An attic under a steep timber roof, a shaft of daylight falling from a round window at the centre

Teach your team to automate the work nobody wants to do.

Live courses and hands-on consulting for small and medium companies, from a team that builds, buys and tests its own software in public.

See the courses
A small company's team seated around a long table in a workshop while one colleague presents a diagram on a wall screenAI-generated

Courses for small companies

Live classes on Claude, Cowork agents and generative images. Customer replies, reports and follow-ups running on your own work.

Join the waitlist Two professionals going through a printed plan together at a table in a small business officeAI-generated

Consulting

We work inside your company: what exists, what fits, and a plan your team can run without us.

Ask about consulting

What we've built

We teach what we run on our own products. One is out in the world. The rest are named, scoped and under way — and we'll show the interfaces when there are interfaces to show.

CrewFinder

LIVE

Crew and skippers find each other before the start gun.

  • Finds crew for a boat, or a boat for a sailor, in minutes
  • Boat pages carry the specs, the verification and the crew
  • Rolling out Portugal first, then Spain and France
See the build
CrewFinder feed of recent sailing outings
CrewFinder boat page with specs, verification and crew
CrewFinder race page listing participants

Scully

BUILDING

Watches a video and asks the question the room isn't asking: where is the evidence for that?

  • Visual identity set, October 2026
  • Model evaluation underway
  • Six verdicts, each with its sources
  • True
  • Outdated
  • Missing context
  • False
  • Not enough evidence
  • Opinion

Mimir

BUILDING

Show it a process once, on video. It keeps your way of doing it, not the internet's.

  • Ships as a Claude skill
  • Installed from a single video

Telescreen

PLANNED

Orwell's screen watched the citizen. This one is turned around, and points at power.

  • Open sources only
  • Licence under review

Everything starts on our own products.

Everything we know, we learned by running it ourselves, with our own money at risk. We don't pass on advice we haven't tested on ourselves first.

We run AI accounts on social networks, measuring everything that trends.

And an algorithm that creates trends of its own.

We test how to monetize software.

Paywalls, upsells and pricing, tried across dozens of products — including our own — to find out what actually works.

We test AI inside software that already exists.

Adding AI to a product that already works is easy to get wrong. We build the integration, put it in front of real usage, and keep only what holds up, on the products we own.

Two systems, always running

We built them for ourselves, and every product we ship passes through both.

Terrarium

Closed environments populated by agents that behave like people — each with its own personality, memory and blind spots. We seal them in and watch: how reasoning holds up under pressure, what makes one agent persuade another, and how a piece of information, true or false, moves through a population until it is believed.

Threshold

Where the paywall sits. When the upsell arrives. How the price is framed, and what the second screen asks for. We test these across dozens of products and keep the numbers, so the next launch starts from evidence instead of somebody's instinct in a meeting.

Enter the attic

Our posts and notes, the tests we ran, and the ones that didn't go our way.

Enter the attic