Prove it before you build.
ProofForge is the workspace between a messy idea and a defensible decision. Map assumptions, run grounded research, design the cheapest test, and pass a gate only you control.
Find the killer assumption
Rank what fails first
Run the cheapest test
Falsify before you build
Own the decision
Continue, pivot, or stop
Core model
The proof map
Four zones turn uncertainty into structure. AI proposes cards — you Confirm what counts as evidence.
Know
Confirmed
- Person
- Problem
- Evidence
Believe
Hypotheses
- Assumptions
- Outcome
Learn
Open
- Gaps
- Tensions
Next
Action
- Experiment
- Decision
- 8Validation phases
- 0AI final decisions
- 100%Human-owned gate
- 1Clearest next test
Why it exists
Most founders fail by building the wrong belief first — without a map, a test, or a decision they can defend.
Outcomes
Proof you can show
Clarity fast
Named user, problem, outcome, and ranked risks after Plant.
Test this week
Cheapest behavioral experiment — not a roadmap you cannot afford.
Decision receipt
What you learned, what remains uncertain, what you chose.
Platform
Idea to proof, end to end
Phase gates, deterministic risk scoring, pilot tracking, and exportable artifacts — not a chat thread with a logo.
Proof Map
Person, problem, outcome, assumptions — every AI card needs your Confirm.
Reality Scan
Cited research with weak-signal labels until behavior confirms.
Risk Engine
Impact × uncertainty × cost-if-wrong. Transparent scoring.
Proof Lab
Interviews, pilots, or pricing tests ranked for your budget.
AI Assistant
GraphRAG on your corpus. Suggests — never decides for you.
Human gates
Decision Gate and approval flows on every project.
Founder Pack
Export map, protocol, observations, and decision receipt.
Memory & Vault
Semantic search across learnings and research.
Page coach
Guidance tied to your live phase and blockers.
Every project exports
- Proof map export
- Experiment protocol
- Reality Scan sources
- Risk matrix
- Pilot observation log
- Decision receipt
- 30 / 60 / 90 plan
- AI guardrail doc
Responsible AI
AI proposes structure. You own the evidence call.
Gemini extracts and researches. TypeScript rules score risk and enforce gates. The Decision Gate never auto-picks Continue, Pivot, or Stop.
- Zod-validated structured outputs
- Weak-signal labels on web research
- Confirm before evidence counts
- Per-idea AI guardrail on overview
- GraphRAG assistant on your project
- Founder-owned decision gate