Selected work · Product case study

Squirrel Board

A shared ideas backlog for the awkward space between we should remember this and we have decided to do it. AI can help with the first part. It does not decide the second.

View the public demo See the product decisions

The problem

Good ideas were becoming accidental commitments

Trish and her AI collaborators do not have an idea shortage. The harder problem is keeping a promising thought without immediately promoting it into a project, forgetting where it came from, or letting momentum make the decision by default.

A normal task list starts too late. Squirrel Board was designed for the earlier, messier stage between this may matter and we have decided to do it.

The product rule

Capture first.

Triage later.

Do not turn every squirrel into a project.

Capturing an idea preserves it. It does not change its priority, activate work, or authorize anyone to act on it.

The design

Authority is part of the product, not a disclaimer

Capture is not commitment

An idea can be saved with its reason and provenance and still remain inactive. Remembering something does not make it urgent.

Agents propose. People decide.

AI collaborators can capture and retrieve ideas or suggest an update. They cannot quietly turn a suggestion into the authoritative record.

Make the history visible

Proposals and decisions stay in the record. A user can see what changed, who contributed, and who made the final call.

Do not build ahead of the evidence

A feature earns its way onto the roadmap by solving a problem that has actually shown up. “That sounds useful” is not enough.

My role

The rules came from using the thing

I defined the governing rules, tested the workflow through daily collaboration, and kept product and governance decisions human-owned. AI collaborators helped explore, implement, test, and document the system under explicit review.

That split was both the way we worked and the thing we were trying to design. How do you make an AI collaborator genuinely useful without giving fluency the weight of authority?

What it demonstrates

What the project proves in practice

Defined a product around a problem we were actually experiencing
Turned authority boundaries into visible product behavior
Made proposals, approvals, and decisions easy to inspect
Protected the private system with authentication and server-side authorization
Kept the scope narrow until actual use justified more
Separated the public demonstration from the private working system

Explore

See the workflow in the public demo

The public version is the original hackathon demo, frozen where it was submitted. The private successor is the one we actually use.

Open Squirrel Board