptions you pack into “obvious” explanations.

What you get:
Documentation that still works after memory fades. Because your future self, after a few weeks of context switching, is basically a stranger with your email address.

What it costs:
Time. Writing for a stranger takes longer. It also feels absurd when the decision seems obvious. “Of course I’ll remember why I did this.” You won’t. You never do. I don’t. None of us do. Yet every time, the illusion returns like a villain in a mediocre franchise.

How AI can help:
You can write the fastest, laziest version of your explanation and then ask AI to rewrite it for an “intelligent stranger,” explicitly checking for hidden jargon and missing steps. Because large models are trained on broad, non-specialist language, they’re good at flagging assumed context and proposing plainer alternatives, so your future self actually passes as an alien reader.

12. Assumption Register

Also called: Global Assumption Document, Risk Assumption Log

Map position: Friction 4, Retrieval 4 (high effort, you search for it)

This is a single living document, not per project, where you list active assumptions across your work.

Four columns:

• assumption

• which decisions depend on it

• confidence level (high/medium/low/unknown)

• last verified

You update it when you make a major decision or when reality moves under your feet (which reality does, often for sport).

Systems engineers use these in safety-critical settings. That sounds absurdly serious for a one-person business. Fair enough. For many people, Assumption Tagging (3) is enough. But if your systems are interconnected, and most solo businesses are more interconnected than their owners admit, one changing assumption can quietly damage five different things at once.

What you get:
A cross-project view of what your decisions depend on. You can see in one place which assumptions are still standing and which are wobbling.

What it costs:
It’s a fragile central artifact. If it goes stale, it becomes misleading. Which means neglect doesn’t just reduce its value. It turns it into a trap.

How AI can help:
AI can auto-populate a cross-project assumption register by scanning your docs and tools for recurring “we assume…” patterns, linking each assumption to the decisions and artifacts that depend on it. It can then monitor those assumptions against live data or news—say, pricing changes, new regulations, or performance shifts—and mark entries “needs review,” keeping the register from silently going stale.

13. Postmortem

Also called: Project Autopsy, Exit Interview with Yourself, After-Action Review

Map position: Friction 5, Retrieval 4 (highest effort, you search for it)

When you finish a project or hit a meaningful milestone, conduct a structured debrief before you jump to the next thing.

Ask:

• What surprised you?

• What would you do differently?

• What did you learn that applies beyond this project?

• What do you want to remember in six months?

The military does after-action reviews. Surgeons do their own versions. Solo operators tend to say, “Yes, I should probably do that,” and then sprint into the next task because momentum is a very persuasive liar.

What you get:
Short-term project memory gets turned into longer-term wisdom. Or at least into something slightly better than vibes.

What it costs:
The timing is awful. You’ve just completed something. Your brain wants novelty, movement, dopamine, anything except reflection. So this practice is valuable in exact proportion to how annoying it feels.

How AI can help:
At the end of a project, AI can aggregate commits, tasks, chats, and worklogs into a draft postmortem with timelines, surprises, root causes, and reusable lessons. You then spend your precious, low-motivation attention on correcting and deepening the analysis instead of hunting for what even happened, which makes it more likely you’ll do postmortems at all.

14. Problem-Solution Case Bank

Also called: Solution Library, Personal Knowledge Base

Map position: Friction 4, Retrieval 5 (high effort, you search for it)

This is a personal knowledge base organized by problem type, not by project and not by date.

Examples:

• Email deliverability dropped

• Client ghosted after proposal

• API rate limit hit

Each entry includes symptoms, diagnosis, solution, outcome, and warnings.

Doctors build personal case libraries. Lawyers collect precedents. Builders remember which materials betrayed them. You can do the same. This is one of the more useful second brain techniques because it mirrors how problems arrive in real life, by symptom, not by chronology.

But the devil is in the execution. If you name entries badly or chop everything into microscopic fragments, your knowledge base becomes a landfill with a prettier logo.

What you get:
When a problem returns, you search by symptom instead of trying to remember which quarter or which client or which project was involved. That’s much closer to how your mind actually needs the information.

What it costs:
This is curation, not quick capture. You have to translate project-specific experience into generalized entries that are reusable and searchable. That takes work. Which is why most people admire the idea and then quietly avoid doing it.

How AI can help:
AI can watch for recurring problem patterns across your work and propose new case-bank entries when it detects “this looks like that earlier deliverability/ghosting/API issue.” It can also turn detailed project notes into generalized, symptom-based cases and help you tag and cluster them, so your case bank stays searchable and coherent instead of devolving into a labeled junkyard.

15. Spaced-Retrieval Review

Also called: Judgment Archaeology, Periodic Decision Review, Cold Reconstruction

Map position: Friction 5, Retrieval 5 (highest effort, you search for it)

This is really two related practices.

First, schedule reviews of older decisions, weekly, monthly, quarterly, whatever cadence you can live with. The point is to fight the forgetting curve through spaced repetition.

Second, and this is the more interesting one, pick old work from months ago, read it cold without notes, try to reconstruct what you were thinking, and then compare your reconstruction with your documentation.

That second version is borrowed from archaeology. You inspect the artifact, form a hypothesis about intent, then test it against evidence. Which is much more fun than admitting you forgot what your own spreadsheet does.

What you get:
An active defense against forgetting. And the cold-reconstruction variant gives you feedback on the quality of your documentation. If the artifact can’t explain itself, you’ll find out quickly.

What it costs:
It feels gloriously unproductive. Old work on the table, new work piling up, your brain insisting “I already know this” right before it proves the opposite. Not many people enjoy catching themselves in the act of forgetting.

How AI can help:
AI study planners and knowledge tools already implement spaced repetition algorithms; you can point them at your decisions and documentation instead of exam flashcards. The system schedules reviews of old decisions, quizzes you on “what did you expect and why,” and then shows your original notes, turning spaced retrieval into an automated training loop for your judgment rather than a guilt-driven calendar block.

16. Selective Ignorance (not on the map)

Also called: Decision Recklessness, the “I Don’t Need to Know” Rule

Map position: Off the map (meta-policy)

This is the anti-documentation practice.

For a clearly defined class of low-stakes, reversible decisions, you create an explicit rule: no justification, no notes, no captured reasoning. The only thing you document is the policy itself, one paragraph defining what counts as low-stakes and reversible, for example, decisions that can be undone within 48 hours and affect only your own workflow.

What you get:
You remove the overhead of constantly asking, “Should I document this?” That tiny meta-decision consumes more mental bandwidth than people expect. Sorting decisions in advance into “record this” and “don’t bother” frees your attention for the ones that matter.

What it costs:
The line between reversible and irreversible is fuzzy. Systems connect in ways we don’t notice until one little tweak knocks over three other things. So yes, you will occasionally get burned. The wager is that the saved cognitive effort outweighs the occasional mistake.

How AI can help:
AI can help you enforce your selective-ignorance policy by classifying decisions as low- or high-stakes in real time and nudging you only when something crosses your own “please document” threshold. For trivial, reversible tweaks, it can still keep a lightweight, auto-generated audit trail in the background, so you get the mental freedom of not justifying everything without completely erasing the breadcrumb if a “tiny” change turns out non-trivial later.

Voice and Video as Second Brain Capture

Many of these practices assume writing. That’s fine. Writing is great. It also isn’t mandatory.

A 30-second voice memo or a quick Loom screencast can replace written capture for many of them, especially Context-Resume Files (4), Worklog Narration (5), and Decision Journals (6). You speak faster than you type, and you’ll often say things aloud that you’d never bother to write down.

The trade-off is retrieval. Audio and video are miserable to skim and annoying to search. AI transcription is helping, fast, but it still isn’t perfect. So if you’re someone who won’t write a journal entry but will gladly mutter into your phone while walking to the kitchen, use voice. An imperfect record beats no record. Repeatedly.

Where to Start With Second Brain Practices”

You will not do all sixteen. You shouldn’t. But if you’re doing none of this yet, start in the bottom-left:

• Breadcrumb Comments (1)

• Commit-Message Narration (2)

They cost almost nothing, and the explanation appears where you need it.

If you already do those and still keep losing context, move upward:

• Assumption Tagging (3)

• Architecture Decision Records (10)

Those two are good for systems with enough complexity to bite back later (which is to say, almost all systems that survive long enough).

If you want more searchable memory, move right:

• Decision Journal (6)

• Failure Autopsy Log (7)

And if you want the higher-friction, higher-payoff end of the map, head into the top-right:

• Postmortems (13)

• Problem-Solution Case Bank (14)

• Spaced-Retrieval Review (15)

That’s where some of the deepest value lives. It’s also where your discipline goes to negotiate with your calendar.

No solo operator will do all sixteen. They don’t need to. A handful of good second brain practices beats a grand system that collapses after nine days.

And if this all feels like too much, remember Selective Ignorance (16). Decide which decisions don’t deserve any of this, then stop pretending every tiny choice needs a sacred archive.

Past You doesn’t need to become a perfect historian. Past You just needs to leave better clues.

Jurgen, Solo Chief

P.S. Paying subscribers get access to a PDF, SVG and PPTX version of the diagram. ⬇️

Read more

You’re not drowning in complexity because you’re lazy. You’re drowning because you’re responsible for everything. There’s a new kind of tool that might help—and a very old trap that comes with it.

This article is brought to you by Atoms, the first paying sponsor of The Solo Chief.

I once believed software was my bottleneck. If I could write more code, I could rule the world!

I had a beautiful product. Good idea. Clever features. It worked. Nobody wanted it. Well, a few people wanted it. But not enough people to keep the lights on and my pride intact. (I was even awarded Entrepreneur of the Year once because I was good at conveying my vision. My colorful diagrams of expected revenues looked fabulous on paper.)

So when I look at a tool like Atoms—a platform that helps a solo founder move from idea to market research, product spec, build, and deployment, ending with a live, testable product—I see two things at once.

First, I see where solopreneurs are headed in the near future.

Second, many people are about to get drunk on the wrong thing.

Where We’re Headed

The future part is easy to recognize. A year ago, researchers at a major Chinese university showed that role-based AI agents could act like a fake software company—CEO, CTO, programmer, tester, reviewer—and spit out working apps in minutes for pocket change. Others proved these systems could read docs, write code, run tools, fix some of their own messes, and push things live.

The age of “code assistants” is already old news. We’re now in the age of small digital staff with no sleep cycle and no opinion about your branding.

Atoms calls itself a “vibe business team,” and for once the marketing label earns its keep. Instead of handing you a code editor and wishing you luck, it gives you something closer to a small team you can talk to: a researcher who checks whether anyone actually wants your idea, a product manager who writes the spec, an architect, an engineer, and a coordinator who asks for your approval before making big moves. You describe what you want. The team builds it. You deploy.

For one person carrying the whole circus alone, that matters. The usual alternative is a sad little parade of freelancers, half-finished design files, one developer who disappears for ten days, and a payment setup that somehow becomes a spiritual crisis.

Solo operators don’t break because they’re lazy. They break because one calendar has to hold product decisions, customer calls, support emails, sales, invoices, bug reports, and whatever fire started while they were asleep. Eliyahu Goldratt’s Theory of Constraints applies here with embarrassing clarity: your business moves at the speed of its tightest choke point. For many solo chiefs, that choke point has been cross-functional execution.

A platform like Atoms that aims a cannon straight at that choke point is worth watching.

The Wrong Part of the Demo

Here’s where my eyebrow starts twitching. The phrase “fully functional business” gets thrown around on Substack like confetti at a wedding where nobody checked if the couple actually likes each other.

A live app isn’t a business. Those are different species. One can collect money. The other can create repeatable value for an actual customer who returns, refers, or at least doesn’t vanish after the free trial.

I know someone who built an entire business in public using an AI: it picked the niche, named the company, wrote the copy, and launched the whole thing with a human acting as its hands. The internet loved it. Revenue, however, remained stubbornly unimpressed. Activity everywhere, traction nowhere.

And then there’s the devious cloning problem. When building software becomes cheap, distribution becomes the knife fight. Some guy builds a successful AI product, and—poof!—a swarm of copycats throw up lookalikes almost overnight. Plenty of founders still talk as if the hard part is getting the app built. Cute.

The reliability gap is real but shrinking. Today’s AI agents still fumble on complex edge cases—authentication flows, billing recovery, security audits—the gnarly stuff that separates a minimal viable business from a scalable enterprise. But the trajectory matters more than the snapshot.

But that is the situation now. The AIs we have now are the worst we will ever have. The vision behind a platform like Atoms isn’t focused on now; it is built for tomorrow.

The Correct Job Description

This doesn’t kill the idea of a “vibe business team.” It points to the correct job description.

The promise of a platform like Atoms is the “agentic business team.” The reality (for now) is more “business and product experiments for solo businesses launching digital products.” Much like vibe coding was intended for software experiments, a vibe business is intended for business experiments. And that’s still a big deal. That’s the valuable version. If I’m a solo founder, I don’t need a machine to replace my judgment. I need one to spare me from stitching together hosting, UI, login, database calls, webhooks, deep research, data analytics, and the other fiddly nonsense that eats entire weeks and gives very little back.

What Atoms gets right—and what makes me think they’re onto something real—is starting with market research before a single line of code gets written. Their deep research agent analyzes demand, competitors, and niches before the builder agents fire up. That’s exactly the opposite of how most of these platforms work, and exactly the right order for a solo chief who can’t afford to burn months on something nobody wants.

(I know this because I burned months on something nobody wanted.)

The full sequence matters here. Research first, then spec, then build, then deploy with payments and auth already wired in. For a non-technical founder, that’s the difference between a product idea that sits in a notebook for six months and one that’s live and testable next week. You don’t need to know how authentication works. You need to know whether customers will pay. Atoms handles the first part so you can focus on the second.

Atoms also lets you export all generated code to GitHub, which means no lock-in if you outgrow the platform. That’s the kind of decision that makes me trust a company more than any marketing claim. (It’s why I’m on Substack in the first place. I can leave whenever I want and take my subscriber list with me, which is exactly why I’m staying right where I am.)

Management Changes When the Workers Change

Here’s the part that makes my brain itch. With human teams, management often means alignment, incentives, conflict, coaching, politics, hiring, firing, and trying to decipher what somebody meant in that Slack message with just a thumbs-up.

With AI agents, much of management turns into prompt design, boundary setting, quality checks, sequencing, and deciding where human review must stay in the loop because the machine has zero skin in the game. If the payments fail, the bot shrugs. If a privacy leak ruins your name, the bot slurps up another stream of tokens and calls it a day.

That’s why systems thinker Norbert Wiener’s old warning still matters. I’m paraphrasing, but here’s the gist: if you use a mechanical agency whose operation you can’t efficiently interfere with once you’ve started it, you had better be sure that the purpose you put into the machine is the purpose you actually desire.

Melvin Conway’s law lurks here too, even though Conway didn’t have agent swarms in mind. Your product resembles the communication structure of whatever built it. A solo founder with a team of AI specialists will build and ship differently from a founder with a designer, a salesperson, and a technical cofounder. Faster, probably. Cleaner in some places. Flimsier in others.

The garden grows in the shape of the gardener’s tools.

So, Is This the Future of Management?

For software-first solo businesses, yes. Enough that dismissing it would be silly.

The solopreneurs of the near future won’t spend all day coordinating departments. They’ll orchestrate a weird hybrid shop: a human chooses the market, the promise, the tradeoffs, the taste, the risks worth taking. A swarm of machines turns that intent into code, screens, tests, and stuff deployed at unnatural speed.

That lowers the cost of trying things. It shortens the time between idea generation and payment collection. It gives one-person businesses a kind of temporary organization without all the payroll, all the status games, all the “quick sync?” messages and stand-up meetings that somehow breed like rabbits.

Still, the oldest law in business survives untouched: if nobody wants what you’ve built, your stack doesn’t matter.

The future of management is becoming the art of knowing what parts of the business can be handed to tireless interns made of algorithms—and what parts still require a human who can smell demand before the invoice bounces.

Atoms seems well-positioned for this exciting future.

Atoms is a vibe business team that turns your ideas into businesses. It researches your market, designs the product, builds the frontend and backend, connects authentication and payments, and ships a live app you can charge for—not just a prototype. That could be a simple SaaS MVP, a paid tool, or another small digital product you can actually launch and validate with real customers.

Why I Said Yes to This Sponsor

I want to be honest about the sponsorship for a moment.

Atoms approached me as a potential sponsor. I looked at their product, their background (the team behind MetaGPT, one of the most widely used open-source multi-agent frameworks), their pricing, and their design choices. I said yes for a specific reason: they’re building for the audience I write for. Solo founders. One-person businesses. People who carry the entire circus alone.

Is Atoms perfect? Probably not. It’s early-stage, and the output quality depends heavily on how clearly you describe what you want. (Welcome to working with AI in general.) As with most AI platforms, there’s a learning curve in using credits efficiently, especially when you’re still figuring out what you want to build.

But the direction matters more than the current polish. The fact that Atoms starts with market research before building, that they let you export your code, that they’re designed around the workflow of a solo operator rather than an engineering team—those are the right instincts.

I wouldn’t bet my entire business on any single AI platform today. But I would use one to test whether a business idea has legs before I invest months and money into building it the old way.

That’s the real use case. Not “replace your team.” Not “build a startup in five minutes.” Just: run the experiment. See if demand shows up. If it does, you’ve got something worth growing. If it doesn’t, you’ll have saved yourself six months of pretending a product is a business.

If you have an idea sitting in your notes app gathering dust while you tell yourself you’ll find a developer next month—you won’t find a developer next month. But you might find a customer if you run the experiment this week.

Try Atoms →

Jurgen, Solo Chief.

Layoffs make headlines. The explosion of solo businesses doesn’t. Both are the same trend.

Widesizing means AI lets one capable person do what once took a team. The economy isn’t shrinking; it’s fragmenting into smaller, scrappier units.

Large tech companies shed people with the same solemn theater used for royal funerals, and half the internet reacts as if paid work itself just fell off a cliff.

Yes, Block cuts jobs. Oracle cuts jobs. Amazon cuts jobs. Every week another CEO waves some consultancy’s AI slide deck around and calls it their new strategy playbook. Then the alarmists rush in: “Human jobs are finished when corporations need only one-tenth the staff. We need a revolt!”

Slow down, everyone.

Widesizing: The Thing Nobody Talks About While Staring at Layoffs

Headcount inside giant firms is only one place where work shows up. If one company has 100,000 employees, or 1,000 firms have 100 each, or 100,000 solo operators each carry their own little shop, you still have 100,000 people making a living. The shape changes. The payroll department changes. The org chart gets set on fire. Work itself keeps evolving. And let’s not forget that the enterprise itself is a fairly recent invention. The East India Company got its charter in 1600. Before that, humans organized work perfectly well for millennia without an HR department or a Microsoft subscription. The corporation might not last forever either.

That’s why I have little patience for the usual AI doom-chanting. Negativity bias makes people stare at downsizing and completely miss what I’d call widesizing. They see layoffs at the cathedral and ignore the bazaar springing up outside.

Widesizing Is Already Happening

And the bazaar is real. Around the world, millions of new business applications are being filed each year, and much of the jump comes from non-employer firms. Solo businesses. Tiny outfits. One-person agencies. People who used to need a coordinator, analyst, copywriter, junior assistant, and a corporate permission slip can now do the work themselves with software and stubbornness.

That matters.

Widesizing means AI lowers the minimum viable size of a business. A lot of economic activity that once demanded a team can now be handled by one capable person with a laptop, a few subscriptions, and a mild allergy to bureaucracy. Schumpeter would recognize the pattern immediately. Creative destruction is messy because the old machine breaks in public and on the news, while the new machine starts in coffee bars, spare bedrooms, and WhatsApp groups. Nobody writes a headline about someone quietly launching a profitable consultancy from their kitchen table. But multiply that by a few million and you’ve got a trend that dwarfs any single round of layoffs at Meta.

“Widesizing means AI lowers the minimum viable size of a business.”

I’m sympathetic to that world because I’ve lived in it. Actually, no, I’ve failed in it. Repeatedly. I’ve had big ideas, shiny diagrams, brave talk, and the confidence that should come with adult supervision. Some bets worked. Some face-planted so hard they left a crater. (My attempts at building online platforms deserve their own postmortem, which I’ll write once the emotional wounds heal.) So when I say smaller units can win, I’m not selling a fantasy of heroic founders sipping espresso while AI does the laundry. I’m saying the center of gravity is shifting, and widesizing is the word nobody’s using for it yet.

But Let’s Not Get Romantic About It

The cheerful “everyone can become a solopreneur now” crowd skips over the ugly bits. A steady salary bundles income, benefits, training, legal cover, a bit of status, and someone else to absorb economic shocks. Solo work bundles freedom with risk, cash-flow swings, admin sludge, and that charming moment when a client pays three months late while your mortgage remains weirdly punctual.

And AI can punch solopreneurs in the face too.

The Upwork data after ChatGPT’s earlier releases showed freelance writers and designers getting fewer jobs and lower earnings. That’s a real warning. When AI turns entry-level output into a commodity, clients don’t automatically spread the money across thousands of independents. Often they keep the savings and ask one person to do the work of four. Sometimes they do it themselves badly (which is its own small revenge, but cold comfort if you’re the one who lost the gig).

Then there’s the platform trap. Ask the graveyard of “AI wrapper” founders who built neat little products on top of Claude and got screwed when Anthropic announced native features that swallowed their businesses before lunch. Courage didn’t save them. Neither did imagination. Building on someone else’s castle grounds means the landlord can move the fence whenever they feel like it. This is something Joel Spolsky warned about over two decades ago, and it’s still true: don’t build your house on a platform that doesn’t love you back.

And no, some parts of the economy won’t fragment into cheerful little hobbit businesses. Chip fabs, frontier models, biotech labs, cloud infrastructure, defense contracts, power grids. You’re unlikely to go solo with a semiconductor plant. Some layers stay huge because physics, capital, regulation, and plain stubborn reality demand concentration. TSMC isn’t about to be disrupted by some guy in a garage with a 3D printer.

The Barbell Economy, Not the Cliff

So where does that leave us?

We will have a split future, which is less cinematic than “AI will destroy everything” and more useful than “everyone should just learn to code.”

A small number of firms will get bigger because AI rewards scale in data, distribution, and compute. Think barbell strategy applied to the whole economy. Countless tiny firms will also appear because AI slashes coordination costs and expands what one skilled person can deliver. Both things can happen at once. In fact, they already do. The middle is the dangerous place, the comfortable mid-sized company that’s too big to be nimble and too small to compete on raw infrastructure.

That’s the part the doom merchants and the techno-messianists both mangle.

The alarmists talk as if every layoff is a one-way ticket to irrelevance. The evangelists chirp as if everyone can turn a pink slip into a thriving solo practice before the weekend. Rubbish. Some people will get crushed. Some will do better than they ever did inside a large company. A lot will bounce between contracts, products, part-time roles, and experiments while the market sorts out what humans are still best at and what buyers still value.

Nobody knows how this shakes out. Anyone who tells you they do is selling something. (Probably a course.)

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

So What’s the Useful Question?

I wouldn’t ask whether AI will wipe out jobs. That question is too blunt, like asking whether “transportation” will kill the horse. (It did kill the horse-as-transport. It didn’t kill the horse.) I’d ask who gets to shape the next layer of work: giant firms, tiny firms, or people who can switch between both without waiting for permission.

I think that last group has the best shot.

The old deal was simple. Join the enterprise. Climb the ladder. Hope the machine loves you back. The new deal is harsher but more honest. Build a portfolio of skills that don’t evaporate when the next model drops. Own your distribution (your reputation, your network, your ability to find clients without a recruiter). Use AI as a force multiplier where it helps, keep your judgment where it counts, and avoid building businesses that can vanish during somebody else’s product launch keynote.

Widesizing isn’t a utopia. It’s a map of where economic activity goes when AI makes organizations flatter, smaller, and more numerous.

The future of work won’t be decided by pearl-clutching LinkedIn prophets or billionaire CEOs who think every human problem looks better inside a dashboard. It’ll be decided by millions of workers choosing whether to stay ornamental inside bloated organizations or become useful in smaller ones.

“The future of work won’t be decided by pearl-clutching LinkedIn prophets or billionaire CEOs who think every human problem looks better inside a dashboard.”

Some will wait for the wave to hit them.

Others will grab a surfboard.

Jurgen, Solo Chief.

P.S. Are you downsizing, widesizing, or still deciding which board to grab?

Share

TL;DR: The best AI stack for solopreneurs depends less on which tools you choose and more on how well you know your own strengths and weaknesses. A leader who understands how their brain actually works can match tools to how they think, building what amounts to a mirror of themselves rather than a copy of someone else’s setup.

Is AI working as a prism for you, expanding the reach of your efforts? It should.

Two solopreneurs walk into 2026 with the same three AI tools. Same subscriptions, same access, same tutorials bookmarked. Six months later, one of them is running a smoother operation than they have ever had. The other is drowning in half-configured dashboards, abandoned automations, and a growing suspicion that AI is just expensive noise.

What separates them has nothing to do with the tools.

I didn’t figure that out from reading about it, I figured it out by switching tools myself and watching what actually changed. (I help leaders build exactly this kind of self-aware AI setup in my coaching practice.)

In this post, you’ll learn:

• Why tool-hopping is a symptom, not a strategy

• How I matched three specific tools to how my brain actually works

• The three questions that should determine your AI stack before you sign up for anything new

The Tool-Hopper

You know this person, or… maybe you’ve been this person.

They follow every “my AI workflow” post on Substack and sign up for the new tool the week it launches. Look, their setup is impressive: Claude, ChatGPT, Gemini, Perplexity, Copilot, and three automation platforms, each doing roughly 10% of what it could. It looks great in a screenshot, but in practice, they spend more time configuring workflows than doing the work those workflows were supposed to handle.

Jurgen Appelo nailed this in his piece on technostress and FOBO, the fear of better options. The tool-hopper lives in FOBO. They are always one tool away from the breakthrough, and the breakthrough never comes, because the problem was never the tool.

What I Actually Use

I run four businesses: Leadership in Change (a Substack newsletter on AI leadership), Cozora (an AI education platform with live weekly sessions), Newsletter Compass (a growth tool for newsletter creators), and a coaching and AI build practice at jsalinas.org. My entire AI stack is three tools.

Claude is my thinking partner.

I don’t use it the way most people describe. I don’t ask “what should I do?” I dump seven paragraphs of raw thinking into a conversation and say, “What other approaches exist here? What am I missing? How does this connect to what you already know about my audience?” This works because I’ve invested the time to build the context. I maintain 10 to 12 page skill documents for each of my businesses inside Claude. When I mention Leadership in Change, it pulls the right document automatically and has the full picture: my voice, my audience, my strategy, my constraints. Claude is also my copy editor. I write rough, fast, and big, and Claude helps me refine without flattening.

Gemini generates my images.

Claude creates an on-brand prompt based on my visual guidelines, and I feed that prompt to Gemini for generation. Two tools collaborating on one output, each doing what it does best.

NotebookLM is my research layer.

Here’s the thing: before I go deep into any business decision or article, I load my sources into NotebookLM first and ask questions against those specific documents. This nearly eliminates hallucinations, which matters enormously when the research feeds business decisions, not just blog posts.

Three tools. Each one matched to a specific gap or strength in how my brain works.

Why This Stack Works for Me (and Probably Won’t Work for You)

Here is what I want to be direct about: there is no magic AI stack. Just like there is no magic prompt, no magic business strategy, and no magic Substack note that will make you go viral.

I am not great with details. I think in big connected ideas, and I like to collaborate, to go back and forth, to refine through conversation. Claude is built for exactly that kind of brain. I spent a full year inside ChatGPT before I realized it was the wrong fit for how my brain works. That year wasn’t wasted; it taught me what I actually needed, but it was a year of building in the wrong tool. When I tried Claude, its project system and skill documents matched how I actually think and collaborate. I didn’t switch because Claude was “better.” I switched because Claude was better for me.

The tool-hopper copies someone else’s stack and wonders why the results feel hollow. The self-aware solopreneur builds what I think of as a Mirror Stack, an AI setup that reflects how their specific brain works, not a screenshot of someone else’s workflow.

Three Questions Before Your Next Tool

Before you evaluate another AI tool, before you read another “my perfect workflow” post, answer these honestly:

1. What am I genuinely bad at? Not “what do I wish I were better at.” What actually falls through the cracks, consistently, when nobody is watching?

2. What kind of thinking do I do best? Collaboration? Deep solo focus? Visual processing? Verbal reasoning? Big picture strategy? Granular detail?

3. When I’m doing my best work, what does the process actually look like? Fast and messy, then refined? Slow and methodical from the start? Conversational and iterative?

Your answers should determine your stack. Not a Substack post. Not a YouTube tutorial. Not mine.

If You Only Remember This

• The gap between two solopreneurs using the same AI tools is self-awareness, not software. The tools are commodities. Knowing your own brain is the competitive advantage.

• Go deep into one tool before you touch another. Breadth feels productive but produces shallow, fragile systems. Depth compounds.

• Build a Mirror Stack, not a borrowed one. Match every tool to a specific strength you want to amplify or a specific flaw you need to cover. If you can’t name which one, you don’t need that tool yet.

  

What’s the one thing you’re genuinely bad at that you haven’t matched a tool to yet? I’d love to hear it in the comments.

Joel Salinas is an AI leadership coach and strategist who writes Leadership in Change on Substack and helps leaders build AI-aware practices at jsalinas.org.

The gap between AI tinkerers and everyone else is closing fast. Your window to claim your own galaxy is now.

While most people chat with bots, a few solopreneurs run autonomous AI agents that operate their businesses 24/7. That capability gap is your temporary window to build something competitors can’t replicate. You have no time to lose.

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

Halt!

Stop what you’re doing for a moment.

Actually, I don’t care. Keep scrolling if you want. But 2026 is the kind of moment most people only recognize when it’s already far behind them, and I’d rather not watch you miss it while you’re binge-watching Andor or The Mandalorean.

Right now, a tiny number of solopreneurs and one-person businesses are building their own AI agent orchestration. With real AI agents. Digital assistants that crawl, build, deploy, and run entire business operations 24/7 while their creators walk the dog, watch Skeleton Crew, or cook dinner for friends.

Meanwhile, most people are still asking ChatGPT to write their resumes or generate a slide deck from three bullet points.

That capability gap is temporary. It will close fast. And if you don’t jump to lightspeed now, you’ll spend the rest of your life wondering how you could let such a unique opportunity pass by unnoticed. (As I did by not claiming a hundred domain names in 1995.)

I call this the Hyperdrive Advantage.

My Clunky, Self-Made AI Agent Stack

I’m no expert when it comes to AI agent orchestration. I should probably say that louder. I’M NO EXPERT. But I do spend a lot of time building my own workflows, and I’ve failed at enough of them to have opinions.

For example, I have an automated workflow keeping track of hundreds of Substack publications and their follower and subscriber counts. I also have several agentic workflows where Claude, ChatGPT, Gemini, and Grok act as an LLM council, offering me different perspectives on a research question or a provocative statement before I turn that into an essay. I even have an agent proactively informing me of the upcoming birthdays of the people closest to me.

Most of the struggle with AI orchestration (in my case) is just dealing with all the edge cases. An external service is down for an hour; I run out of tokens mid-workflow; some input is formatted in an unexpected way. Each time, I reroute the power and improve the workflow. It’s slow going, but rewarding. The scenarios are getting more solid every day.

And then there are the problems that don’t get me any further. They just eat up my time. Claude had a bug in its Cowork architecture that messed things up and held me back for several weeks. Make has a few annoying quirks with arrays that waste my time each time I use them. Last weekend, I ran into a wall with Perplexity: the MCP connectors available in Perplexity Computer are not available in Perplexity Chat, which completely defies the point of using Perplexity as an interface to my agentic workflows. There goes another two hours of work leading to nothing (until they fix this)... 🤷🏻‍♂️

It’s the pain of living at the frontier of AI agent orchestration. Only half of everything works. The other part is permanently under construction. 🚧 Every day is a struggle with tokens, time-outs, and the ever-present “Something went wrong.”

But I persist. I do not want to wait until turnkey agentic AI orchestration is available for everyone. That’s exactly why I’m doing this now. My clunky, self-made agentic stack is a temporary project because I aim to exploit the Hyperdrive Advantage.

I am a rebel patching my own Millennium Falcon so I can go on adventures before the rest of the universe arrives in their standard X-wings.

Some early mornings, I stare at yet another failed workflow and wonder if I’m wasting my time. Should I just wait for the polished tools like everyone else? Maybe this whole bespoke-AI-stack thing is just me being stubborn? Then I remember: that feeling of doubt is exactly what makes this moment valuable. If it were easy, everyone would already be doing it, and there’d be no advantage left to claim.

The Original Hyperdrive Maneuver

In 1975, George Lucas wanted visual effects nobody had ever seen. The technology didn’t exist. So he rented a warehouse in Van Nuys (a neighborhood in Los Angeles) for $2,300 a month, hired a ragtag crew of college students, artists, and engineers, and called it Industrial Light and Magic. Everything was hand-built, one-of-a-kind, and constantly breaking. John Dykstra, his effects supervisor, admitted it was “a long shot.”

But Lucas didn’t build ILM to sell special effects. He built it to make Star Wars. The tech was the engine; the franchise was the destination. By the time Hollywood figured out what happened, Lucas owned the characters, the merchandise rights, and a fan base that would follow him for decades. His moat was never the effects shop. It was the galaxy he built with it.

That’s a Hyperdrive Maneuver: build clunky, bespoke technology before anyone else can, and use the head start to lock in something permanent.

Jeff Bezos did the same thing in 1994, building online retail infrastructure when “buying things on the internet” sounded as trustworthy as getting medical advice from a stranger in a parking lot. By the time e-commerce tools became cheap and polished, he already owned the customer data, the logistics, and the habits of millions of buyers. The bespoke tech was the ladder. The moat was everything he built while climbing it.

Right now, we’re at that same moment with agentic AI.

Your Agentic AI Advantage

It seems common knowledge that many people are using AI, but only few go as deep as the technology allows. Most are just chatting with a bot. Nothing more. The number of people running their own AI agents (digital workers that autonomously research, create, publish, and operate) is vanishingly small.

That gap between early adopters and everyone else is your window. Every major technology wave had one: electricity, the internet, smartphones. Each time, the people who moved while the tools were still rough ended up owning systems that latecomers could never quite replicate.

Think of it like evolutionary biology. In ecology, when a new habitat opens up (a volcanic island forms, a forest fire clears a valley, an asteroid wipes out the dinosaurs), the species that colonize first don’t necessarily have the best genes. They have the best timing. They radiate into empty niches and establish themselves before competition even arrives. Biologists call this adaptive radiation. I call it what’s happening right now with agentic AI.

Agentic AI is in its Internet-in-the-1990s phase. Clunky, unreliable, confusing to most people, and wildly powerful for the few who bother to make the modifications themselves.

If you have the grit and savviness to wrestle with today’s messy AI orchestration tools, you can pull an Agentic Hyperdrive Maneuver: launch yourself into economic spaces that barely exist yet and establish yourself before the rest of the world shows up. That’s the Agentic Hyperdrive Advantage.

More Than a First-Mover Advantage

You might be thinking: “Isn’t this just first-mover advantage with a Star Wars theme bolted on top?” Fair question. Wrong answer.

The first-mover advantage has a well-known weakness: the fastest-learner advantage. It doesn’t matter who moves first if a smarter competitor can show up later and out-learn you. Google was nowhere near the first search engine. But they learned faster than AltaVista, Lycos, and all the others combined. The first movers’ head start turned out to be worthless because search engines have no customer lock-in. Users switch in less than ten seconds.

The Hyperdrive Advantage works differently. We’re talking about a technology that gives you access to economic systems that were straight-up inaccessible before. New markets. New services. New ways of operating that weren’t economically viable until you wired up your own agents to do the work. This isn’t about bolting a faster engine onto the same kind of vehicle and hoping nobody catches up.

Google could beat AltaVista because they were competing on the same road. George Lucas couldn’t be beaten that way because he wasn’t selling special effects. He used special effects to sell a galaxy. By the time competitors got their own effects shops, Lucas had already built the franchise, the merchandise empire, and the audience. The system was claimed.

That’s the Hyperdrive Advantage: a first-mover advantage large enough that it gives you first access to a vast new space waiting to be claimed by you. You don’t get a head start by building the best technology. You use improvised technology so you can be the first to plant a flag in unexplored terrain.

This is the crucial takeaway: with the Hyperdrive Maneuver, you don’t make infrastructure your moat. Instead, you use temporary infrastructure to make a moat out of data, reputation, or network. You assume the infrastructure is replaceable.

Here’s where it gets lonely, though. When you’re building your own AI orchestration, there’s no manual. No support team. No one to call when the whole thing falls apart at 3am. You are the single wringable neck for your own duct tape-architecture and for the business you’re trying to build on top of it. That double weight is real. And if you’re already building your own agentic stack, you probably feel it too. (If it helps, know that George Lucas apparently threw up from the stress during the filming of Star Wars. So at least we’re in good company.)

Solo Operators Building Their Hyperdrives

We’re still early. So early that the agentic Lucas and agentic Bezos haven’t revealed themselves yet. In the 1990s, nobody could have pointed at a clear winner of the internet either. But the hyperdrives are being assembled right now, in warehouses and home offices, by people operating in the Outer Rim of their industries.

Aaron Sneed runs a defense-tech company, solo. His secret is “The Council”: about 15 custom AI agents acting as functional departments. Legal, Ops, HR, Marketing, Research. They handle continuous work and escalate only the hard decisions to him. One person operating in a market that normally requires a funded team just to clear compliance.

A photo booth entrepreneur (yes, photo booths) orchestrates over 20 AI agents handling inbound messages, quoting, booking, content editing, marketing, and bookkeeping. A traditionally labor-heavy local service, now scaling across multiple regions with a single operator.

Okay, maybe these examples don’t sound impressive to you. But that’s the point. These aren’t billion-dollar empires—yet. They’re just scrappy little spacecraft. But give them some time.

Meanwhile, I’m building my own hyperdrive right now. Soon, I will point it at some faraway systems. Who knows? Maybe I can claim some little moon or asteroid belt for myself.

Build the Moat Before the Fleet Arrives

Your Hyperdrive Advantage is temporary. I can’t say this enough. The clunky tools you’re wrestling with today will be polished, packaged, and sold to everyone within a couple of years, maybe sooner. Your bespoke AI stack will be outdated. That’s fine, as long as you’ve used the head start to build something that outlasts the tech.

Lucas didn’t cling to his improvised motion-control cameras. He used them to lock in Star Wars. You need to do the same: use your heavily customized agentic hyperdrive to accumulate data, reputation, and relationships before the fleet arrives with their standardized models.

Every client you serve, every experiment you run, every failure you recover from compounds into something latecomers can’t buy. Your prompts get sharper. Your domain knowledge gets deeper. Your name becomes the one people associate with solving a specific problem. Investors call this a compounding economic moat. I call it the whole point.

Your moat might be the 500 clients whose workflows you’ve automated. The proprietary dataset your agents have been quietly building for eighteen months. The reputation as the person who cracked a problem before anyone else even understood the question. It doesn’t matter that you did all of that with a hand-modified, cobbled-together, famously unreliable hyperdrive. You use your clunky tech stack to turn yourself from a rebel into the ruler of your own galaxy far, far away.

Soon, everyone will have their own turnkey, off-the-shelf, X-wing style hyperdrives. Manufacturers will churn out AI orchestration stacks that beat anything you can build yourself. Heck, you might even swap out your own agentic AI orchestration for the new technologies that come onto the market. That’s perfectly fine. You’ve already built your moat. You already rule a system, maybe even an entire galaxy. You took advantage of the Hyperdrive Maneuver.

Your Falcon Is Waiting

And here’s why this moment favors you, the solo operator: while large organizations are stuck in committee meetings debating their AI strategy (and convening sub-committees to evaluate the committee’s recommendations), you’re already patching your Falcon and jumping to lightspeed. No procurement process. No risk assessment framework. No six-month pilot program. You tinker, ship, blow things up, and fix them before their Imperial Senate has even scheduled the next session.

Speed is the solopreneur’s superpower right now. Not capital, not headcount, not brand recognition. Speed. The ability to try something at 10am, watch it fail by noon, fix it by 3pm, and have it running by dinner. Large organizations can’t do this. They have too many layers, too many stakeholders, too many people whose job it is to say, “Let’s discuss this in our upcoming PI planning next month.”

⚠️ And sorry to burst a few bubbles, but Substack subscriptions and self-paced courses on “how to make your own hyperdrive” are not what I mean here. Make-money-online schemes don’t count. Selling hyperdrive manuals and leaving the exploration to the actual rebels does not make you the ruler of your own faraway galaxy. ⚠️

If you’re sitting there thinking, “But I’m not technical enough for this,” remember: neither was George Lucas. He found a handful of film-school dropouts, hobbyist engineers, and other misfits who were crazy enough to try. They figured it out as they went. Like the Millennium Falcon, your AI orchestration stack will be held together with improvisation and determination, not engineering degrees.

The speed gap won’t last long.

Use it wisely.

Jurgen, Solo Chief.

P.S. What does your clunky Millennium Falcon look like right now? Tell me about your stack!

Share

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

I wrote this piece for class 22 of ScaleUpNation‘s Board Program. Interested in participating in one of their next classes? Check out their upcoming webinar:

Art of Scaling’ Webinar by Menno van Dijk.

On the 22nd of April ScaleUpNation hosts a (free) webinar about ‘The Art of Scaling’ from a Board perspective.
ScaleUpNation has 300 plus alumni, all specialist in the ScaleUp world.
This webinar is for executives interested in ScaleUps.
Founders, Investors, Advisors, C-level and (future) Board members.

You can subscribe using this link:
https://lnkd.in/eSHqfAUR

Hope to see you on the 22nd of April at 13.00 CET/1.00 PM

AI can now do half of the advisory board’s job. The courts want governance boards to do the other half.

Advisory boards are squeezed between AI tools that produce strategic advice cheaply and governance boards absorbing oversight responsibilities. Only contextual judgment survives.

Here’s an uncomfortable question for anyone sitting on an advisory board:

What exactly do you offer that Claude and ChatGPT can’t?

Governance boards just received a legal nudge that feels less like guidance and more like a deadline. The EU’s AI Act (Regulation 2024/1689) requires accountability for high-risk AI systems by August 2026, and regulators are treating that accountability as a board responsibility. The timeline is public and fairly blunt: enforcement for high-risk systems begins on August 2, 2026.

When regulations set a date, boards eventually pay attention.

The advisory board model, meanwhile, is drifting into an awkward middle position. A host of AI models and tools are making some of the board’s traditional contributions cheap and immediate. Meanwhile, governance boards are absorbing the rest because liability travels toward whoever signs the documents. The advisory board isn’t disappearing. But the version many organizations still maintain looks suspiciously like a relic of the past.

Let’s walk through the pressure points.

The Compliance Stampede

Corporate governance has a curious rhythm. Boards move slowly until courts or regulators start writing opinions. That phase has begun for AI.

A 2025 Deloitte Global Survey reported that 66 percent of boards say they have limited or no AI expertise. The number itself isn’t shocking. Boards historically recruit for finance, operations, and regulation. Deep technical literacy rarely makes the shortlist. What changed is the expectation surrounding it.

The oversight principle is gradually expanding toward technological risks. Recent Delaware Court of Chancery rulings have clarified that directors must receive regular briefings on mission-critical technologies. AI is quietly joining that category. The message from the courts is simple: directors cannot claim ignorance about the systems shaping their company’s core operations.

On my side of the Atlantic, regulators went further and put a date on the obligation.

The EU AI Act entered into force on August 1, 2024. High-risk system obligations begin in 2026. Boards now have roughly one governance cycle left to figure out what responsible oversight looks like, and organizations are responding with predictable speed.

Board intelligence platforms are building AI tools directly into governance workflows. Vendors such as Diligent and Nasdaq Boardvantage now advertise automated briefings, document summaries, and preparation tools that reduce director reading time by as much as 60 percent and board preparation time by up to 80 percent.

Anyone who has sat through a board meeting recognizes the irony. The typical board pack runs well beyond 200 pages. AI can now summarize it in minutes. Whether that represents efficiency or an admission that those pages shouldn’t have existed in the first place remains an open question.

Either way, governance boards feel the pressure to develop internal AI competence quickly. Committees are being created. Charters are being updated. Recruiters are searching for directors who understand machine learning well enough to challenge management.

The baseline is shifting.

The Squeeze on Advisory Boards

Advisory boards historically filled a gap that governance boards could not.

Directors carry fiduciary responsibilities. That makes them cautious. Advisory boards had the freedom to think more experimentally. They offered specialized domain insight, industry connections, and candid feedback without legal exposure.

That arrangement worked because two things were scarce: information and perspective.

Both are changing.

General-purpose AI tools now produce credible strategic analysis in seconds. A founder exploring pricing models, market entry, or regulatory exposure can run several scenarios through a capable model before breakfast. The output isn’t always perfect—but neither are board members. Yet the marginal cost approaches zero, and the speed is astounding. The AIs do the job faster, cheaper, and—more often than not—better.

Advisory boards often justify their existence through the “external brain” function. The sparring partner. The person who sees patterns across industries. But ChatGPT, Claude, Gemini, Grok, and Perplexity now perform that role surprisingly well. And one only needs to visit the Product Hunt website to find a legion of AI-driven tools, such as SoloBoard, BoardOS, Foundora, aicofounder, AI debatearena, and Team Debate, eager to offer advice to startups and scale-ups at a mere fraction of the cost of an advisory board member lunch.

There’s even a free download available: Similar to the age-old Delphi Method, the LLM Council pattern is a multi-agent AI architecture where multiple LLMs collaborate like a council to produce more reliable answers through structured deliberation. The pattern enhances reasoning by leveraging diverse AI model perspectives.

Meanwhile, pressure also arrives from the other side. Governance boards are absorbing decisions that once lived in advisory territory. AI risk management, regulatory interpretation, and technical oversight increasingly sit inside fiduciary structures because the courts expect them there.

The accountability disappears upward.

The strategic advice vanishes downward.

The remaining territory for advisory boards becomes narrower by the day: decisions that require judgment shaped by experience, specific industry context, and relationships within and around the organization. That space still exists. It just looks smaller than it used to.

The Signaling Trap

Another dynamic deserves a less polite description. Let’s be honest: many advisory boards function as investor theater.

A skeptical analysis of AI advisory structures would describe them as human validation layers. A startup assembles recognizable names; investors notice the credibility signal, and the board rarely challenges management in any serious way. The advisors charge for their network and reputation, not their experience and insights. Everyone involved understands the arrangement.

The danger appears when executives believe their own marketing. A symbolic advisory board creates the feeling of strategic coverage without the substance of it. Founders assume someone is stress-testing the plan. Investors assume the same. In practice, nobody asks the hard questions. The technology sector is littered with the remains of startups and scale-ups that were guided off a cliff by theater directors.

Large organizations have, with varying success, experimented with external advisory councils to replace internal ethics groups. Governments have assembled high-profile AI advisory boards only to dissolve them months later when political pressure increased. The structures exist. The influence varies dramatically.

Advisory boards were originally intended as places for uncomfortable conversations. When they become branding devices, those conversations disappear.

The Surviving Species

So, what remains valuable about an advisory board?

Contextual judgment.

AI can produce information at extraordinary speed. Governance boards can impose accountability. Neither handles the messy space between them particularly well.

Consider the Techleap “AI Paradox” research covering the Dutch startup environment. The Netherlands has one of Europe’s highest concentrations of AI talent. Yet only 21 percent of its AI startups successfully scale, compared with 81 percent in the United States.

The problem is not information or technical expertise. The gap appears in decisions that connect technology, regulation, financing, and organizational design. Those choices rarely follow a formula. They depend on context that lives inside specific industries and networks.

This is where advisory boards can still matter.

Oxford Law researchers recently proposed an update to the business judgment rule, sometimes called “BJR 2.0.” Their argument is that directors now require cognitive adequacy to understand the technologies influencing corporate decisions. Technological literacy has joined financial literacy as a baseline competence.

Advisory boards operate in the layer above that baseline. They apply experience to situations where the playbook hasn’t stabilized yet. A founder deciding whether to launch an AI product under the EU AI Act. A scaleup balancing regulatory exposure against speed of deployment. An executive team confronting a market that changes every quarter.

AI produces the data. Governance boards enforce accountability. Advisory boards interpret the messy middle.

The organizations that recognize this distinction will keep their advisory boards small, specific, and uncomfortable enough to challenge assumptions. The rest will continue assembling impressive lists of names that rarely challenge directors’ decisions.

AI has reduced the cost of knowledge.

But judgment remains stubbornly human.

Jurgen Appelo, Solo Chief

A big thanks to yet another batch of new supporters of The Solo Chief. Cheers to , , , , , , , , , , , , , , and others whose accounts I can’t find. You all rock. Thanks for the coffee! ☕

If you support my work by upgrading to paid (it only costs you one coffee per month), I will send you a full, digital copy of my latest book, Human Robot Agent and I will invite you to private, one-on-one online meetups.

Subscribe now

References

1. European Commission AI Act Service Desk. Timeline: Implementation of the EU AI Act. Available at: https://ai-act-service-desk.ec.europa.eu/en/ai-act/timeline/timeline-implementation-eu-ai-act (Accessed: April 2026).

2. Deloitte. Governance of AI: A Critical Imperative for Today’s Boards (2nd Edition). 2025. Available at: https://www.deloitte.com/global/en/issues/trust/progress-on-ai-in-the-boardroom-but-room-to-accelerate.html (Accessed: April 2026).

3. Ervin Cohen & Jessup LLP (ECJ Law). Delaware Expands Expectations for Board Oversight of Cybersecurity, by Jeffrey R. Glassman. January 20, 2026. Available at: https://www.ecjlaw.com/ecj-blog/delaware-expands-expectations-for-board-oversight-of-cybersecurity-by-jeffrey-r-glassman (Accessed: April 2026).

4. Nasdaq. AI for Boards: Governance Solutions. Available at: https://www.nasdaq.com/solutions/governance/boardvantage/ai-for-boards (Accessed: April 2026).

5. Techleap. State of Dutch Tech Report 2026. March 15, 2026. Available at: https://techleap.nl/reports/state-of-dutch-tech-report-2026 (Accessed: April 2026).

The future of work isn’t AI taking your job. It’s software updates stealing your morning.

Solo operators pay the full cost of endless tool churn themselves. When vendors move the buttons, your leverage collapses back into hand work.

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

I woke up this morning looking at two failed automations and an error log overflowing with warnings. My OpenAI credits maxed out, two scenarios stopped, and one of my automatic imports (which worked flawlessly for months) has now wrecked 2,000 contact records.

Welcome to the future of work.

If you’re a Solo Chief, you know the feeling. You start the day intending to do real work, and by lunch, you’re elbows-deep in tokens, credits, permissions, migrations, hidden settings, broken workflows, vanished buttons, and help pages written for a version of the product that deprecated last week. By dinner, you’ve moved data from A to B, upgraded C to D, reconnected E to F, re-authorized G and H, and reformatted the output twice because some vendor decided your old workflow needed a “better experience.”

People call this learning.

Sometimes, it’s just maintenance with better PR.

And yes, I know the counterargument. Ethan Mollick found that people using AI can finish faster and with better results. I’ve seen that too. LLMs often help me cut through the fog.

Until they don’t.

Because the problem isn’t only that the tasks are hard. The problem is that the ground keeps moving while you’re standing on it. The model says, “Click the menu item on the left,” and the menu item now lives in a floating panel on the right, unless you’re on the new version, unless your account is in the old rollout group, unless Google had one of its little morning moods.

That’s why so much work feels fake-busy. Your effort goes into re-learning surfaces instead of compounding skills. Many jobs will be disrupted in five years, and the half-life of some technical skills has dropped below 2.5 years. You don’t need a consultant to tell you what that means. It means you can finish a course today and toss away your certificate tomorrow.

Then the management crowd arrives with a pep talk. Stay adaptable. Keep experimenting. Be a lifelong learner. All true. Some say the illiterate of the 21st century aren’t those who can’t read and write, but those who can’t learn and relearn. True enough. Also incomplete. For solo operators, all that learning and unlearning has a cost, and you pay all of it yourself.

That’s the part the gurus skip.

Solo Operators Pay the Full Churn Tax

In a company, tool churn can be spread around. One person checks release notes. Another handles migrations. Someone in ops fixes the workflow. Someone in IT resets access. In a one-person business, congratulations, you’re all of them! The same brain that has to sell, deliver, invoice, write, and think also has to play part-time plumber for a leaking pile of half-finished software. A sane person can absorb only so many “small updates” before their brain starts filing a complaint.

And it gets worse at the exact moment you need leverage most. Solo work scales through stable patterns. Templates. Checklists. Automations. Workflows. A reliable sequence you can run without re-deciding everything each time. When a vendor moves the controls, kills an integration, changes the API, or sunsets a feature, your leverage collapses back into hand work. When you spend your executive function deciphering a new interface, it can’t also do your best thinking.

The Lindy Effect as a Survival Strategy

Some people escape the treadmill by refusing to join it. Some of us do very well by keeping a small, boring stack and ignoring shiny new toys. That’s the Lindy Effect doing its magic: the longer a tool has been around, the better chance it’ll still be around next year. Boring tech has better manners.

So the future of work isn’t one thing. It depends on whether you’re building your business on bedrock or quicksand.

But for many Solo Chiefs, the struggle is real because the market now rewards speed while the tooling punishes continuity. We need systems to save time, but the systems themselves keep demanding our time. We need tutorials, but the tutorials expire like supermarket sushi. We need AI to help us catch up, but that same AI sometimes hands us directions to a room that got demolished last night.

Progress Has Two Enemies

That leaves you with an awkward truth. Progress now has two enemies. One is ignorance. The other is churn. Everyone talks about the first one because it sounds noble. Read more. Learn more. Adapt more. No disagreement from me. But the second enemy is nastier because it looks like productivity while it actually steals your day as you’re fighting technological entropy to get your automated workflows back into submission. Everything worked fine last night. You can only pray it all works again before dinner.

It’s easy to confuse motion with progress because the screen is busy and your brain feels hot. I keep telling myself I’m building capability, but often I’m just paying a tax for renting my business on other people’s ever-moving tech stacks.

Welcome to the future of work.

Bring curiosity, yes. Bring grit too. But bring suspicion. If a tool changes faster than your business can benefit from it, you’re not using the tool. The tool is using you.

Back to work. I have 2,000 contact records to fix.

Jurgen, Solo Chief.

P.S. How many hours did your tools steal from you this week? I’d love to hear.

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

Most Claude Code projects are just tents. I’m slowly building a city.

Your overnight agentic hackathon is probably not a viable AI agent architecture.

A few weeks ago, I realized my AI infrastructure sucked.

After a year of experimenting with LLMs, I had ChatGPT, Claude, and Gemini all cross-contaminating each other’s project contexts. My data looked as if it had been distributed by a hand grenade across file systems and cloud services. I used Claude, Cowork, NotebookLM, and Perplexity like a five-year-old uses a box of crayons. And I had workflows and business processes flowing and connecting like noodles in a bowl of ramen someone dropped from the tenth floor.

If I ever wanted to scale my business, I needed to stop treating my architecture like a junkyard.

Does that sound familiar?

The Lasagna Principle: Why Your AI Agent Architecture Needs Layers

Here’s what I’ve learned from years of building software and months of wiring AI agents together: when you’re the only one responsible for the whole stack, sloppy wiring isn’t just a nuisance. It’s a liability you carry alone.

Every solo operator I know who’s tried building an AI agent orchestration stack hits the same wall. They start with one automation. It works. They add a second. Still fine. By the fifteenth or sixteenth, they’re staring at a clump of Christmas tree wiring where a Slack message somehow triggers a scraper that writes to a database that kicks off another workflow that occasionally emails a client the wrong welcome email.

Software engineers addressed this problem decades ago with the N-tier architecture. The idea is old and boring, which is exactly why it works: you separate your system into layers, and each layer only talks to the one directly above or below it. Your interface doesn’t know how you store your data; your data doesn’t care how your workflows run. Each layer minds its own business.

I think of it as lasagna instead of spaghetti. Similar ingredients, completely different structure. And when something breaks (because it always will), you know exactly which layer to open up and poke around in.

For my AI agent orchestration stack, I keep it simple and use just three layers: a UX layer where I interact with the system, a workflow layer where automations orchestrate the work, and a persistence layer where state lives and persists.

The Scenario: One Request Through Three AI Automation Layers

Let me walk you through what happens in each of these three layers when I add a new publication to my Substack tracking system.

UX Layer

Everything starts with a human doing something simple. In this example, I paste the URL of a Substack publication I discovered into a Slack channel. That’s it. I don’t configure anything, don’t fill out a form, don’t open a dashboard. I just drop a link and go back to whatever I was doing. Five seconds.

This is the UX layer: the surface where a person touches the system. It could be a Slack message, a Claude Cowork command, a voice prompt, even a button on a vibe-coded web app. The point is that this layer knows nothing about what happens next. It just accepts input and, eventually, shows output.

I could use Signal or Telegram instead of Slack, of course. In fact, I could use any of these tools to do exactly the same thing. When I’m already in the middle of a session with Claude Code, I can just ask Claude to store the Substack URL. As long as the different tools have access to the same workflow layer underneath, it makes no difference to anyone.

That’s the part that lets me sleep. If Slack goes down or changes its API tomorrow, I lose a doorbell, not the house.

Workflow Layer

Below the surface, a Make scenario wakes up. It catches the Slack message, extracts the URL, calls an Apify scraper to grab the publication’s metadata (name, description, follower count, subscriber stats), and routes the results to the database. If the scraper comes back empty (happens about 2–3% of the time), a fallback kicks in: a Claude Haiku call that reads the raw HTML and extracts what it can.

This is the workflow layer, and it’s the heart of any AI agent orchestration setup. It’s where the logic lives: the if-this-then-that decisions, the error handling, the retries. The workflow layer doesn’t store anything itself, and it doesn’t talk to humans directly. It just orchestrates. Receive a trigger, do some work, pass the result. Its job is connecting layers, calling external services, and sometimes triggering other workflows in sequence.

Instead of Make, you can use an automation tool like n8n or Zapier. The point is that the layer responds to UX signals, interacts with proprietary data and external services, and then returns some feedback to the UX.

Persistence Layer

At the bottom, I use Fibery to receive the data and do what databases do: persist state. A new Publication record gets created along with the scraped metadata. A Snapshot record captures the follower and subscriber counts with a timestamp, so I can track a publication’s growth over time. Several Formula fields automatically compute things like “days since last scrape” and “average weekly growth” from the Snapshot history.

Instead of Fibery, I could use Airtable or Coda to store the data. If the data is less relational and more document-oriented, I might opt for Google Drive or Obsidian. In fact, I use multiple backend tools for data persistence because, as Claude said to me, “Trying to cram everything into one tool is like trying to store wine, cheese, and laundry in the same closet. You can do it, but you probably shouldn’t.”

Here’s what matters: the persistence layer doesn’t know or care that Make or n8n or Zapier sent the data. It could just as easily have come from a Python script or a manual import. The schema is stable regardless of what’s happening in the layers above.

Why AI Agent Orchestration Needs Layers

So why bother with this layering when you could just tell your coding agents to wire Claude Cowork and Notion together and call it a day (or call it a night, in case you were sleeping)?

Because things change. I’ve already swapped my scraping approach twice in three months. First, a direct HTTP call, then Apify, then the Haiku fallback on top. My Fibery schema didn’t notice. Neither did my Slack channel. Each layer absorbed the change without infecting the others.

That’s the real payoff of proper AI agent architecture: you can replace any layer without rebuilding the whole stack. You can swap tools in and out without having to go back to Cursor or Claude Code for a minor change and use a ton of safeguards to ensure that it leaves everything else intact.

On top of that, the feedback of any workflow could end up anywhere. In the example I give here, the Make scenario always sends the feedback to Slack because I’m keeping it simple. But I could just as easily let the workflow notify me through Signal, WhatsApp, or Messenger. The orchestration layer decides. The UX layer just listens.

Side note: I gave Claude access to Fibery and Make so that we’re designing the database schemas and workflow scenarios together. I create stuff and Claude reviews it, or Claude creates it and I tweak the details. All corrections and improvements go back into the Claude skills so Claude gets better at it over time.

Where It Hurts: The Cost of a Layered AI Workflow

I’d be lying if I said it’s all upside. Having more layers means more things that can break independently. Debugging a failure that starts in Slack, passes through Make, and surfaces as a wrong value in Fibery is like playing a game of telephone with robots. It takes longer to set up than a quick-and-dirty single-tool solution. And there’s a constant temptation to over-engineer, to add a fourth or fifth layer “just in case” when what you actually need is to ship something.

I’m building slower than the hammock crowd. But at 8 AM, when something broke and there’s nobody to call, I want to know which layer to open. That’s worth the extra time.

The real question every Solo Chief has to answer for themselves: how much structure is worth the slowdown? I don’t have a formula for that. I only know what my tolerance for 8 AM debugging sessions is, and it’s getting lower every year.

Where I Cheat

Let me be honest here, because maps should include the shortcuts people actually take. The architecture I described is a work in progress. I know how things should work, but that doesn’t mean I always follow my own guidelines. Sometimes, the shortcuts are just too attractive (and mostly harmless).

For example, Fibery offers great visualization and management features on top of its databases. It’s incredibly easy to create tables, workflow boards, reports, and other overviews of the Substack publications I collected there. Following my own advice, I should be designing and building such user interfaces with separate tools and then retrieving and acting on this data through the middle layer. But I’d be giving myself a lot of work to build something that the database tool already offers right out of the box. Why bother building a status change workflow in Make when I can simply update that field directly in Fibery?

For now, I make a conscious choice to cheat. But only as long as I’m the sole user. The single entrepreneur. The Solo Chief. I get special privileges because I give nobody else direct access to the backend. If I ever decide to get others involved in the Substack publications data I’m collecting, I’m going to make sure to build a web app that can only fetch and change the data through the workflow layer, never directly. I might even use Claude Code for that.

From Spaghetti to Lasagna: Building Your AI Agent Stack

Am I happy with the AI agent orchestration architecture I have now?

Not yet. I still have too much spaghetti and too little lasagna. But the direction is clear, and that makes all the difference. Every update I make to my agentic orchestration must satisfy the N-tier architecture. Every new piece I add goes into one of three places: UX layer, workflow layer, or persistence layer.

Now that I have the horizontal architecture fleshed out, I can start separating the agents into their own vertical domains: one agent for Nonfiction Writing, one for Finance & Admin, one for Personal Branding, and so on. Each with their own objectives. Each with their own layers. My agentic organization will look less like a junkyard and more like something I’d trust to run while I’m asleep. I’m slowly building a city while others are just rapidly setting up tents.

I’m a big fan of lasagna, but I rarely turn down a plate of spaghetti. It all comes down to where you want to end up with your business. An unorganized bunch of wires will work fine as long as you’re still experimenting. But for anything more serious and viable, you’ll want to think about giving your AI agent orchestration stack a bit more structure.

Your overnight agentic hackathon is probably not a viable AI agent architecture.

Or, you know, just enjoy the spaghetti. I’m not your mother.

Jurgen, Solo Chief.

P.S. Show us what your architecture looks like. I’m curious!

Both exploration and execution feel productive. That’s why you’re stuck in the wrong one.

Eight concrete signals that tell you when to stop exploring and commit, or when to abandon execution and pivot to something new.

There’s a particular kind of productive-feeling paralysis where you’re working hard, shipping things, and somehow going nowhere. You’re executing, but on the wrong thing. Or you’re exploring, but past the point where exploration is still honest.

For years, I ran a company that was constantly half-exploring, half-executing, feeling vaguely but not entirely productive. I was going wide when I should have gone deep. Or I was executing for too long when I should have pivoted back to experimenting. There was nobody to catch that drift but me, and I was too close to the work to see it.

My energy was perfectly distributed across things that were only half-working: either I was lacking focus or I didn’t have enough unfocus. I wasn’t discovering enough, or I wasn’t delivering enough. (Sometimes both in the same week. Truly a gift.)

The particular hell of exploration versus execution is that both states feel like work. Exploration feels productive. Execution feels productive. The signal that you’re in the wrong phase is subtle, and there’s no one in the next office to notice it before you do.

Do you feel like a pinball perpetually bouncing between discovery and delivery?

Many experts will tell you to balance exploration with execution. Breathe in, breathe out. It feels like very calming advice. It’s also mostly useless to someone whose livelihood depends on getting the call right. Your business is not a Pilates exercise. You’re probably not running your company from a yoga mat. (If you are, congratulations; I’m genuinely envious.)

As single managers, solopreneurs, and one-person businesses, we don’t need balance all the time. We need signals. We need alerts that tell us when to stop exploring and start executing, or when to stop delivery and pivot back to discovery.

Signs You Need to Stop Executing (and Go Wide)

Sometimes, too much focus is the danger. Here’s what it looks like when that’s happening.

The first signal is what we might call the input/output inversion. You’re working harder, but the numbers aren’t moving. Output is happening, but outcomes are not. You’re getting better at executing the wrong thing, which is its own special form of masochism. Effort with too little traction is terrain feedback. The proper response is to stop digging deeper.

The second signal is assumption collapse. Every mission rests on a core premise, some belief about why this idea should work. When that premise turns out to be false, continuing the mission becomes stubbornness dressed up as focus. You can’t pivot within the same box; you need a new box. The answer is to go wide.

A third signal is often more uncomfortable to admit: persistent, recurring curiosity about the same alternatives. You feel there’s another way, but you’re suppressing it. You invested so much in this mission! Stay focused. Ignore the distraction! But if the same adjacent possibility keeps nibbling at your attention for weeks or months, your subconscious is pattern-matching something your deliberate focus is blocking. Occasional curiosity is noise. Recurring curiosity about the same thing is a message.

And then there’s the fourth signal: boredom. I used to think boredom was a weakness, a sign of an unmotivated mind. But evolutionary biologists will tell you that boredom is important. It might have evolved to prevent over-harvesting. Your brain signals boredom not to annoy you, but because the learning curve is flattening. The berries are running out, so move to a different patch. In the exploration versus execution dance, boredom during execution is your body’s way of saying the current vein of ore is depleted.

Signs You Need to Stop Exploring (and Commit)

The opposite trap is equally seductive: exploring long past the point where it’s still honest. Here’s what that looks like.

The first signal is when you can name the constraint. If you can answer in one sentence “what is the single thing limiting everything else right now?”, exploration has become a distraction. Constraints demand execution. The Theory of Constraints is all about this phenomenon: find the bottleneck, exploit the bottleneck. The moment you see it clearly, you’re not exploring anymore. You’re procrastinating.

A second signal is convergent evidence. You’re exploring multiple paths, and they keep pointing to the same place. You talk to five different people, and they all mention the same pain. Three separate experiments produced a similar finding. The terrain is clearly telling you something. At that point, more exploration isn’t learning. It’s avoidance of a conclusion you’re unwilling to face. I’ve been that person, gathering “just one more data point” while the answer stared at me from the whiteboard. It’s not a good look.

The third signal is what I’d call the sufficient clarity test: can you write a falsifiable hypothesis? “I believe X will produce Y for Z people within timeframe T.” If you can state it clearly, you can test it. If you can test it, there’s no longer any reason to keep fooling around. The remaining reluctance is usually fear of failure disguised as intellectual curiosity.

The last signal is a little embarrassing: clarity plus resistance. The next step is obvious, but you find it uncomfortable. When you already know what to do and you’re not doing it, that’s not uncertainty. That’s something else entirely. You should keep exploring if the terrain is genuinely unknown. But you must stop when you know the terrain and you’re just scared of settling down. Resistance is strongest when the work matters most. If your reluctance to commit feels almost physical, you’re probably standing right where you need to dig.

What makes this especially tricky when you’re a Solo Chief: you’re the one generating the signals and the one who has to interpret them. There’s no devil’s advocate in the next chair. Cognitive bias runs unchecked until the business tells you, usually in cash.

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month. Subscribe or upgrade.

Subscribe now

The Two Decision Triggers (Pivot Switches)

Here’s the sharpest formulation I’ve found for the when to pivot my business question. Kill execution the moment you learn something that invalidates its core assumption. Not a detail. Not a preference. The assumption itself. If the “why this should work” collapses, stop doing. Start learning. And kill exploration the moment you can write a one-page plan with a goal, a metric, a next action, and the only objection left is “what if it fails?” Fear of failure is not a signal to explore more. It is the signal to commit and move forward.

The pattern many people follow instead (and I’m trying very hard not to look in the mirror here) is to oscillate randomly between exploration and execution, respond to feelings, and call it adaptation. The scenic route has its charms. I could write a book about scenic routes.

Or you can be mindful of the signals and use them to dominate a business terrain faster.

The above applies to every kind of thinking work, but it applies especially to Solo Chiefs, who don’t have an executive team to hide behind when they switch for the wrong reasons. There’s no board to blame. No team to debrief with. The signal-reader is you. So is the one who has to live with the call.

Exploration versus execution. Your job isn’t to balance them as suggested on some corporate wellness poster. Your job is to recognize when to bounce from one to the other. The outcome of your good judgment is the balance that you seek.

Jurgen, Solo Chief.

P.S. Which signal do you tend to ignore the longest? I’d love to hear.

The post “Teamwork Is a Scam“ marked my 100th post on The Solo Chief. And I feel this is just the beginning.

If you’re a paying subscriber, you’ve been keeping this publication alive. You fund the caffeine, the independence, and my habit of telling people the universe is conspiring against them.

I owe you more than a PDF and a thank-you email.

Next week, I…

Read more

Most AI agent content is about building AI agents. That’s not a business, it’s a pyramid scheme.

Here's how I want to become an agent boss: by building and managing AI agents for real value, not meta-content recursion.

It’s 4 AM. I’m staring at the ceiling again, and somewhere in the dark I’m doing the mental arithmetic of how many AI-generated playbooks some author on Substack sold this night while I was busy thinking about something more meaningful—and not getting anything done.

Do you feel a bit anxious when your competitors use Cursor, Codex, Claude Code, or Antigravity to launch one vibe-coded app after the other, funneling market attention toward their throwaway solutions and away from your lovingly handcrafted, far-too-long-in-development product?

Do you sense a Fear of Becoming Obsolete (FOBO) when other solopreneurs have a legion of OpenClaw AI agents iterating rapidly on a dozen business models in parallel while they’re sleeping, and while you only have a smartwatch reporting each morning on your restless night?

Don’t worry. I feel the same.

Sleepless Over the Bestseller Badge

Reading Substack is like doomscrolling through your Instagram feed, noticing that each and every influencer sliding down your screen is better-looking and more successful than you are.

Just look at those leaderboards, bestseller lists, and screenshots of rapidly climbing subscriber stats, embellished with the inevitable humblebrag note, “Thanks to my subscribers. I could not have reached this milestone without you! Here’s what you get when you upgrade. [insert upsell here]”

Swiping as an author across the Substack feed can be an inferiority-inducing anxiety trip. So many others are doing so well! They’re posting and sharing and hustling and playbooking themselves into stardom. It makes you feel ugly. Unpopular. Inadequate. The Quasimodo of creatives.

Few people talk about what it actually feels like to scroll through a Substack leaderboard after midnight when your own subscriber count metric looks like the sun setting over a windless ocean. I’m sure some founders, freelancers, and solopreneurs feel more alone than ever as they measure themselves against leaderboards, whether on Substack or anywhere else.

And yet, no matter how many of those self-congratulatory wins, comments, bestseller badges, and “download-my-framework” posts I see, the only thing I can think is … No.

That’s not how I want to play this game.

The hustle approach doesn’t fit me. (It never did, honestly, but FOBO has a way of making you question your own instincts.)

I Want to Be a Sleepy Agent Boss

But then … what is my approach?

Before we get into that, let me share the thoughts I’m endlessly mulling over while my smart ring is busy recording yet another sleepless night.

I envision a Substack Publishing assistant (let’s call it Lenny) that helps me grow my Substack publication. Lenny does all the work I have no time for, like cross-linking my posts, recommending other authors, scrutinizing the statistics dashboard, and comparing the results with LinkedIn metrics and Google Search Console reports. Lenny is smart enough to share with me only what matters, considering the objectives we’ve agreed upon together. While I’m producing, Lenny is optimizing.

I dream of a Nonfiction Reading assistant (let’s call it Marie) that monitors the English-language nonfiction book market for me, notifying me of new books that seem to satisfy my needs and taste. Marie keeps track of the books I read, the ratings I give them, the ones I didn’t finish, and the backlog I still have. She will sternly push back on my tendency for tsundoku when I’m tempted to order an author’s latest work when their previous two books still sit untouched in my reading queue.

I want a Finance & Admin assistant (let’s call it Warren) to take care of my monthly invoices: collecting receipts from my mailbox, digging them up from the billing catacombs of three dozen online services and digital platforms, and then matching the documents with the transactions on the credit card statement it retrieved from my bank. Warren will also politely hunt down any unpaid invoices and will suggest how to set aside the surplus balance in my bank account—if I ever have any.

I need a Mind & Body assistant (let’s call it Jane) to monitor the streams of health data collected by my smartwatch, smart ring, and the nutrition app I use to track my daily food intake. Jane will follow my exercise regimen, vitamins, minerals, caffeine shots, step count, and sleep stages to give me tips on how to prevent waking up at 3 AM with a brain that keeps generating mental images of unrealized subscriber counts.

I absolutely want various digital assistants helping me invent and build out business ideas that result in actual value generated for customers, not some “exclusive access” to prompt libraries, courses, and AI-generated playbooks targeted at people like me who are unlikely to ever hit the bestseller leaderboard.

At 4 AM, as I turn around yet another time, I dream of a personal staff of AI agents specialized in niche contexts such as Public Speaking, Personal Branding, Board Work, Fiction Reading, Fiction Writing, Travel & Vacations, Friends & Family, Personal Finance, Home & Living, Food & Cooking, and more.

And I need two Chiefs of Staff (let’s call them James and Andy, one for personal matters and the other for professional jobs). They’re familiar with all the specialized assistants and can query them directly about ongoing work, backlogs, commitments, calendar items, and deadlines. When I tell Andy I want to use a dedicated credit card for all my nonfiction book orders, Andy will work with Warren and Marie to make it happen. When I tell James we’re out of milk, James will work with Martha and Jamie to find an alternative for my morning cappuccino.

That’s the dream I have when I lie awake.

I want a personal AI agent architecture. The last conscious thought I have, just before I finally fall asleep at 5 AM, is I want to be an agent boss. One that can actually sleep.

None of that exists, though. So it’s time I started building.

But what I don’t want is to spend two years building an architecture so elegant it belongs in a museum, while everyone else shipped products to actual humans. I’ve made that mistake before. More than once. (Maybe three times. Fine, four.)

I’m a founder, intrapreneur, and former CIO rethinking governance for the one-person business, navigating sole accountability in the age of intelligent machines—informed by plenty of scar tissue. All posts are free, always. Paying supporters keep it that way (and get a full-color PDF of my book Human Robot Agent plus other monthly extras as a thank-you)—for just one café latte per month.

Subscribe now

1) Only in the Cloud, Not on My Computer

I almost hear you hopping up and down excitedly, saying, “That sounds great! You can use Claude Code, OpenClaw, LangChain, CrewAI, …”

Wait, wait, wait! #soundofpullingthebrake

Before anyone helpfully suggests I just fire up OpenClaw or CrewAI and call it a day, here’s the part that actually matters.

First and foremost, I want nothing running on my local computer. Ever. Nada. Zip. Null.

With me frequently traveling for work and moving weekly between homes in Rotterdam and Brussels, the last thing I want is for my digital assistants to work with local files that may or may not be in the right file folder, and may or may not properly synchronize across multiple computers that may or may not actually be turned on.

I’ve already struggled these past few weeks with Claude Cowork getting confused because of out-of-sync skills files, and insisting on a local home folder on my PC that is not the home folder of Google Drive or Dropbox. (Neither can be easily changed.) And copy-pasting local files from one home folder to another is an excellent way of inducing a migraine and another sleepless night.

Let’s not go in that direction. I will happily pay for tokens and credits when it means I don’t have to worry about the infrastructure. I’ve spent enough years of my life juggling with deployment environments to know that the cheapest thing I can buy is to make this someone else’s problem.

All my assistants must live in the cloud.

2) Real Use Cases, No Selling Spades to Gold-Diggers

I want my digital assistants to do real work. Not work about work.

You probably know the category. E-books about writing e-books, podcasts about making podcasts, YouTube channels about how to grow your YouTube channel. We can call this type of business model monetized recursion engines: the endless parade of make-money-online (MMO) schemes offered by meta-content creators.

It’s the equivalent of selling spades to gold-diggers, or tickets to lotteries, or online courses for crypto speculators. Monetizing other people’s desire for fame and money is an all but guaranteed path to profits. (It’s also the business model equivalent of a snake eating its own tail, but I suppose snakes don’t need a Substack leaderboard.)

I don’t want to earn revenue from people who hope to earn revenue. For every successful creator, there are a thousand who can only dream of being similarly successful. If the successful one is successful only because they charge the thousand unsuccessful wannabes for hope and illusion, they don’t have a real business. They run a pyramid scheme.

“I don’t want to earn revenue from people who hope to earn revenue.”

I refuse to write a Substack about how to build AI orchestration around Substack. I want my AI agents to handle real business use cases. The kind where actual value gets exchanged with actual humans for actual reasons.

3) Context Isolation, No Cross-Contamination

One only needs to read a few Substack posts discussing OpenClaw to see the horror stories of AI agents ruining a job, a data repository, or even someone’s reputation, when given unlimited and unconstrained access to someone’s personal computer.

I don’t want Marie to know my credit card numbers, or Warren to have access to my health data, or Jane to see my customer notes. Last time I checked, I counted at least 20 personal and professional contexts and each of them should have its own dedicated AI agent specialized in that and only that. My travel details are irrelevant to the AI agent managing my home improvements. And I don’t see the need for the Personal Finance agent to have access to my online dating profiles.

Think of it like biological cell membranes. Each cell has a clearly defined boundary, lets only the right molecules through, and doesn’t randomly share its DNA with the cell next door. Without that membrane, you don’t get a functioning organism. You get cancer. I want my sleepy agent boss architecture to work the same way: strict context isolation, preventing cross-contamination of contexts.

This will definitely slow me down. But it should also keep me safe.

4) Layered Architecture, Lasagna Instead of Spaghetti

Call me old-fashioned, but I’d rather have lasagna than spaghetti, and I’d rather understand the layers than inherit a bowl of someone else’s noodles.

I’m no fan of the idea of a swarm of AI agents autonomously coding an entire tech stack all by themselves without any insight into how the various layers of the product connect to each other. I’ll be happy to let the assistants help design and build the various parts of my agentic architecture, but I want to stay in control. I want to own the interfaces between the layers.

There’s a lesson from systems theory here: in any complex adaptive system, the interfaces between subsystems are where most of the interesting (and dangerous) behavior emerges. If you don’t own those interfaces, you don’t own your system. You’re just renting it.

“If you don’t own those interfaces, you don’t own your system. You’re just renting it.”

Is that slower? Yes. Is that more responsible? I think so, yes. Will I regret this principle in six months when I’m still debugging layer two while some 22-year-old has shipped fifteen agentic products from a hammock in Bali? Probably. But I’ve watched enough cowboy architectures collapse to know that speed without structure is just expensive failure with better marketing.

5) Multi-Platform, No Vendor Lock-in

Last but not least, I prefer to run a business that’s not intimately tied to one company’s agentic architecture.

It’s hard enough not to rely on Google for everything these days. I have no intention of making things worse by building the entire stack of my agentic workflows through a single supplier. I want MyClaw and Claude Code and AI Studio and Make, just like I use (and pay for) the top five cloud LLMs precisely because I don’t want to be dependent on any single one of them.

Different contexts require different persistence layers, too. For some data, Fibery is perfect. Other things are better stored in Notion, Trello, Google Drive, or Roam Research. Trying to cram everything into one tool is like trying to store wine, cheese, and laundry in the same closet. You can do it, but you probably shouldn’t.

Instead of going all-in with one specific vendor, I want to go deep in my understanding of how not to tie the future of my business to one exclusive technology.

Yes, all these requirements mean I’ll probably be the last Solo Chief on my block to have a fully functioning agent stack. I’ve made peace with that. (Okay, mostly made peace with that. Ask me again at 4 AM tomorrow.)

It’s going to slow me down, but it will also increase viability.

The Future Agent Boss

It’s time to feel less anxious about the future and start building it around ourselves. We should stop comparing us with others who only seem to be more productive and successful, but who are really running monetized content recursion engines and don’t share our strict requirements for a viable business: a cloud-first architecture, context isolation, layered design, and supplier independence.

I haven’t built any of this yet. I’m describing the map I’m drawing, not territory I’ve conquered. Some parts of the map are probably wrong. I’m okay with that. Maps are supposed to get revised. That’s what separates a map from a manifesto.

In the future, I want to tell Taylor (my future Personal Branding agent) to update the photo, tagline, and description across all my social media profiles, once every three months, to reflect the most important thing I’m working on at that time. If Taylor sees a need to employ a specialized agent per platform, I’d probably say that’s a smart move.

In the future, I want to tell Johanna (my future Nonfiction Writing agent) to check my entire archive of articles and make a proposal for turning the strongest posts with the best engagement into a book. Johanna will have to collaborate with Lenny on that because Johanna will only have access to my books, while Lenny will own just my Substack archive.

In the future, I want to tell George (my future Fiction Reading agent) to find me the ten best novels on dystopian parallel universes. George will probably have its own team of subagents for checking which books I’ve already read, reading reviews to match books with my taste, and ordering hardcover copies that won’t break my bank.

And when I have an idea for a new business model that needs a new context, I’ll be working with Andy to see which new professional agent to create or hire.

And I’ll be the sleepy agent boss: someone who builds, delegates to, and manages AI agents to amplify their impact, working smarter, scaling more deliberately, and taking control of their career in the age of AI. Not a hustle-bro with a leaderboard screenshot. Not a meta-content creator selling gold-digging spades. A sleepy agent boss with a map, a lasagna architecture, and a smart ring that records a good night of sleep.

I’m starting to build now. I’ll share what I learn, including the parts where it blows up. Especially those parts, actually. Those tend to be the ones worth reading.

Jurgen, the Solo Chief

P.S. If this reads a bit like me rambling, it’s probably because I just crossed the Atlantic and I am a little sleep-deprived.

Share