What assignments do you give when coding agents can finish last year’s homework with a single prompt? In the second iteration of Stanford’s CS146S course, Mihail Eric brought open-source maintainers directly into the classroom. Students had to get up to speed on an unfamiliar repository, propose modifications, and explain their changes to the very maintainers who might merge their patches.

Latent.Space
Nov 8 · 121 min
Listen to the conversation (121 min)
What assignments do you give when coding agents can finish last year’s homework with a single prompt? In the second cohort of Stanford’s CS146S, Mihail Eric brought open-source maintainers into the classroom. Students had to familiarize themselves with an unfamiliar codebase, propose changes, and defend their work to the maintainers who might merge their patches.
To celebrate reaching 1M subscribers on the AI Engineer YouTube channel, we invited Mihail, who teaches The Modern Software Developer at Stanford and leads Monaco AI. He open-sourced the course syllabus and shared what parts of the curriculum held up against real-world practice within his own engineering team. Monaco built bespoke agent workspaces, bought off-the-shelf code review tools, and gave on-call agents access to Datadog and GitHub. We discussed how he decides which tools his team needs to own versus buy.
Watch the full interview and screencast on YouTube.
Show Notes
-
Stanford CS146S — The 10-week syllabus and assignments.
-
The Emperor Has No Clothes — Mihail building a coding agent from scratch.
-
Demystifying Claude Code — Mihail dissecting real-world agent harnesses.
-
Monaco’s Agent Workspaces — How the team builds isolated workspaces for each agent.
-
How to Kill the Code Review — An essay exploring what happens when code generation outpaces review bandwidth.
-
Basis’s Authority Map — How agents distinguish actual current behavior from legacy architectural decisions.
-
Stripe’s Minions — A deep dive into Stripe’s background coding agents.
-
AI Engineer CODE Summit — The technical summit mentioned in the intro.
Timestamps
0:00 AI Engineer 1M subscriber celebration and CODE Summit
0:51 Mihail Eric and Stanford CS146S
5:26 What junior engineers need to know today
10:02 Open-source maintainers reviewing student code
15:04 10-week syllabus overview
24:04 Inside coding agents: loops, tools, and harnesses
32:18 Context engineering for large codebases
38:15 Core skills for grokking diffs
46:42 The Monaco.com domain dispute and the airplane banner
50:23 What Monaco builds for sales teams
59:21 Inside Monaco’s engineering team
1:01:40 “Buy to learn, build to own”
1:05:24 Making codebases agent-friendly and closing the on-call loop
1:12:37 When code review becomes the bottleneck
1:20:37 Security, dependencies, and permissions
1:30:13 Background agents and Stripe’s Minions
1:34:16 MCP portals and why infrastructure is converging
1:41:22 Building software factories
1:45:24 What work is left for developers?
1:52:16 From solo leverage to team autonomy
Intro clip excerpted from The Thinking Mode, EO Studio.