AI drives activity, not outcomes. Outcomist helps you figure out what you actually want before taking action.
People frame decisions wrong. They ask AI to execute before they've clarified the real question. Outcomist surfaces what's actually being decided — then helps you act on it.
Three real examples:
"Should I take $6K/month client work or focus on my $2K MRR product?" Outcomist revealed: $60K savings (12-14 months runway), product grew 4X in 6 months, 97% retention rate. The "stable client" was $40/hour when broken down. "The scarcity is in your head, not your bank account." Recommendation: say no, go all in on the product.
"10 customers want Feature X, 2 customers want Feature Y. Which should I build?" Outcomist revealed a product identity crisis — Feature X serves employers pushing RTO, Feature Y serves employee safety. They can't coexist. Recommendation: build Feature Y, clarify who you're actually building for.
"I want to redesign my office to make it more inspiring" Outcomist revealed the real need was a personal sanctuary — games, memorabilia, a retreat space — not work optimization. Two completely different rooms.
/explore [your decision or question]
- Pattern recognition (30 sec) — identifies what's really being decided and reframes it
- Targeted questions (5–10 min) — builds context one question at a time, adapts based on your answers
- Clear recommendation — specific to your situation with concrete next actions
No hedging. No "here are 7 options." A clear stance.
/explore Should I hire a contractor or build this myself?
from outcomist_core import DiscoveryFlow
from outcomist_core.llm import AnthropicProvider
flow = DiscoveryFlow(llm_provider=AnthropicProvider(api_key="your-key"))
session, response = flow.start_session("Should I take this client work or focus on my product?")
print(response.content)See outcomist_core/README.md for full library docs.
- Epic 1: Deep Discovery — ✅ shipped (v4.8). 100% pass rate on 10 regression tests.
- Epic 2: Proposal Review — 🔄 in design. Tangible, high-quality proposals for user validation.
- Epic 3: Build & Track — ⏳ planned. Execution support for validated proposals.
- Progressive disclosure — 30 seconds first, deeper only if needed
- Honest evidence — pattern recognition yes, invented statistics no
- Adaptive questioning — questions evolve based on your answers
- Clear stance — one recommendation, not a list of options
- One question at a time — never interrogate
Chris Park — Senior PM, Microsoft Office of the CTO, AI Incubation group. Engineering degree from Waterloo. 17 years shipping product.