Narrowing what we say we do
We took AI off the public services list. Not because we stopped doing the work — because saying we do everything is the easiest way to be hired for nothing.
Until last week, the services page listed four things: custom software, AI / ML, cloud infrastructure, training and support. AI / ML is gone now. The page is back to three categories the studio actually specialises in, plus the support work we sell on retainer.
Two things prompted the cut.
The first is that nobody was hiring us for AI work. They were hiring us for industrial software that contained AI somewhere inside. A content pipeline that uses an LLM under the hood is custom software with a small AI surface. Listing AI as a separate practice was telling buyers we sell something that, in practice, we deliver as part of something else.
The second is more honest. “AI / ML” on a small studio’s services page reads like a trend tag. We have engineers who can ship it when it earns its place inside a system. We don’t have a separate AI practice in any meaningful sense. Pretending otherwise was a positioning lie that cost us nothing today and would cost us a real conversation tomorrow.
The AGB still lists AI solutions as something we can be contracted for. That’s the legal contract surface. The website is what you read before you decide whether we’re worth a call. Those are not the same document.
The four practices that survived the cut: custom software, web development, cloud infrastructure, training and support. Each one has shipped work behind it — BugPin or work running for clients we can’t name. Each one is something we’d rather do too much of than too little.
If you’re building something that needs an LLM somewhere inside it, that’s still us. We just don’t think it’s the front door.
Continue · next note
Self-hosting SeaweedFS for industrial workloads
Notes from running SeaweedFS as the object store behind a closed-source enterprise platform for a year. What works, what surprised us, and what we would do again.