I like that you’re positioning this as an “AI PM that happens to design” instead of yet another screen generator. Most tools are great at producing artifacts and terrible at preserving the reasoning behind them. If Figr could reliably spit out a tight decision log/constraints list from each session (what we considered, what we rejected, and why), that alone would replace a lot of hand-wavy product docs.
The 200k UX pattern corpus sounds powerful, but that’s also the scary part: pattern bias overpowering the specifics of a product. The more you can show “this suggestion came from your own data” (analytics, funnels, support tickets) and let teams tune how opinionated the pattern-matching is, the easier it is to trust it for things like onboarding and billing flows rather than just happy-path demos.
Figr do give out the decision logs and even saves it for your future sessions as memory. Your patterns, decisions, work preference, and more are stored.
On the 200k+ ux pattern, it is more to guide the ux building process to avoid hallucinations and make sure solved problems aren't hallucinated. For example there are only 8-10 different ways you can design a settings section.
Mainly we refer to your product context and use it to think through and build specs, flows and designs.