How Yoon found a winning EdTech idea after eight pivots

Executive overview

Pivoting eight times in a year, Yoon kept missing not market fit but founder-market fit — the personal conviction to pursue an idea for a decade. A thyroid cancer diagnosis at 18 gave him a decision filter: "If I only had 10 years left, would I still do this?" That question cut through credential-chasing and sidetrack ideas instantly.

The Ikigai framework — the intersection of what you love, what you're good at, what the world needs, and what pays — pointed him to education. Domain expertise from prior ML work in EdTech, a clear pain point (professors grading 1,000 submissions mechanically), and fast in-person validation got Pensieve to a $6.8M seed round.

Bold bets taken from a 10-year mortality lens compound into surprisingly large returns.

The 10-year decision filter

• Diagnose your motivation before the market: will you still care in 10 years?
• The filter eliminates credentialling detours ("get my PhD first, then start")
• Purpose sustains daily energy through the difficult early stages
• Without founder-market fit, correct YC methodology still fails

Finding the idea with Ikigai

• Four criteria must overlap: love, skill, world need, and payment potential
• Yoon's three candidate sectors: education, energy/climate, healthcare
• Prior ML work in EdTech gave him domain knowledge and operator experience
• Watching a startup grow 20x taught him how educational businesses actually scale
• Education cleared all four Ikigai criteria; it became the non-negotiable domain

Validating AI grading

• Pain point: instructors grading 1,000+ submissions become mechanical, accuracy drops
• First test: a Figma mockup (not a deployed product) shown to a Columbia professor
• Professor dismissed the AI tutor demo and immediately asked to use the AI grader
• Each successive instructor conversation strengthened the signal
• Pensieve reduces grading time by 60–90% without accuracy loss

Getting the first ten colleges

• Sent cold emails to faculty offering a campus visit if slots were booked
• Only flew in when the calendar filled; skipped campuses with no bookings
• Met instructors 1-on-1 to build relationships and convert them to users
• Resourcefulness: mapping available resources and stretching toward out-of-reach ones

Roadmap and larger mission

• 2026 goal: licensed across hundreds of US universities
• Next: direct-to-student AI-native learning platform
• Thesis: future AI-native schools shift to socialisation; knowledge acquisition moves to intimate 1-on-1 AI tutoring at home
• Core human skill in the AI era: making good decisions under uncertainty with incomplete information
• Every conscious decision in uncertainty accumulates into a unique path no AI can replicate