Summary

Wes McKinney describes how his three-person team at Kenn Software merges hundreds of pull requests per week while maintaining a low bug rate by practicing what he calls 'agentic engineering' — a human-led, spec-driven process that explicitly rejects fully autonomous AI loops. He frames the core distinction as 'vibe coding is not caring at scale,' arguing that design, architecture, and taste must remain human responsibilities. The workflow centers on adversarial spec review, small-increment implementation with roborev verification, and converting ephemeral agent documents into living architecture references. Kenn has also built a custom toolchain (Forge, Ghosthub, Kata, AgentsView, roborev) because existing tools were too slow and friction-heavy for high-throughput agent-assisted development. A published 'Clanker Constitution' governs how coding agents behave, counteracting their poor default behaviors around communication, sloppiness, and overstepping.

Key Insight

High-throughput agentic engineering is only sustainable when humans own design and taste decisions at every stage — autonomous loops don't produce quality, they just produce volume.

Spicy Quotes (click to share)

• 8

  Vibe coding is not caring at scale.

• 9

  I think fully autonomous, no-human-in-the-loop pipelines are bullshit: anyone who is telling you that you can engineer agents looping on each other's output, step away from the keyboard, and get good quality output on the other end, in almost all cases, is either a) clueless or b) selling you something.

• 7

  The work produced by the latest frontier models (5.6-Sol and Fable) is extremely sloppy and almost never suitable for production without substantial hardening.

• 6

  Design and taste are not delegated to an agent.

• 7

  The clankers (what we call the coding agents, since 'agent' gives them too much credit) are not in charge, we are.

• 6

  Pull request descriptions need to use plain language and lead with outcomes, and not be a wall of text of 'robospeak' which seems to be the default behavior nowadays especially with Claude Fable.

• 5

  We are biased towards human-operator-centric workflow that is intended to minimize the amount of code slop that lands in our repos.

• 5

  Never delegating critical work to an autonomous coding loop.

Tone

opinionated, pragmatic, critical