YC AI Startup Coach

Problem

Many AI startups stall not because they lack models, but because they lack judgment: unclear whether the problem is real, unclear which stage validation should target, and unclear whether feedback is evidence or self-comfort. YC AI Startup Coach turns vague founder uncertainty into concrete checks, moving a statement like 'I want to build a product' into a testable picture of who has the problem, how often, how severe it is, and why current solutions fail.

How it works

• Input handling: accepts natural language or structured JSON, infers a mode such as idea_validator, technical_builder, or yc_partner_office_hours, and falls back to onboarding when context is ambiguous.
• Dual-stage diagnosis: declares both Anthropic stages such as Idea/MVP/Launch/Scale and Steve Blank steps such as Discovery/Validation/Creation/Building, using the gap between them to locate self-deception.
• Evidence first: treats sign-ups, usage, payment, retention, and referral as real signals, while requiring interviews grounded in past behavior rather than future assumptions.
• Stage checklists: outputs next actions, framework rationale, and time horizons through idea_stage_checklist, mvp_stage_checklist, launch_stage_checklist, and scale_stage_checklist.
• Pivot judgment: applies Ries-style pivot_or_persevere logic and Moore-style chasm_crossing_checklist to decide whether to narrow the segment, reposition, build a whole product, or keep validating.

Limits

It is useful for startup diagnosis, interview outlines, MVP scope, PMF signal checks, and operational audits. It should not replace real user interviews, legal, financial, or payment decisions, and it should not package unvalidated ideas into a fundraising narrative.