Hey PH 👋 I'm Fathy. I build Specky solo out of Vienna. The thing that made me start: building got cheap — Cursor, Claude Code, v0, Bolt — but knowing what's worth building never did. And the work that should answer that question, the interviews and signals and reasoning behind past calls, disappears into docs nobody re-reads. Six months later nobody can say why they built the thing. So Specky keeps a product memory. Your Slack, Jira, GitHub, calls, and research stay connected, the AI works on top of that instead of a blank prompt, and when it drafts a spec every claim links back to the signal it came from. After you ship, it checks whether anyone actually used it. Easiest way to see it: paste any product URL at specky.space — no signup, about 30 seconds, and it builds a memory on that product. I'm not going to ask for upvotes. What I'd genuinely like is your take on the positioning — "product memory layer" is a bet and I'm not certain it's the right frame. Tell me if it doesn't land. I'll be here all day.

Specky - Build faster with a product memory layer

Replies