Collaborative AI Engineering: One Dev, Two Dozen Agents, Zero Alignment — Maggie Appleton, GitHub

By ai.engineer

Categories: AI, Tools

Summary

Coding agents have made implementation cheap, but team alignment is now the real bottleneck—current tools funnel agentic output into outdated systems, creating critical feedback loops that arrive after implementation rather than before it.

Key Takeaways

  1. The time between logging an issue and an agent opening a PR has collapsed to minutes, but review happens after implementation when it's too late to course-correct, shifting all alignment burden onto pull requests.
  2. Coding agents operate with unshared local plan modes that teams never review collectively before deployment, eliminating critical alignment checkpoints that used to exist in the planning phase.
  3. Software development process has historically included planning, building, and review phases with multiple Slack/Zoom touchpoints—but agentic development collapses this timeline, removing early alignment conversations entirely.
  4. Current tools (GitHub, Slack, Jira, Linear) weren't designed for agentic development workflows; they funneling masses of agentic outputs into platforms built for outdated software development processes.
  5. When production becomes cheap, opportunity cost becomes the real cost—teams need shared spaces to align on what to build before agents start, not debate implementation details after.

Topics

Transcript Excerpt

Okay. Are we all good? Right. Uh so yes, this talk uh is called uh one developer, two dozen agents, zero alignment. Uh this is the case for why we need collaborative AI engineering. So first a very quick intro. I'm Maggie. I work uh at GitHub as a staff researcher engineer. Uh at least that's my title. I'm actually a designer back when that was like a separate thing to engineer. Um and next is the labs team within GitHub. So we work on kind of more experimental risky bets than the rest of the or...