(aka slammy barson) Data scientist - Econometrics and Computer Science Impulsive builder of things Love for San Francisco - Host of the Bay's Local Podcast
Repositories
saml212/matrix-codi-rank-blindness
Code for 'The Gradient Does Not See Rank: Rank-Indifference in Matrix-CODI on ProsQA' (ICML 2026 MI Workshop; arXiv:2609.03090)
saml212/cascade
Podcast automation pipeline — extract the essence from raw podcast content into concentrated, shareable clips. Deepgram + Claude + FFmpeg.
saml212/thelocalpod-links
saml212/matrix-memories
Matrix-valued fast-weight state as a representational medium for language models: measurable rank, super-linear capacity, exact in-context composition, and the measurement methodology behind the results (negative results included).
saml212/harbor
Framework for evaluating and improving agents
saml212/good-humored
RL environments for humor. All rights reserved.
saml212/awesome-mcp-servers
A collection of MCP servers.
saml212/harness-builder
Scaffold a Claude Code (or Codex) harness that lets one LLM agent work unsupervised for hours and refuse to ship slop.
saml212/stone-launcher
An Android launcher for people who'd rather read. Twelve focused apps, no feed, no autoplay, with an opt-in blocker for the apps that pull you into doomscrolling.
saml212/caddy
building CAD with AI
saml212/parameter-golf
Train the smallest LM you can that fits in 16MB. Best model wins!
saml212/looper
Research framework for testing how AI agents learn from experience
saml212/fft
fft in mojo