Briefd

Context

Founders rarely write usable design briefs — not through carelessness, but because a brief demands vocabulary they haven't needed yet. The result is a familiar pattern: a project starts on a vague sentence, scope creeps, and both sides are frustrated by week three. I wanted to test whether a language model could close that gap.

Contribution

I designed the five screens in Figma, wrote the prompt layer, built the frontend in React, wired up authentication and the database, and shipped it to production.

Constraints

• One day from start to live, working alone
• Cost per brief had to stay near zero so the tool could be free to use
• No developer, so every technical decision had to be one I could execute myself

Key decisions

• Used a single open text field instead of a structured form, because a form asks founders to already know the answers — which is the exact problem the tool exists to solve.
• Tied rate limiting to user accounts rather than IP address, so shared office wifi and VPNs don't lock people out and the limit can't be trivially bypassed.
• Scored every brief out of 100, which turns a one-shot generator into something people iterate with — a 62 sends them back to add the detail they skipped.
• Chose React over WordPress, because the product needed authentication, a database and API calls rather than pages.

Outcomes

• Live and free at mybriefd.com with real Claude generation, user accounts and per-user rate limiting
• Five screens designed and built in a single working day
• The one-prompt input means brief quality depends on how much the user writes — a single structured follow-up question would likely fix that without losing the simplicity