Loop

Review bottleneck

You might say: verification bottleneck, PRs nobody reads

Opening the throttle makes more pull requests. They wait in the human review queue, so lead time grows. The team closes the throttle again, or the queue caps throughput by itself, and every agent added makes the queue longer. Speeding up the agents does not help, because the constraint is the reviewer. The way out is to let evidence carry low-risk work. An agent-ready codebase, verification and a merge policy by risk class take that work out of the queue.

What it does
Shows why the factory's speed is set by how fast its output can be trusted, and where to work to raise it.
Ask
Is the review queue growing faster than merges, and which share of it could merge on evidence instead?
Trouble
Pull requests sit unreviewed for days, go stale and need rework, or reviewers start approving what they have not read.

Loop: balancing

Throttle → Pull request → Review queue → Lead time ↺

Pattern

Accelerator with a brake

Sources

  1. Addy Osmani, Agentic Code Review, What the 2026 data actually showsWe poured machine-speed output into a system built for human-speed work. The bottleneck did not disappear; it moved to verification
  2. Gene Kim and Steve Yegge, Vibe Coding: Building Production-Grade Software With GenAI, Chat, Agents, and Beyond, From Line Cook to Head Chef: Orchestrating AI Teams (book excerpt)When coding is no longer the bottleneck, the rest of your organization becomes the bottleneck.
  3. Mohannad Hammadeh et al. (AWS), Scaling agentic coding past the human bottleneck, Scaling agentic coding past the human bottleneckAgentic coding today scales only as far as the person approving every step
  4. Mohannad Hammadeh et al. (AWS), Scaling agentic coding past the human bottleneck, Scaling agentic coding past the human bottleneckWhat caps agentic development is supervision bandwidth, not model quality.
  5. Gene Kim et al., The Phoenix Project, Chapter 7, p. 90 (5th Anniversary Edition excerpt, ISBN 9781942788294)You don’t factor the availability of any of the other nineteen stations in your decision
  6. Matthew Skelton and Manuel Pais, Team Topologies, Chapter 1, p. 12 (publisher's excerpt)Ultimately, the team becomes a delivery bottleneck, as their cognitive capacity has been largely exceeded, leading to delays, quality issues, and often, a decrease in team members’ motivation
  7. Matt Pocock, AI Coding Dictionary, Human reviewAgents raise the volume of code produced, so review becomes the bottleneck.

Use it on your project

teardown-apply: hold "Review bottleneck" (review-bottleneck) from the teardown at https://teardownatlas.com/software-factory/ against my project

Paste it into a coding agent that has the teardown skills.

An AI agent wrote this map with the teardown skill. It can contain errors: check the sources before you rely on a part. About and legal