domain

research

7 cards · 4 operators · 6 tier-A claims · 3 synthesis patterns.

Strongest claims

1. AI capability is not evenly distributed, it spikes where labs have data, rewards, and verification loops Andrej Karpathy
2. You can outsource thinking, but not understanding, verification is the new human job Andrej Karpathy
3. An LLM should maintain a wiki, not re-derive knowledge per query Andrej Karpathy
4. The middle is hollowing out, execution gets automated, leaving spec-writing and verification as the high-value human tasks Eugene Yan
5. Close the feedback loop by mining session transcripts for patterns to promote into config Eugene Yan
6. A trace alone teaches nothing; learning requires feedback attached to the trace Harrison Chase

Adjacent domains

• ai-native · 6 co-occurrences
• engineering · 6 co-occurrences
• leadership · 3 co-occurrences
• product · 1 co-occurrences

Synthesis patterns in research

• Evals are data analysis, single judge, binary rubrics, error analysis first
• Frontline customer contact is the PMM substrate
• Verification, not execution, is the irreplaceable human job

7 insights in research

• Exploring and understanding complex systems · Anirudh Rayaprolu
• AI capability is not evenly distributed, it spikes where labs have data, rewards, and verification loops · Andrej Karpathy
• You can outsource thinking, but not understanding, verification is the new human job · Andrej Karpathy
• An LLM should maintain a wiki, not re-derive knowledge per query · Andrej Karpathy
• The middle is hollowing out, execution gets automated, leaving spec-writing and verification as the high-value human tasks · Eugene Yan
• Close the feedback loop by mining session transcripts for patterns to promote into config · Eugene Yan
• A trace alone teaches nothing; learning requires feedback attached to the trace · Harrison Chase