Weekend Reading List — 11/04/2026
Have a great weekend.
Links
“I Still Prefer MCP Over Skills” - David Mohl. I’ve been bullish on Skills as a core emerging primitive for agents, but this is a pretty compelling take (from a technical POV) which argues that MCP is still a more well-suited architecture for agents to interface with other services (as opposed to Skills+CLI’s). Good food for thought.
“AI Agents Could Make Free Software Matter Again” - George London. Interesting and good read on the history of open source, the ideological battle for “software freedom” and what’s changing in a world of cheaper and cheaper code. All in all, a very balanced take on the trade offs between convenience and freedom and why openness (counterintuitively) is the best way to build a moat in an AI-first world.
“Why is continuous learning so hard?” - Sean Godecke. Interesting read on the challenge of applying continuous learning to large language models, fine tuning vs RL and why we’re likely going to have to continue down the path of building scaffolding around (or on top of) models to drive improvements in performance.