Our products

Built from our own work with AI

We adopted AI early and put it to work, building agentic systems for ourselves, and that showed us where the opportunity is. The result is two products: our Agentic System Builder, from Q1 2027, and our Agentic Product Discovery, from Q2 2027.

Agentic System Builder

The blueprint we build client systems with

Q1 2027

Our Agentic System Builder turns what we have learned building agentic systems into a blueprint for the whole system. It combines the three worlds: agents built to current best practice, a machinery of checks and records that makes the work visible, and people in the loop. Each one on its own falls short.

We describe your process as stages, with what goes in, what comes out and which decisions are made. The builder works out every check, approval task and record from that, and the checks run as code. A person approves the exact content they reviewed, and if it changes, the approval no longer counts.

It is for organizations with document-heavy processes where decisions have consequences, and it takes on its first client builds in Q1 2027. You receive the system it builds, designed to run in your own environment. You never receive the builder, and your material never enters it.

Talk to us about a build

1Run · A-0142

README.md
method.md
run-record.md
1-intake
supplied
2-eligibility
3-assessment
assessment.md
parts
4-references
5-panel-review
6-recommendation
7-decision-letter

23-assessment / assessment.md · v3

## Findings

F1Need is shown in the applicant's own attendance records.

↳ evidence: supplied/project-plan.md §2

F2Every milestone in the delivery plan has a named lead.

↳ evidence: parts/delivery.md

F3Reach beyond the pilot area is planned, not yet shown.

↳ no evidence line

## Open questions

Q1Confirm the venue is held for the full term.

3Check · gate G3

Failed

assessment.md › F3

Finding F3 has no evidence line.

One action: add an evidence line to F3, or remove the finding.

Checks are code, not AI.

4Sign-off receipt

Stage 3 accepted at gate G3

Signed by: Panel chair · Review panel
When: 06 Mar · 14:12
Content code: K7Q-42M
Bound to: assessment.md · v4

Change the document and this code stops matching.

A workflow built with our Agentic System Builder, in use · fictional grant review · demo data only, no client data

Agentic Product Discovery

AI-native product discovery that lives in your repository

Q2 2027

Our Agentic Product Discovery is a standalone system for AI-native discovery, built on current agent SDKs. Its method comes from more than 200 digital products our founder has defined, tested and overseen. A team of AI agents and a lab of synthetic customers work problem first, from the market and the problem to competitors, trends, a roadmap and go-to-market, and a person makes each key call.

It follows documentation as code (DaaC): your product documents live in your repository, versioned and reviewed like code. Each one reaches the repository only through a pull request your team reviews and merges, and edits your team makes there flow back into the discovery.

So it helps beyond discovery. As the project moves, it helps you keep the documentation current, and it flags any document for review when a finding it rests on changes. It is for product managers, founders and product teams who keep their documentation in GitHub, and it launches in Q2 2027.

Tell me at launch

A panel listing a product's discovery documents with their review status, including one document changed in GitHub. — 1 · documents

A panel of discovery agents and a synthetic research lab with a persona generator, interviewer, survey runner and research analyst. — 1 · documents