Faster coding shifts the bottleneck to the outer loop

Canonical version: Faster coding shifts the bottleneck to the outer loop.

Luca Rossi makes a point about AI coding that matches my thinking: most AI coding workflows optimize the INNER loop, the rapid cycle of describing a feature and generating code. But that's a fraction of software delivery. The OUTER loop (deciding what to build, shaping the solution, releasing, learning from production) is where delivery actually lives.

Speed up only the inner loop and the bottleneck doesn't disappear; it moves:

• Reviews — AI generates more pull requests than humans can reasonably inspect
• Product shaping — teams build faster than they can define what good looks like
• Customer signal — you ship the wrong thing, faster

His three principles for fixing the outer loop:

1. Organize around outcomes, not handoffs. Cross-functional teams span more of the workflow, with reviewable artifacts (specs, plans, decision records) instead of baton passes.
2. Mechanize deterministic work. Automate routing, validation, and analysis; reserve AI for interpretation and judgment. The same principle as Use code for control flow, agents for reasoning, applied to team workflow instead of agent architecture.
3. Make the workflow self-improving. End-to-end reflection turns friction into systematic fixes instead of recurring complaints.

The goal is to go from faster coding to faster DELIVERY. As I discussed in AI makes it harder to remain idle, raw throughput went up, and the constraint moved to the humans doing the deciding, reviewing, and learning around it.

References

• LinkedIn post by Luca Rossi: https://www.linkedin.com/posts/lucaronin_these-days-most-ai-coding-workflows-optimize-share-7488141742517694464-IoWW

Related

• Luca Rossi
• Use code for control flow, agents for reasoning
• AI-Assisted Development Workflow
• AI makes it harder to remain idle

About Sébastien

Ready to get to the next level?