Debate Degrades Reasoning
Single-round debate degrades LLM reasoning in symmetric settings — 2,100 evaluations, 11 conditions, two benchmarks
Structure Beats Scale
How Structured Review Outperforms Brute-Force Generation in LLM Code Synthesis
Long-form thinking and short-form musings.
Debate Degrades Reasoning
Single-round debate degrades LLM reasoning in symmetric settings — 2,100 evaluations, 11 conditions, two benchmarks
Structure Beats Scale
How Structured Review Outperforms Brute-Force Generation in LLM Code Synthesis
A Third Brain That's Less Stupid
July 20, 2026 · 16 min read
The Model Is Faster Than Your Eyeballs
June 18, 2026 · 9 min read
When the Picture Is the Data
May 23, 2026 · 14 min read
The Honesty-Engagement Tradeoff Is a Measurement Failure
April 9, 2026 · 15 min read
The Factuality Ladder
April 1, 2026 · 12 min read
Building ClawClamp: Autonomous AI Agents Without Losing Sleep
March 29, 2026 · 5 min read
Structure Beats Scale
March 11, 2026 · 11 min read
The Only Limit Is Noticing
February 28, 2026 · 9 min read
What I Learned Running 19 Sprints With AI Agents
February 20, 2026 · 12 min read
Today I Stopped Babysitting AI Agent Teams and Started Coaching Them
February 18, 2026 · 11 min read
Stevo's Writing
White papers, blog posts, and various musings.