Blog

Machine Learning

Energy-based models, autoregressive models, learning theory, and formal methods for AI.

• Why Energy-Based Models? The Toy AR-vs-EBM Argument

• Score Matching: Training EBMs Without Ever Computing Z

• Autoregressive Models + LLMs Exponential Error-Compounding Argument — Is It Real or Fiction?

• Asking the Right Question: Formal Methods as Scalable Oversight

• Uniform Stability in Machine Learning Theory

Music

Jazz, arrangements, and other musical projects.

• Metallica Goes Jazz: A 2010 High School Arrangement, Reanimated by AI

• Jazz Improvisation

Meta-Research: How to Do Research

Research practice — picking directions, moving fast, writing, and thriving in the community — plus the self-development that supports it (e.g., mindfulness).

• Velocity in Research

• Vectoring: Attack the Biggest Uncertainty First

• Thriving in the Research Community

• How to Write a Statement of Purpose for Graduate School

• Mindfulness as a Cure for Anxiety

• Building Synergetic Teams for Research and Startups

Agentic Coding

Working with AI coding agents: workflows, prompt design, and writing in the AI era.

• The Em Dash Is Only an AI Fingerprint If You Didn't Already Use It ;)

• Embrace AI or Be Left Behind — By People, Not Machines

• Two TLDRs Are Better Than One: A Small Prompt-Design Fix That Matters More Than It Looks

• What I Learned Building a Correctness-Gated Multi-Agent Workflow for Research