QFM105: Machine Intelligence Reading List - March 2026
Source: Photo by Ecliptic Graphic on Unsplash
This month's Machine Intelligence Reading List runs from autonomous research agents to the copyright status of AI-generated art. Andrej Karpathy's autoresearch lets single-GPU agents conduct ML research overnight, while his MicroGPT tutorial and its interactive explainer give anyone a from-scratch understanding of transformer internals. Manuel Schipper shares how he runs 4-8 parallel coding agents with tmux and Markdown specs. Simon Willison catalogues agentic engineering patterns, and StrongDM's AI team shows what it looks like to build serious software without developers ever reading the code.
Frank Lantz asks Why No AI Games?, the Supreme Court lets stand the ruling that AI art can't be copyrighted, Anthropic publishes research on persona vectors for monitoring character traits in language models, and their leaked Mythos model reveals a reported step change in capabilities. Addy Osmani flags comprehension debt -- AI-generated code is growing faster than anyone can read it -- and Sebastian Raschka's LLM Architecture Gallery provides a visual reference for the model zoo.
As always, the Quantum Fax Machine Propellor Hat Key will guide your browsing. Enjoy!
Links
Regards,
M@
[ED: If you'd like to sign up for this content as an email, click here to join the mailing list.]
Originally published on quantumfaxmachine.com and cross-posted on Medium.
hello@matthewsinclair.com | matthewsinclair.com | bsky.app/@matthewsinclair.com | masto.ai/@matthewsinclair | medium.com/@matthewsinclair | xitter/@matthewsinclair
Was this useful?