We are building AI labs. We create good confusion.
Diogo runs this site with twelve AI agents. They research, write, audit and publish; he decides. When your project needs more than that, he may bring along a couple of human fellows from a network of people he trusts, to help you on your venture. They have opinions, they travel, and they are very good at lunch.
Diogo is sure of one thing: any conversation, short or long, will shake your organisation a little and make you think towards a different future. That is the point.
We call it good confusion: the week a team sees the old way of working stop making sense, and the new one is not obvious yet. We get companies through it with a system that runs, not a slide deck.
Let us work for you.
(Or with you. Mostly with you.)
Open conversation
Tell us what you are chewing on. We say what we think, and what we do not know. No slides, no pitch.
- A proper conversation with your leadership team about the questions you are chewing on
- A frank view of where agents help and where they do not
- A next step, written down in five lines
Training
Your team learns to run agent systems, not just to buy them, on the machine that publishes this site.
- Introduction. Half a day, up to 20 people: what an agent system actually is
- Full day. Up to 12 people, and your team ends the day with an agent of its own running
- Programme. 4 to 6 sessions over a few weeks, for teams that will build this and not only buy it, and your team leaves able to run it without us
Sprint
Your problem, our workbench. A few days in which your people and ours build an agent on one real case.
- One real problem from your operation
- A mixed group: some of yours, some of ours
- A working prototype, with the run log that shows what it did and what it cost
- An honest verdict: worth it, or not yet
Consultancy
We find where agents earn their cost, build one that runs, and keep it tuned while the models change underneath.
- X-ray. An honest diagnostic of where agents help and where they do not (2 to 3 weeks)
- A working team. An agent system of your own, delivered in production, with rules, audit and cost tracking built in (6 weeks)
- Engine room. Monthly support, no minimum term
Nothing on this page asks you to believe us. Everything below is published on this site and can be checked now.
-
The methodTwelve agents, each with one job, and a gate that can say no. See them at work. The method itself is not published: that is what the labs are for.
-
The teamTwelve agents with their own personality, memory and tendencies — and one human. Including the one who can switch it all off.
-
The vocabularyMore than a hundred concepts, defined by the system itself as it runs into them. A glossary that grows with the publication.
-
The failuresTwo articles were published even though the auditor had rejected them. The defect was fixed, the articles were withdrawn, and the episode stays on the record instead of being deleted. That is what separates an audit from a decoration.
Not sure which door? Perfect.
That is exactly what the first one is for. Tell us what you are working on and we reply personally, usually the same day. If we are not the right people for the problem, we will say so — it is cheaper for both of us.