How This Ex-Meta L8 Engineer Ships 40 PRs a Day with AI Agents | Kun Chen

By Peter Yang

Categories: Product, Startup

Summary

An ex-Meta L8 engineer ships 40 PRs daily by letting AI agents handle 90% of coding while he focuses on detailed planning and validation. The key: invest heavily in specs and measurable goals upfront, then parallelize 20-30 concurrent agent sessions to eliminate human bottlenecks.

Key Takeaways

  1. Shift time allocation from coding to planning: spend most effort crafting detailed specs and measurable goals rather than writing prompts, which lets agents run autonomously for longer periods without re-prompting.
  2. Parallelize 20-30 concurrent agent sessions across multiple projects simultaneously. This eliminates waiting time and ensures productive context switching while agents work, similar to managing large scope at big tech companies.
  3. Remove yourself from code review loops entirely. Once you stop reviewing every line, you become the bottleneck. Trust agents with first-pass code and only validate at the end when ambiguity requires human judgment.
  4. Create feedback loops where agents validate their own work continuously. Structure workflows so agents code and validate iteratively without waiting for human approval between cycles, enabling 40 PRs per day.
  5. Redesign team workflows for agent-scale productivity. Traditional processes built around manual coding don't scale when agents produce 10x more output. Move humans out of the loop as much as possible.

Related topics

Transcript Excerpt

If you review every single line of code, you become the bottleneck. So I don't reveal this first pass code from the agent. Eventually I got to a point where I find myself never catching anything the agents don't catch. I typically have like at least five different sessions actively running. On average, there's like 20 to 30 agents running. Most of the time, uh, it's like 20 to 40 kind of PRs every day. Our workflows and how our teams work were built at a time when we spend most of our time coding. But when you start to write like 10 times more PRs, we are not ready for that to really scale up how much we can get from the agents. We have to move ourselves out of the loop as much as possible. Hey everyone, today I'm really excited to welcome my friend uh, an L8 engineer from Meta at Microsof…

More from Peter Yang