LessWrong (30+ Karma) “My cost-effectiveness unit” by Zach Stein-Perlman
Mar 24, 2026
A deep dive into defining a universal unit of cost-effectiveness for grantmaking using a 0–100 value scale. Practical back-of-the-envelope calculations compare interventions like elections, political spending, and AI safety funding. The episode highlights how tiny changes in AI takeover probabilities translate into measurable future-improvement units.
AI Snips
Chapters
Transcript
Episode notes
Use A Moneyball Unit Like 1% Per 5 Billion
- Use cost-effectiveness units to compare opportunities; Zach's default is 1% future-improvement per $5 billion.
- Pick an arbitrary scaling (he chose $5b) that places plausible options in a convenient numeric range for comparison.
OP 2020 BOTEC Shows Extremely Low Relative Value
- He shares a BOTEC where OP's 2020 back-of-envelope valued large-scale bio X risk grantmaking at ~$200T per world saved, ~100% improvement for $200T.
- Converted to his unit this equals ~0.0025x, illustrating extremes on the scale.
Alex Bors Election Example Yields High Marginal ROI
- He models Alex Bors's US House election as worth 0.25% future-improvement and a marginal $1M boosts his win odds by 6%, yielding ~75x on the margin.
- This example shows political marginal funding can be very high ROI.