In trial
AI workshop
Write a request in your team chat. An AI agent implements the change, we review and approve it.
Lab
We try out new tools in our own operations first. What proves itself becomes a building block for our clients.
In trial
Write a request in your team chat. An AI agent implements the change, we review and approve it.
In development
Private AI assistance for your inbox: sorts, summarises and drafts replies, on your own infrastructure.
Early stage
A trust layer for invoices and receipts with DATEV integration: verifiable origin, automatic filing.
Every product keeps its roadmap in its own repository and publishes a public version from it. Status and progress come straight from the work, not from a slide deck.
A product, not a demo.
One learner signs up, studies vocabulary, and has a reason to come back tomorrow.
Lexicon, identity, spaced repetition, review sessions, collections, English and German UI.
Something worth switching to.
The app chooses what you learn next from the graph; words carry context and sound.
Words carry the sentences they live in.
The graph picks what you learn next; words are spoken; the app works without a connection.
The first paying customer.
A teacher runs a class, sets work, sees who needs them, and can author their own sets.
A teacher runs a class, assigns sets and sees who needs them.
Teachers make their own sets instead of teaching somebody else's curriculum.
Revenue that does not need us in the loop.
A teacher can pay for it themselves, and the funnel that reaches them exists without us in the conversation.
Entitlements, Stripe billing, marketing pages and enquiries.
The rest of the classroom job.
Tests are set, sat and marked, with the integrity story that implies.
Tests are set, sat and marked. Backend first, screens next.
Share links, streaks, an operator console, class reports, Latin morphology and Latin audio.
Entries beyond languages, a moderator role, and a review queue that sets and lexicon proposals pass before other learners see them.
Concept cards for subjects like biology and history, the archived subject sets back through the review queue, and AI-drafted sets a moderator approves.
Reasons to bring other people.
Learners study together, compete, rate what they use, ask for what they miss, and can buy sets from other creators.
Groups learners form themselves, leaderboards and duels, a request board, ratings on public sets, and sets for sale.
Contracts rather than seats.
More than one teacher under one roof, with oversight, administration and a parent's view.
Organisations, platform admin, parent portal, messaging, materials, reports.
A repo lanes can work in.
Local environment, database core with RLS, auth, identity, rooms — and the skizze transplant.
Database, auth, identity, rooms, local environment.
The blueprint workshop moves in from the monorepo; 81 e2e specs green.
A cell that books a monastery.
Occupancy plan, booking kernel, approval flow, demo data, public site and 3D tour.
Booking kernel, approval flow, per-room week grid, demo week.
The campus page and the tour, bilingual.
A blueprint every cell can apply.
Catalog as git data, reference tables, roles, cookie sessions, portal sign-in, staging cell.
Versioned package, reference tables, hydration at boot.
blueprint_members, Better-Auth wiring, provider port, new e2e suite.
Sign-in, password reset and the six procedures that had no screens.
Namespace, CNPG cluster, deploy workflow, OTel.
A cell that can be run.
Observability, backups, promotion to production by PR, runbooks.
Promote production by PR, backup and restore rehearsed.
A product rather than a project.
AMD or Fresenius as the second institution; the blueprint proves itself on two applications.
A second institution on the same blueprint.
Real users in the cell.
People migration and credentials — once the data-processing agreement and client sign-off exist.
Subject map, credential issuance, invite path.