Since roughly // broadly 2020, I’ve been running a solo-powered minor empire. I have no employees, and my only office is my home office, filled as it is with cat hair and various comic books. My business is: me, a laptop, a set of AI tools that scale the parts I hate, and a personal network. I’ve done brand strategy, naming, GTM, messaging, content and growth marketing for a list of folks I respect and genuinely give a shit about - SaaS cos, VC firms on 3 continents, and an initiative I still believe has the power to change the world.

And I’ve done it all without once - ever - wishing I had a team, or wishing I was any “bigger” than I am.

Well, I’m coming up on the 6th anniversary of going my own way, and I wanted to take a moment to put down on paper what worked and why, how I think about the future of what I do, and why the structural economics of being a solo-operator are only going to improve from here.

The trad agency model is a staffing arbitrage play. You hire 23-year-olds, bill them out at senior rates, and scale with automation. You technically lose money on the first project with each client (the first being the only time senior staff do the work) and make it up on the long-tail retainer work that hopefully follows. Founders sell, juniors grind, and middle managers translate between the two groups. To varying degrees of success...if you’ve ever found yourself wondering why so much agency work feels so mediocre, when those same agencies are winning awards etc, this is why. The person who sold you on the engagement, the person entering the Lions and the person doing the work on your actual brand are almost never the same person. There’s an incredibly lossy compression mechanism in between, and it strips out most of the taste, judgement and contextual awareness that made the pitch compelling in the first place.

This model is the go-to because - thus far - it has been the only way for an artisan to scale their work to the level of financial freedom // validation the entrepreneurial set promote as the be-all-end-all of human existence.

When I started Studio Self, I had just left a tech startup who had been acquired by MYOB, and along the way we’d struggled with precisely that experience: attempting to scale our own capabilities by hiring someone else, only to find ourselves mired in a level of mediocrity that can only be defended by a recognisable brand. My goal was to eliminate that experience entirely.

The thesis was that creative agencies should be small. They should be born small, and they should stay small, and they should be entirely focused on ~the work. The agency is a different beast to a tech startup, and it should act as such. Call it an anti-LARP manifesto.

There has never been a translation layer. There’s no game of telephone between “what the client actually needs” and “what the junior assigned to the account understood from the brief.” I have shipped every engagement with full context intact, from first meeting to final deliverable.

I have made a deliberate choice to not scale the creative aspect of my work. Note the word “creative.”

For the first few years, that meant my workload was roughly 1-2 clients max.

The constraint on solo operations has always been bandwidth; one person can only do so much in a day, and the non-creative operational overhead of running a business (invoicing, scheduling, CRMing, formatting, bookkeeping, task // project management etc) eat into the hours and the cognitive bandwidth available for actual creative work. I challenge anyone who has spent more than an hour a day looking at a spreadsheet to pull a piece of good actual Work out of their ass. In my experience, the work suffers any time the creative mind is faced with documentation.

This is dangerous, because the appeal // selling point of an agency is the creator who drives it. Call this the Mind (the originator of the business.) They and their specific output are the product.

In a traditional agency, you solved this by hiring operations people. And of course, to justify the cost of the ops folks, you then hired other creatives, and the agency enshittification almost inevitably began. Adding layers between the Mind (the originator of the business) and the client inevitably leads to a decay in quality. If you wanted to maintain solo-operator status, you solved it by working nights // weekends and then slowly burning out. The road to alcoholism is (believe me) paved by folks who attempted to do it all.

LLMs have changed this math completely.

The choice a creative makes about how they use AI is critical, and I think it can evolve, but you need to be clear about it. The distinctions between what you do and don’t outsource to a thinking machine with a non-apparent brain matter enormously, and I worry that most folks running solo businesses are getting this - shall we say - a little backward.

I have a simple formulation when it comes to AI:

My purpose in my work is to be creative.
My purpose is to enjoy my work.
I enjoy the work that is creative.
I do not enjoy the work that is not creative.
Therefore...well.

I use AI for task and project management, operations, proposals, documentation, formatting, scheduling, meeting transcripts, project timelines, email triaging (but never, ever replying or management), and the thousand small tasks that used to consume over 50% of my working week. Example: setting an agent loose on my Google Drive to clean up a folder of poorly filed (whoops) drafts, or to sort out unpaid invoices, or to capture and share receipts with my accountant. I use AI to surface topics and ideas from Hacker News and other sources that I don’t want clogging up my personal RSS reader. I apply automation heavily to this work because there is nothing creative about it, and it brings me no joy. The quality of that output is probably better than I produce manually, because AI doesn’t go down a mental tangent at 4pm on a Monday while staring at invoices and become submerged in self-hatred. I am liable to do that. I use AI for coding and dev work - largely through Mistral - and I’ve found it invaluable in shaking the dust off the coding skills I first learned modding Wolfenstein 3D at the tender age of 14. Lastly, I use it to come up with YouTube titles and descriptions, and I use Descript for AI-assisted editing.

But I don’t use AI to write copy, or to develop brand strategy.

I have been a heavy user of Grammarly in the past, and I’ll admit to that - once upon a time, it was like having an editor in your pocket. But their increased reliance on AI has made their editing tools next to useless and any output run through them has become homogenous and unreadable, and after a review of the last 6 months of work, I’ve finally kicked them to the curb. I’d rather my work include however many technical imperfections Grammarly might have fixed, as long as I can avoid Grammarly’s phrasing proclivities...

I don’t use AI to make taste-based decisions about what a company should sound like, or look like, or feel like. And the reason isn’t actually a romantic attachment to the artisanal purity of human creativity, as much as I might believe in that - it’s entirely strategic.

The market is already flooded with AI-generated “creative” work. You can feel it in the sameness of the YC landing pages, and how every Series A-B deck now has the exact same tone of voice. Everything has started to blur together, and nobody sounds specifically like themselves. It’s beginning to feel like everyone has the same marketing team, because everyone’s marketing team uses the same agency, and that agency is ChatGPT. I’m not anti-AI, but I’d certainly count myself amongst those who dislike bad outputs...

My theory is that messaging and work that sounds like it was produced by an actual human with actual taste and actual opinions (even if those opinions run the risk of being ~wrong) are going to stand out more than ever. Which is, I suspect, going to be seen as the great irony of the AI era. The technology that makes it trivially easy to create competent creative work simultaneously makes merely competent creative work nearly worthless. The value transfers to the staff that can’t be generated: points of view, idiosyncrasies, judgements, strong opinions. The premium shifts from execution to context, and personal // individual context will always beat documented context.

Taste is - almost by definition - a solo-operator product. It doesn’t survive committees and Slack channels and stands. It lives in one person’s brain and how that brain produces work.

So the model I’ve landed on - AI handles everything that doesn’t require creative judgement, and a human handles everything that does - turns out to be accidentally well-positioned for the next decade of competition. I’m faster than most agencies, because I don’t have coordination costs, and I’m better than AI-only solutions because the work I produce has (arguably) a pulse.

There’s a Borges story (isn’t there always) called “Perre Menard, Author of the Quixote” in which a fellow attempts to write Don Quixote word-for-word to the original, but through his own creative process. It’s worth a read. Borges point was that identical outputs can have completely different meanings, depending on who produced them, and why. I find myself returning to this, when I look at AI-generated brand work, sitting next to human-created brand work. They look, sound, feel similar on the surface - but the difference is in the intention. It’s in whether someone actually made a choice to put a comma here or there, or whether an algorithm produced the most statistically likely arrangement of words. The right clients know the difference, even if they can’t always articulate it, and they’re increasingly willing to pay for it.

The history of business is (and I’ve ranted about this before) a history of declining coordination costs. Ronald Coase, in his 1937 paper “The Nature of the Firm” argued that companies exist because of the transaction costs of coordinating work. You hire people instead of contracting everything out, because managing 100 contracts costs more than an org chart etc. But every time coordination costs have dropped, the optimal firm size has dropped with them. IE - the internet made it possible to run a 10 person company that would have needed 50 people in 1989. Cloud computing and SaaS tools made it possible to run a 5 person company that would have demanded 10 in 2005. AI is making it possible to run a 1 person company that would’ve asked for 5 in 2020. Etc.

I think it’s fair to assume we’re still early on that curve.

The tools available to solo operators today are absurdly powerful compared to even 1-2 years ago. I can generate documents, produce financial models, deploy websites and self-hosted tools, manage multi-channel marketing campaigns, handle customer support and run project management workflows without hiring anyone. This is a boon to the creative side of my brain that hates every single one of those tasks.

What hasn’t changed - and what I don’t expect to change for a long time - is that the top of the value chain still requires human judgement. The lower-level work is getting cheaper, and probably should; but for experienced folks who know how to sell that experience, the premium is rising. Judgement dense, taste-heavy, context-dependent tasks justify a price at a multiple of a Claude Max subscription, and for good reason. Which creates an interesting (to me, at least) opportunity: a single operator who uses AI to handle the entire operational substrate of a business while personally delivering the high-judgment creative work that clients actually value. The margins on this model are high, because the revenue scales with the operator's expertise and reputation while the costs stay essentially flat.

There’s a “cult of personality” aspect to this, too. I think we care more about individual founders and what they represent than we used to. We invest in companies because a human being has become the face of the operation, and we like that face, and we like the things they say. You can see this playing out everywhere from Mr. Beast to Elon Musk, and no matter how distasteful you or I might find their brands, it’s a real phenomenon. Arguably, it’s a good chunk of what makes my own business work: people like me. They like the way I think, and they want to associate with it or apply it to their own brand.

Could this model produce a billion-dollar solo business? I don’t think the math is as crazy as it sounds. I’m not sure it’s something I have any interest in pursuing, but I find the idea fascinating...

Software companies with zero marginal cost products have shown that revenue can scale independently of headcount - if a solo operator builds a productized service // SaaS tool // content platform, or they can pull off a brand licensing operation on top of their core expertise, and they can use AI to handle everything else, the revenue ceiling starts to look very different from what we've historically associated with one-person operations.

You don't need to sell a billion dollars worth of consulting hours. You need to build a leveraged asset on top of consulting insight.

I'm not predicting this will happen next year, but the structural barriers that made it impossible are being removed one by one, and faster than most folks seem to realize. The first person to do it will probably be someone who looks a lot like the current generation of solo operators: technically sophisticated, creatively excellent, strategically sharp, and running their entire operation on a stack of AI tools that didn't exist three years ago.

My “mini-empire” is this blog, this email newsletter, products like Kerouac and Distributism etc, powered by a solo agency. I wake up (most days...) passionate about hacking away at all of it, and I deeply give a shit about the clients I have and the words I get to put on the page.

And if I were starting that “empire” today, I don’t think I’d do anything differently. I’d still go solo from day one, and I’d still choose to position myself at the higher end of the market, and I’d still guard my own creative work jealously.

The advice I’d give to anyone else:

1. You need to actually be good at something. Yes, the solo model magnifies skill, but it also magnifies any // all mediocrity. If you’re a B-minus operator hiding in an agency where the brand does the work, going solo will expose your shortcomings with alacrity.
2. You need an actual opinion and an actual perspective on your domain that clients can’t get from a chat bot. I’ve had a fair few VCs and tech “luminaries” advise me to be less opinionated over the years; play it safe, don’t ruffle feathers etc. But that’s the entire value proposition - a real person with a real worldview. If your only pitch is “I’ll do competent work at a reasonable price,” you’ll be automated out of every opp. A chat bot won’t tell you its frank opinion on xAI and Elon Musk. I will, and I’ll do it without being asked twice. This is not a weakness.
3. You need to build systems. As many as you can. And make them legible, too... every repeatable process should be automated, every template refined, every workflow documented. The goal is to make the operational side of your business so efficient that it functionally disappears, leaving you with nothing but the high-value creative work that only you can do.

130 years ago, musicians and orchestras were concerned that the advent of the phonograph would put them out of business. We’re several tech disruptions removed fro that moment, and recorded music has created entirely new categories // genres // grifts of music that were previously impossible, but live music has only become more valuable.

The solo operator is in a similar place.

AI-generated creative work is the recording. Your work is the live performance. And as the recordings proliferate and all start to sound exactly the fucking same, the live performance becomes the thing people will cross state lines to experience.

I’m Six years in, and I don’t think I’m likely to change course now. And I’m frequently surprised by how durable the model has actually turned out to be...I've watched agencies and tech giants alike go through layoffs and restructurings and pivots and identity crises, and I've watched AI startups promising to “automate creative work” launch with terrible fanfare and then discover that their output was too generic to be even slightly competitive.

Meanwhile, the solo model keeps working // compounding. The core value proposition becomes more relevant as the alternatives get noisier and more desperate...

My goal from here is to follow whatever side quests Studio Self happens to turn up, let it continue to support this blog, and keep having conversations with founders and teams who I actually like (important.)

I think the future probably belongs to the small companies // individuals more than sprawling conglomerates. There is a huge opportunity for people who use AI to remove everything that isn't judgment from their workload, and apply that judgement to good products and good services. My theory is that one person, with an AI-augmented operational layer plus taste is the company of the future - and I'd bet on that again and again.

Finally - if you’re interested in talking about how to work with me at Studio Self, you can reply here or drop me a line: [email protected]