AI-native product engineering · Knowledge and feedback
The product's memory, kept by design.
Where what you know about your product lives, how it changes, and how what is learned comes back.
01The problem
Products forget. Documents drift. Decisions get lost.
The why disappears
Six months in, nobody remembers why a rule exists. The person who knew left, the document was never updated.
Two sources of truth
The deck says one thing, the code another, the ticket a third. Each change has to be made three times, and is not.
AI makes it worse, unmanaged
An agent without a model guesses. Fast output on a wrong assumption is the most expensive kind.
02One connected model
Knowledge is the source. Everything else is a view of it.
One concept per file
Events, rules, specs, decisions, learnings: each one a small, versioned, linked record. Readable by people and by agents.
Typed links
This spec implements that capability; this decision supersedes that one; this learning impacts that block. Traceability is in the model, not in someone's head.
Generated views
Indexes, diagrams, the history, the dashboard: produced from the model, never edited by hand, never stale.
03How a decision is made
One decision, one proposal, one debate, two approvals.
- 01Proposal · Anyone, including an agent
- 02Debate · In the open, on the text
- 03Alternatives · Recorded, with why they lost
- 04Approval · Two people, separate roles
- 05Decided · Never edited; superseded
A decided record never changes in place. A new record supersedes it through the same flow, so the history is always readable. People who never use Git take part from a browser.
04The learning loop
What is learned in one product improves the next.
- 01Record · A learning, with evidence
- 02Harvest · Routed where it belongs
- 03Decide · Accepted or rejected, with a reason
- 04Return · As a new version
- 05Measure · Time from learning to return
Learnings are written at every retrospective, incident, validation or moving metric. What belongs to your product stays in your product. What improves the method or a building block returns to every product as a version.
05Where it lives
Your product's knowledge, in your product's repository.
| Area | What it holds | Who writes it |
|---|---|---|
| Domain | Events, commands, rules, boundaries, glossary | Discovery sessions, validated by your experts |
| Product | Capabilities, journeys, personas | Your product owner with the VM8 lead |
| Architecture | Systems, contracts, blocks in use, decisions | VM8 engineers, approved with you |
| Specs | Functional spec, acceptance criteria, UI spec, per capability | Agents draft, your product owner approves |
| Learnings | What was learned and where it goes | Everyone, at every retrospective |
Domain
What it holdsEvents, commands, rules, boundaries, glossaryWho writes itDiscovery sessions, validated by your expertsProduct
What it holdsCapabilities, journeys, personasWho writes itYour product owner with the VM8 leadArchitecture
What it holdsSystems, contracts, blocks in use, decisionsWho writes itVM8 engineers, approved with youSpecs
What it holdsFunctional spec, acceptance criteria, UI spec, per capabilityWho writes itAgents draft, your product owner approvesLearnings
What it holdsWhat was learned and where it goesWho writes itEveryone, at every retrospective
Published as a site you read in a browser. VM8's own method and principles are mounted alongside, never copied, so your knowledge says only what exists because your product exists.
06What it is worth
Memory you can hand to anyone.
Onboarding in days
A new engineer, a new partner, an auditor: they read the model, not the code, and understand the product.
Due diligence, ready
Every decision with its reason, every change with its author and date. Nothing to reconstruct.
Agents that know your product
The same model is what the agents read. Precise context, less AI spend, fewer wrong guesses.