SafetyPage

Alignment Forum and LessWrong · Forum post · LessWrong

Research Lessons I Learned at MATS & Redwood

Jo Jiao · 12 Oct 2026

Read at lesswrong.com ↗Read at sourceNo licence statement found

Outline · 82 paragraphs

  1. Better models of the problem speed up empirical iteration¶5
  2. Develop predictive models and update them¶9
  3. Update with uncertainty in mind¶13
  4. Predict senior researchers’ research takes¶20
  5. Use metacognitive strategies to make predicting and updating habitual¶29
  6. Spend more effort on research communication¶34
  7. Writing good Slack update threads¶43
  8. Making good plots¶51
  9. Be calibrated about the process¶58
  10. Good confusion feels different from bad confusion¶60
  11. Sometimes it’s better to burn money than time¶65
  12. Small time commitments become costly in attention¶67
  13. Know which level of ‘being done’ to aim for¶70
  14. Closing Thoughts¶77