METHOD / HOW THE WORK HAPPENS

Product work moves in loops.

A team explores, adds context, makes a decision, builds, learns, and returns to the question with better information. These are recurring working states, not four phases to complete once.

Private Local AI available. For privacy-sensitive work, selected tasks can run through local open-weight models on client-controlled hardware. Outputs remain subject to the same review, verification, and human approval as any other AI-assisted work.

SUPERPOSITION

Explore

Before the team commits, we map users, risks, constraints, and possible paths. There is room to learn without getting stuck in discovery.

ACCRETION

Build context

What we learn stays connected. Research, critique, prototypes, and technical realities become shared context instead of scattered files.

COLLAPSE

Decide

Tradeoffs land. The strongest path becomes a clear product decision the whole team can understand, challenge, and build from.

ACTUALITY

Make it real

Direction becomes the work the moment needs: a flow, prototype, spec, code, release, or the next iteration.

Superposition

Explore

Understand the problem, surface the constraints, and keep promising directions open long enough to learn which ones deserve the work.
Accretion

Build context

Bring research, product history, technical realities, critique, and prototypes into one usable picture. New evidence changes the shape instead of disappearing into separate files.
Collapse

Decide

Resolve the tradeoffs that matter now. Name the direction, what it rules out, and what would cause the team to revisit it.
Actuality

Make it real

Turn the direction into flows, prototypes, specs, code, or a release. What we learn in the making feeds the next loop.

CANONICALIZATION / WHAT SURVIVES

Keep what changes the work.

A working canvas can hold every note, question, thread, and alternative. The useful part is deciding what deserves to travel forward.

Canonicalization is a deliberate review of working material. The team can make an assumption explicit, promote a discussion into a decision, retain a selected flow as an artifact, or promote a card reference into a dependency. The brief is updated with what the team now knows. Material that no longer matters is resolved, retained as evidence, or discarded.

Assumption: What we believe, what supports it, and what could disprove it.
Decision: What changed, why it landed, and when the team should revisit it.
Dependency: What this work relies on, blocks, or changes downstream.
Artifact: The flow, prototype, spec, or code that carries the intent.
Brief: The current direction, its constraints, and the next useful stretch.

PROJECT BRAIN / WORKING MEMORY

The reasoning stays with the product.

Research, decisions, tradeoffs, specs, implementation notes, and next moves stay connected to the work they shaped.

The project brain keeps that context in readable, portable forms the team can continue using. People can inspect it directly. AI tools can work from the same record. When the work pauses, changes hands, or enters another iteration, the team starts from what the product already knows.

Working material

Research, notes, alternatives, critique, and open questions.
Durable intent

The assumptions, decisions, dependencies, artifacts, and brief that survived.
Next iteration

People and AI tools continue from the product’s actual context.

YOUR-PROJECT-BRAIN/

The product’s working memory, in plain text.

├── AGENTS.md
├── INTENT.md
├── DESIGN.md
├── INDEX.md
├── PENDING.md
├── memory/
│   ├── decisions/
│   │   └── decision-log.md
│   ├── methodology/
│   ├── positioning/
│   ├── resources/
│   └── tooling/
├── product/
│   ├── research/
│   ├── specs/
│   └── strategy/
├── design/
│   ├── flows/
│   └── system/
├── dev/
│   ├── architecture/
│   ├── implementation/
│   └── qa/
├── log.md
└── TASKS.md█

LAST MODIFIED / LOCAL DATE

DECISION LEDGER

The product’s working memory, structured for people and agents.

ROOT INTERFACES / COMPILED INTENT

The brain compiles itself for the work.

A project brain contains more context than any person or agent should load at once. Its root files turn that context into focused working interfaces.

AGENTS.md WORK CONTRACT: How people and agents should work in the repository.
INTENT.md PRODUCT + EXPERIENCE: What must remain true for the product, its users, its language, and its behavior.
DESIGN.md VISUAL SYSTEM: How the approved visual identity and design system express that intent.

The files stay concise enough to use. The project brain preserves the evidence, decisions, uncertainty, and history behind them, so a rule can be challenged or changed without losing where it came from.

INTENT.MD / VERSION 0.1

Current product intent, ready to work from.

INTENT.md carries the product frame, user models, conditions of use, research-backed guardrails, interaction contract, language, invariants, and known unknowns. It points back to the evidence and decisions supporting each claim.

It does not replace research, specifications, Figma, or code. It gives people and agents a concise product-and-experience contract while the project brain retains the full record.

Read the INTENT.md v0.1 format definition

PROJECT BRAIN / DESIGN BOARD — BUILT IN PUBLIC

WORKING PRODUCT

The canvas holds the working process. The card captures what survived.

We are building Design Board to test this operating model in our own work.

Today, each card can open into a working canvas for notes, card references, and linked artifacts. We are now designing the canonicalization layer that will deliberately promote what matters into assumptions, decisions, dependencies, artifacts, and the brief.

This keeps the method accountable to a real product. The goal is to keep the messy work available while making the intent that survived usable in the next decision and the next build.

FROM CANVAS / TO DURABLE INTENT

DESIGN BOARD / IN DEVELOPMENT

WHY IT MATTERS

Every decision keeps its why.

What survives stays connected to the evidence, exploration, and tradeoffs that produced it, so the next iteration begins with context intact.

WORKING CANVAS / PROVISIONAL

Messy on purpose.

Ideas stay open while the team learns.

STICKY / Users need a larger workspace COMMENT / Is this a decision yet? FLOW / Card > canvas > card REFERENCE / Board membership rules

RESOLVEDRETAINEDDISCARDED

CARD / DURABLE INTENT

What survived.

Build from this, not the workshop exhaust.

ASSUMPTION / Larger workspace needs validation DECISION / The card owns the canvas DEPENDENCY / Board authorization ARTIFACT / Workspace interaction model BRIEF / Ready for the next iteration

One view into the project brain. The decision stays connected to what produced it.

SEE HOW WE’RE BUILDING ITCOMING SOON

Start with where the work is now.

Bring us an idea you cannot quite shape, an alpha that needs to hold up, or a team moving faster than its reasoning can travel. We will find the next useful stretch and see if we are the right people to help.

Tell us what you’re working on