This is a longer piece than the others. It's a sort of resignation letter, addressed to nobody in particular, written for the founders and freelancers and agency-runners who might recognise the shape of it.
By the autumn of 2024 I'd spent twenty years running web agencies. The last one had twenty-five staff at its peak. We did good work. Some of it I'm still proud of — I can name three sites we built between 2018 and 2022 that I'd stand behind today. Most of it I'm not proud of, not because it was bad, but because it was fine. Fine work that paid the bills, satisfied the client just enough to not get fired, and quietly disappeared into the long tail of the internet a few years later.
I'm writing this to explain why I left, and what I'm trying to build instead.
What was wrong with the work
The thing I want to say first is that the peoplein agencies are mostly excellent. The designers I worked with were thoughtful, the developers were rigorous, the project managers were patient. The work problems weren't about talent. They were structural.
The structural problems were three:
The pricing punished restraint.A florist who needed three pages was charged for a fixed-overhead minimum that meant five pages was almost the same price as three, so we'd talk her into five. The pricing model rewarded scope creep. The customer ended up with two pages they didn't need, which they'd then ask us to maintain forever.
The handoff was hostile. We'd ship the site at the end of a project, take final payment, and move on. The maintenance retainer was a separate sale that most customers couldn't justify. Six months later the customer would email asking for "a small change" and the agency would either bill the discovery hour to figure out the change, or write it off, or quietly stop replying. None of these were honest options.
The economics excluded the customers we most wanted to help. The £30,000 brand redesign was right for a £5M-revenue company; the £3,000 templated build was right for the same company's marketing team to bash out a microsite; neither was right for the £150,000-revenue florist who needed a working website she could afford to maintain. We'd quote the florist £2,500 because that was the floor of what we could deliver. She'd say no and we'd watch her go to Wix. Two years later she'd be back asking for a rescue. The agency model was structurally hostile to the customer most likely to need real help.
I told myself for years that this was just how the industry worked. That the small businesses we couldn't serve weren't really our market. But I'd grown up watching my parents run a small business — a working café in a small town that struggled, year after year, with the same kinds of operational problems my customers had — and I knew that the businesses we were turning away weren't a marginal segment. They were the actual majority of UK small business by count.The agency model just didn't have a way to serve them.
What was wrong with me
The other thing I want to say is that the agency model wasn't only bad for the customer. It was bad for the people running it, in slow ways.
Agency life, at the level of a 25-person studio, is mostly people-management. You spend your week in 1:1s, in pipeline reviews, in sales calls. The actual work— the design, the writing, the architecture decisions — is done by people you've hired, who are mostly good at it but sometimes aren't. When the work is good you take some pride in it but the credit isn't really yours. When the work is bad you bear the responsibility but you can't really fix it without pulling rank.
The trap of running a creative agency past about ten people is that you become an HR organisation that occasionally produces websites. I didn't want to run an HR organisation.I wanted to make websites. I'd been getting further from the work for a decade.
By 2023 I was sitting in a sales call with a coffee roastery — a couple in their forties who'd saved for years to open a small batch shop in north London — and I was doing the maths in my head while they talked. The maths said: "we can't help you for less than £4,000, and you don't have £4,000."I knew before they finished the brief that I'd be sending them off to find a freelancer. I gave them my polite "this isn't quite the right fit" line. They thanked me. They left.
I sat with the after-feeling for a long time. That couple was who I'd started the agency to help. I was now turning them away because the agency couldn't help them. The agency had grown into a shape that was hostile to the customers I cared about most. I'd built the wrong thing.
What changed in the toolkit
The reason I didn't quit immediately, in 2023, is that the toolkit didn't yet support what I wanted to do. I'd been watching the AI generative work since GPT-3 — interesting but flat. The early image-gen work was fascinating but unusable for production. The early text-gen work could write a paragraph that sounded plausible but couldn't hold a brief together for a multi-page site. The work was happening but it wasn't yet useful for our work.
By the spring of 2025 that had changed. The text-gen models could hold a brief; the image-gen models could produce brand-matched imagery; the orchestration layers could chain enough decisions together that a multi-page coherent site was a tractable problem. The maths I'd been doing in my head — "if a single experienced operator could orchestrate the build, the maths flips" — was no longer hypothetical.
I sat with it for a few months because the maths flipping doesn't mean the work is good. There's a difference between "the platform can produce a working website in twenty-five minutes" and "the working website would meet my agency standards." For most of 2025 I was building privately, testing the platform against the standard I'd held in the agency, and being honest with myself about whether it was meeting the bar.
By autumn 2025 the bar was being met. Not on every site; not on every brief; but reliably enough on the kinds of small business briefs my agency had been turning away for a decade. The platform could produce a site that I, as the senior creative director on the agency, would have signed off on. In twenty-five minutes. For roughly the cost of a coffee.
That was the moment the resignation letter started writing itself.
What I'm trying to build instead
ThemeSmith is the version of agency I always wished I could offer. Same standards. A fraction of the cost. Editorial quality, by software, run by one person who's done this work for two decades and is determined to bring the bar with him.
The shape of it has six commitments, in roughly decreasing order of how much they cost me:
One: the editorial standard holds, regardless of price. A £15/mo Starter site has to meet the same standard a £30,000 agency build would have. The platform compresses the production work; it doesn't compress the standard. If I can't deliver the standard at the price, I drop the customer, not the standard.
Two: I never charge for work the customer doesn't see value from.No discovery fees. No project minimums. No retainer-without-deliverables. The customer pays for the build (one-off if they prefer; subscription if they want maintenance) and that's it. The traditional agency apparatus of phase-billing and retainer-bargaining is gone.
Three: the maintenance is included. The slow rot we wrote about (in the slow rot in Issue 01) is the structural failure of the agency model. The site ships and then dies. We've fixed that by making the maintenance non-optional. Every tier includes the work that keeps the site good, because charging for the maintenance separately was always the pretence that produced the rot.
Four: the customer can leave. The lock-in pattern most subscription services use — cancel and your work goes dark — is a small tax on customer trust that compounds across the relationship. We don't do it. Cancel, get a static snapshot of your site, host it wherever you like. (Wrote about that in why we send a snapshot when you cancel.)
Five: I tell the truth about who we're for. The customer who needs a £30k brand redesign should go elsewhere. The customer who needs enterprise commerce should go elsewhere. I'd rather lose the wrong customer at the brief stage than fail them six months in. (Wrote about that in who shouldn't hire us.)
Six: the work is mine again. The thing I missed about pre-agency life — being responsible for the work, being able to fix it, being the senior eye on every project — is restored. The platform handles the production; I handle the editorial judgement. For the first time in a decade I'm closer to the work than I am to the people doing it.
These six commitments are, structurally, the resignation letter from the agency model. They're what changed when I left.
What I miss
I want to be honest about this. The agency had things I miss.
The team.The 1:1s I described as a trap were also the part of agency life I most enjoyed. I miss the designers I worked with. I miss the developer who could rewrite a CSS architecture in two days while I was in meetings. I miss the project manager who could read a customer brief better than the customer's own marketing director.
The scale. Twenty-five staff is, structurally, a community. There was a Christmas party. There was an office. There were inside jokes. There was a thing.ThemeSmith, currently, is one person at a desk near Brighton. That's right for the shape of business I'm trying to build, but it isn't a community.
The £30k briefs. Once or twice a year, the agency would land a project that was genuinely complex — a startup with a real budget and a real ambition, a charity with a national campaign, a national arts venue with a deep brand history. That work was fun in a way the small-business work usually wasn't. It stretched the team. It pushed the craft. I miss that texture. ThemeSmith is built for working small businesses, and the texture is different — quieter, more patient, more about the slow accumulation of small good sites.
I've made my peace with these losses, mostly. The six commitments are worth more to me than the things I left behind. But it would be dishonest to pretend the move was costless.
What the next years look like
The year ahead is the first cohort. A hundred small UK businesses, signing up at the founder offer rates, getting sites built and tended. The cohort will tell me what's working and what isn't, and I'll write about it (the quarterly Notes from the dispatch desk format is for that).
The year after that, if the first cohort goes well, is hiring. I anticipate three positions in late 2026 — senior engineer, brand designer, customer ops lead. Not because I want to scale fast, but because the work calls for them. The Christmas party will return, in some form.
The year after that, I don't know yet. I'm trying not to over-plan. The mistake I made at the agency was scaling past the size that suited the work.I'd rather get to the right size and stop than aim for an impressive size and miss the work.
What I want to be writing in 2030 is something close to this same letter, but updated — the platform got better; the cohort grew; the editorial standard held; the maths still works.If that's still true in four years, ThemeSmith will be the version of agency I always wished I could offer, at scale.
To anyone in the same place
If you're reading this and you're running an agency that's growing into a shape you don't like, or freelancing in a model that's stretching you thin, or sitting in your own version of the 2023 sales call wondering what the right shape of your work is — you're not alone, and the toolkit has changed.
The shape of work that wasn't possible in 2020 is possible now. The customer base you couldn't serve before, you can serve now. The standard you held yourself to before, you can hold at a tenth the price now. The maths flipped.It's not magic and it's not a marketing line; it's a real shift in what one experienced operator can do with the right platform.
I built ThemeSmith for the customers I'd been turning away. You can build the equivalent for whatever customer you've been turning away. The 2026 toolkit will support it.
That's the actual point of this letter. The agency model isn't dead — for the customers it suits, it's still the right answer. But it never suited the working small business, and now it doesn't have to. That's the resignation.That's the new shape of the thing.
The kettle's on. The work is starting.
— Chris Painter, May 2026