Decide before scaffolding
An idea must earn reject, park, experiment, merge, or promote before project creation becomes available.
04 / Local-first idea admission
Decide what an idea deserves before creating another repository.
A local-first desktop product that turns rough ideas into evidence-backed dispositions, then keeps project creation behind an inspectable approval boundary.
Product brief
- Product
- Roughwise
- Status
- In progress
- Target user
- A technical builder who collects more plausible product ideas than they can responsibly validate or maintain.
- My role
- Product framing, Electron architecture, domain modeling, typed IPC, local persistence, replay behavior, trust boundaries, interaction design, and verification.
The problem
A rough idea can become a repository before its assumptions, evidence, overlap, or real next decision are visible. That creates activity without proving the idea deserves a project.
01 / Workflow
The product, end to end.
Follow the product from first input to a reviewable outcome.
- 01Capture a rough idea
- 02Separate evidence from assumptions
- 03Ask one material question
- 04Recommend a disposition
- 05Review the reasoning
- 06Approve or reject promotion
02 / Decisions
Architecture with a reason.
Electron · React · TypeScript · node:sqlite · Typed IPC · Vitest · Playwright
Proposal is not authority
Research and agent output can recommend a disposition, but only the capability gateway may authorize a mutation.
Keep the record local
Revisions, decisions, and event history persist outside a chat so the admission trail can be inspected and replayed.
Failure paths
What can go wrong is part of the product.
- An illegal disposition transition is requested
- A persistence transaction or migration fails
- Private research payloads cross the public boundary
- The renderer attempts direct filesystem or Node access
Verification
What supports the claim.
- Packaged macOS Electron runtime with typed IPC
- Forty-four unit and integration tests
- Packaged and unpackaged runtime smoke tests
- Restart, migration, rollback, and replay coverage
Portfolio boundary
Complete enough to review. Honest enough to trust.
- In-progress local desktop foundation
- The current interface is a walking skeleton, not the complete admission workflow
- Research uses deterministic fake data; no live provider is connected
- Project promotion and execution are not implemented
- Single-user local state only; no public demo