Private Equity's AI Problem Is Not a Model Problem
Why LLM rollouts at investment firms stall out at search, what the four workflows actually demand, and the products that should exist but don't.
Writing
AI Droplets
Notes, essays, project updates, and paper breakdowns—experiments, ideas, and what I am reading in AI.
Writing
Context Is a Data Structure, Not a String
Why LLM agents degrade before hitting token limits, and how treating context as a graph with tiered compression changes everything.
Build Your Blog with Astro
A simple guide on building and deploying your own content website using Astro and Cloudflare.
Closing the Loop on ML Evals
A practical framework for turning evaluation insights into deployable improvements.
Pulse
Weekly roundups of what shipped in open source and what landed in ML research. Assembled automatically, annotated by me.
GitHub Pulse: Week of March 1, 2026
Agents learned to work in teams, SpacetimeDB collapsed the stack, and WiFi replaced cameras for human sensing.
ML Research Pulse: Week of March 1, 2026
Diffusion language models get their PyTorch moment, CUDA kernel generation outpaces compilers, and VLM efficiency breaks through with 89% token reduction.
ML Research Pulse: Week of February 1, 2026
Training-free improvements, process-level agent rewards, and efficient reasoning dominate this week's ML research. 84 papers analyzed across multimodal models, efficiency optimization, and AI agents.
GitHub Pulse: Week of January 31, 2026
AI agents went mainstream this week. OpenClaw hit 132k stars, MCP is becoming infrastructure, and vectorless RAG is challenging embeddings.