Mihail Eric on Why Model Weights Are Eating System Prompts
Stanford's Mihail Eric inspected Claude Code with a proxy. Here is why system prompts are shrinking as frontier models absorb tool instructions.
The best ideas from 30+ top podcasts, every week.
Stanford adjunct professor and Monaco Head of AI Mihail Eric walks through his 10-week curriculum for CS146S, 'The Modern Software Developer,' explaining how traditional computer science education must evolve as AI coding agents commoditize junior engineering tasks. The discussion explores practical system design patterns for AI-native teams, including agent harness architectures, context engineering workflows, and solving the code review merge bottleneck. Eric also details how Monaco builds internal background agents like Monicoder and on-call bots while strategically deciding when to build custom infrastructure versus buying specialized vendor tools.
Stanford's Mihail Eric inspected Claude Code with a proxy. Here is why system prompts are shrinking as frontier models absorb tool instructions.
Mihail Eric explains Monaco's rule for AI infrastructure: buy vendor tools to learn fast, then build internally when control matters.
The week's best AI, tech, business, science, and culture podcasts in one short read. Free, every Sunday.
You get both newsletters, The Throughline and The Carry. No spam. Unsubscribe with one click.