The founding note. What we are building, why now, and what we are betting on.
What we are
cybercyber is a one-person AI product lab. We design and ship AI products, AI features, and the agent infrastructure that runs them — both for ourselves and for a small number of clients each year. The site went live this week. This is the founding note: what we're building, why we're building it now, and what we're betting on.
The first thing worth saying: cybercyber is not an agency. It's not a studio. The distinction matters, and it isn't decorative. Agencies and studios do good work — some of the best work being shipped in 2026 comes out of small studios with strong points of view, and the world would be poorer without them. But the lab model is structurally different from both, and trying to be a studio while operating as something else is how brands lose coherence in their first six months.
So we're naming it directly, on day one. cybercyber is a lab. Here's what that means, and here's why we're building it.
The personal note
I've been doing product design and UX for more than a decade. Across that time I've worked at agencies, consultancies, FTSE 100 companies, an NYSE-listed business, several startups, a one-person operation, and as employee number eight at a company building from zero. I've been the most senior designer in the room and the most junior. I've shipped work in environments where every decision goes through six approvals, and in environments where the decision is whatever I decide it is at 11pm. I learned thing the hard way, sometimes got lucky, and sometimes my luck ran out. I shipped things I'm proud of. I shipped things I'm not. I worked with people I respect, doing work I'd defend.
But somewhere across all those environments it became clear that the shape of work I do best doesn't fit the shape of any of them, exactly. Agencies are calibrated around throughput. Consultancies around frameworks and deliverables. Large companies around political process and quarterly cycles. Startups around the founder's velocity and opinion. Solo work around the limits of one person's bandwidth. Each of these models works, and excellent work comes out of every one. None of them quite matched how I actually operate at my best.
The thing I'd noticed about my own best work — across product design, UX, brand systems, AI features, businesses I built, products that shipped, products that didn't — was that it required two opposite disciplines, applied at the right moments. Sometimes the work needed time. The decision in the middle of a build that's almost right but not quite. The willingness to delete a week's output and start again because the version on screen wasn't going to last. The patience to sit with a problem until the right answer arrived rather than the convenient one.
And sometimes the work needed the opposite. Ship it now. The version that's good enough. The forty percent solution that gets into users' hands and learns from contact rather than waiting for the perfect ninety percent. The discipline of moving when moving is what's actually called for.
The hard part is knowing which is which. Most environments default to one or the other — either they over-think and never ship, or they ship constantly and never go deep. Sometimes the UX discipline is overly academic, sometimes UI work too ignorant of UX. The work I'm proudest of came from moments where I had the authority to make that call, build by build, decision by decision. Most environments don't structure the work that way. cybercyber does.
What changed in 2026
The reason a one-person lab is even possible right now is that the production bottleneck collapsed.
Five years ago, getting from idea to working product required a team. A designer, two or three engineers, infrastructure people, someone to wire up the auth and the payments and the deployment. Every product worth shipping was a multi-person endeavour, which meant that every decision worth making had to survive being filtered through multiple opinions, calendars, and political constraints. The good ideas that made it through were the ones that survived the gauntlet, not necessarily the ones that were best.
That's gone. AI tools collapsed the production layer to a weekend. A single designer who can prompt well, write code, and operate the modern AI stack can now ship work that used to require four people. Not faster — at the same speed, with the same craft — but with a third the headcount and none of the political overhead.
This isn't a marginal improvement. It's a structural change in what's scarce. The bottleneck used to be production. Now the bottleneck is judgement: knowing what to build, how to build it well, when to take time, and when to ship. Production is free. Judgement isn't.
cybercyber is built around that observation.
Why a lab specifically
Agencies sell hours. Studios sell aesthetic point-of-view. Labs sell judgement and output.
The economic model of a lab is engagement-priced — fixed scope, fixed price, fixed timeline, with the work itself as the unit. Labs don't scale by hiring more designers. They scale by getting better at the decisions that compound. The incentive structure rewards refusing work that wouldn't be worth keeping, refusing scope creep that would compromise the bar, and refusing the convenient answer when the right answer needs more time.
The reason cybercyber operates on the lab model is that it's the only model that aligns with what we're actually selling. The premium pricing isn't there to extract margin from hours. It's there to fund the discipline of refusal. And the lab framing makes the temporal decision explicit — sometimes the right move is to take time, sometimes the right move is to ship, and the lab's job is to know which is which on any given build.
Why we build our own products too.
cybercyber is operator-led. We build our own AI products and businesses alongside client work. This isn't a side project — it's structural to what the lab is.
There are several products in active development inside the lab right now. Some will ship. Most won't, because the discipline we apply to client work is the same discipline we apply to our own: most ideas don't survive contact with reality, and we kill them quickly when they don't. The ones that do ship will be standalone businesses, not portfolio pieces.
The reason this matters for client work is that the judgement we sell stays fresh. We're not consultants advising on AI products from the outside. We're the people building them, in our own stack, on our own dime, every week. Patterns we learn from our own builds inform the work we do for clients. Things we discover in client work inform our own products. The two compound.
This also means cybercyber will only ever take on a small number of client engagements per year. Not because we're being precious about it, but because the math of operating both sides of the lab requires the focus. We'd rather do a few engagements properly than spread thinly across many.
What we are betting on
The bet is simple: in 2026, the studios and agencies that hold the line on judgement will keep doing excellent work — they always have, they always will, and the best of them are calibrated for exactly that. The labs that emerge alongside them will do something slightly different: operator-led, narrow, willing to take time when it matters and ship fast when it matters, structured around the kinds of decisions that compound.
There's room for all three. cybercyber is just clear about which one it is.
The internet is drowning in B+ AI output. We build the version worth keeping. Sometimes that means taking the time. Sometimes that means shipping on Tuesday. Knowing the difference is the whole job.
That's the bet. The lab is open.
---
cybercyber is a lab for AI products, AI features, and the agent infrastructure that runs them. We work with a small number of clients each year. The ones who can tell the difference. Email hello@cybercyber.ai.
