Launching Framework — 100+ thinking frameworks + AI that actually fills them in
Hey IH 👋
Today I'm launching the web version of Framework — a curated library of 100+ thinking frameworks (SWOT, OKRs, RICE, Eisenhower, Premortem, MoSCoW, Lean Canvas, BCG Matrix, ICE, North Star, and 90 more) with an AI co-pilot that fills the canvas in for you and exports a one-page strategy brief.
The short version: pick a framework, type one sentence about your situation, and the AI drafts the whole canvas in ~4 seconds. You then refine in plain English — "push harder on distribution risks" — and the AI rewrites just that slice. At the end, you hit "Generate brief" and get a typeset one-page PDF: Context · Top quick wins · One big bet · What to cut.
Why I built it
A year and a half ago I was sitting in a coffee shop in Shanghai trying to do a SWOT for a product I was deciding whether to ship. I opened a template. Stared at four empty boxes. Reordered the boxes. Closed the laptop. Walked home and wrote a one-pager in plain prose that skipped the analysis entirely.
This wasn't laziness. It was the same thing I'd watched every PM I'd worked with do, every founder I'd advised do, every consultant I'd partnered with do. The frameworks are genuinely excellent. But the blank version of a framework is hostile — you don't know where to start, you don't know if your first answer is good, you don't have momentum. So you skip the framework and write the plan, and the plan is worse than it would have been.
The unlock isn't AI replacing your thinking. It's AI removing the activation cost of starting the thinking. A SWOT is four boxes — the AI doesn't need to be creative, it needs to give you a competent first draft so you can argue with it. That's the thing AI is now genuinely good at, and it's the thing nothing else does.
I shipped the iOS version first (`Framework`) and people kept asking for the laptop version. So I rebuilt it on the web — same Supabase backend, so existing iOS users get the web app on day-0 for free.
What's actually different from "just use ChatGPT"
I tried that for a year. The problem with chat is it returns prose, and prose isn't a SWOT. You can't rank a paragraph. You can't drag-and-drop a sentence. You can't export a one-page brief from a chat transcript.
So Framework gives the AI a shape to think inside — a real 2×2, a 4-quadrant SWOT, a scored RICE table. The AI fills the shape. You don't have to learn how to prompt it. And every AI-drafted item is visually marked (soft coral background) so you always know which thoughts are yours and which are the AI's — that turns out to matter a lot for strategy work.
Tech stack (for the people who'll ask in the comments)
– Next.js 15 with static export (no Node runtime in production)
– Supabase (Postgres + Auth + Edge Functions)
– Stripe in live mode for Premium / Ultra
– Anthropic Claude — Sonnet for Premium, Opus for Ultra — accessed via a single edge-function router that handles streaming + structured output
– Vercel for the static host, Cloudflare in front
– React-PDF for the typeset brief export
– TypeScript end to end
The AI side took the longest. Getting Claude to reliably emit framework-shaped JSON (not "here's a SWOT in markdown") while streaming partial results to the browser was 80% of the AI work. I ended up with a registry of structured-output schemas — one per framework — and a tiny task-name protocol so the same edge function powers canvas drafts, refines, and the brief wizard.
Pricing
Free forever to browse all 100+ frameworks and try the AI with a small credit allowance. Premium is $14.99/mo (or $143.99/yr) for unlimited briefs, 50 AI explorations/month, and PDF export. Ultra is $29.99/mo (or $289.99/yr) for Claude Opus + 4× the credits. No trial card grab. Cancel in one click.
What I'd love feedback on
1. Which frameworks you wish were in the library (I'm planning to ship 20 more in the next month).
2. Whether the AI feels too eager (drafting items you didn't ask for) or not eager enough.
3. The brief format itself — is "Context / Top quick wins / One big bet / What to cut" the right four sections, or is there a better skeleton?
Replying to every comment today and tomorrow. Thanks for reading 🙏
