Winsen Labs · An AI laboratory

Reimagine and Rebuild the Future of Work.

Frontier models can already read, write, reconcile and reason at the level of a capable professional. What they cannot do is know your company — who approves what, how your teams write, which client is sensitive, where the exceptions live. We research that gap and build the infrastructure that closes it. Work is better with Winsen.

Start with your problem See what we build

The question

AI arrived for individuals.
It has not arrived for companies.

Your teams have chatbots. They visit them, ask them, copy from them, and then go back to working the way they always have. The licences got bought and the pilots got run, and the underlying process is unchanged. This is not a failure of adoption or of enthusiasm. It is that a model knows everything about the world and nothing about your company: not the last four years of decisions, not the reasons behind them, not the exception your finance team makes for one supplier every March.

That context is real, and it is expensive, and almost none of it is written down in a form a machine can use. Winsen Labs works on turning it into infrastructure — so that AI stops being something your people visit and becomes something your organisation runs on.

What we build

Three pieces of infrastructure. Two of them open source.

Winsen One builds a working model of a company and puts AI workers on top of it. Platos is the runtime those workers live on. Rocketman is how we ship software with coding agents, kept inside the repository where the agents can read it. We run all three ourselves well before anyone else has to.

Product · In development01

Winsen One

A world model of your company.

Winsen One connects to the systems you already run and builds a living model of how your organisation works: your people, your precedents, your language, and the reasons behind decisions nobody wrote down. AI workers operate on top of that model, and every action they take is a proposal until a human approves it. The model belongs to you, and it is exportable.

Winsen One is still maturing. We are working through it with a small number of organisations before opening it more widely.

Winsen One at winsen.ai

Product · Open source02

Platos

The runtime agents actually live on.

Most agent frameworks are built for the demo. Platos is built for the Tuesday six months later, when the agent is load-bearing and something upstream has changed. Model-agnostic, monitored by default, self-hostable in one docker compose up, and yours to run on your own infrastructure.

Visit platos.dev GitHub

Product · Open source03

Rocketman

Project management that lives in the repo.

When coding agents write most of the code, the plan needs to sit where the agents can read it. Rocketman keeps a project hub inside your codebase as a single offline file, with a skill stack that walks an idea from spec to shipped without leaving the repository.

Rocketman on GitHubnpx @winsendotai/rocketman

Working together

We start with your problem, not our product.

We take on a small number of engagements each year, and the first week has no solution in it. We sit with your team and map where the work piles up, because we would rather find out that your real bottleneck is an approval chain than sell you an agent that does not touch it. Then we deploy into one narrow workflow, measure what your people actually approve, and widen only when the numbers make the case. Everything we build with you is yours: the model, the code, the runbook.

01Week 1

Understand

A working session with your team, with nothing to sell in it. We map where work actually piles up, which approvals become bottlenecks, and which of your processes depend on knowledge that only lives in someone's head. Most of what we need is already in the room.

02Weeks 2 to 4

Deploy narrow

One team, one workflow, on your stack and inside your approval rules. Narrow is deliberate: it is the only honest way to find out whether the thing works before it is load-bearing anywhere that matters.

03Weeks 4 to 12

Measure

Approved-work volume, time returned to your people, error rates. You watch the same dashboard we do, which means we cannot tell you a story about the numbers that the numbers do not support.

04When justified

Expand

We widen to the bottleneck we found together, and only once the measurements make the case for it. If they do not make the case, we say so. That is a real outcome of this process, not a failure of it.

Bring us the workflow that hurts How engagements run →

Winsen Academy · Coming up

We also teach this.

Winsen Academy is our school for working this way — not prompt tips, but how these systems behave and how to build with them when the tools change again next month. Taught by the people doing the research and sitting inside real companies. The first courses are being written now.

Join the Academy list →