The 'overnight success' narrative destroys useful failure data
Problem
Startup post-mortems are written after fundraising or acquisition. The real failure reasoning disappears.
Hypothesis
Branching decision logs written during the process have 10x more signal than retrospective post-mortems.
Weakness
Survivorship bias in who writes post-mortems at all. Also: most founders won't log while actively struggling.
Next step
Find 3 founders willing to log weekly decision branches for 6 months.
Branch log
keptReal-time decision logging vs retrospective writing
rejectedBuilding a post-mortem database — already exists (e.g., Failory), low signal
why it turnedTalked to 2 founders: both said their mental model changed significantly within weeks of a pivot. Retrospectives miss the pre-pivot logic.
unresolvedPrivacy problem: founders don't want to log publicly while failing.
Impact
→ triggered 1 connection
→ rediscovered 3 times
Connections
Related tunnels
Why cities fail to adopt micro-mobility even when infrastructure exists
enBike lanes exist but usage remains low. The infrastructure argument doesn't explain adoption gaps.
Fermented food knowledge transfer problem between generations
enGrandmothers hold the real fermentation knowledge. It doesn't get transferred because there's no structured way to capture tacit skill.
医療AIの現地化問題：症状の語彙は文化で異なる
ja日本の患者が「だるい」「重い感じ」と表現する症状は、英語の医療訓練データに存在しない。
Regulatory sandboxes create knowledge islands, not transfer
enFinTech sandboxes in Singapore, UK, UAE operate with different rules. The learnings don't cross borders.