“We need more scaffolding”
In which Jessitron and Avdi discuss using LLM agents to make improvement to development environments for both LLM and human coders. Also: the vindicat…
Podcast: Machines of Code and Grace
“How do we give it eyeballs?”
Jessitron and Avdi discuss using an agent with Obsidian, Change Intent Records, and closing the verification loop.
“No poop candy!”
Avdi and Jessitron talk about brown M&Ms, ASCII trumpets, and why we need telltales to ensure our agents haven’t missed something important.
“Historically is like six months ago”
In which Jessitron and Avdi discuss the merits of larger and smaller LLMs, and the challenges of code review in codebases with a lot of history.
“I’ve often wished I could fork myself…”
Jessitron and Avdi discuss defining AI “Feature Owners” in order to avoid accidental breakage.