# All Code Is Slop (Except Mine) — Lauren Tan, SpaceXSI

> For agents: start with the [agent guide](https://ratstack.sh/llms.txt). Every page is Markdown by default; add `Accept: text/html` for HTML.

[AI Engineer](https://www.youtube.com/@aiDotEngineer) · Published 2026-11-14 · 17:31 · S tier · Agents and MCP

[Watch on YouTube](https://www.youtube.com/watch?v=O_zcc6Xl12g) · [All RatTube videos](/tube)

## Summary

[Lauren Tan](/lore/lauren-tan) describes how her team keeps code quality up while agents write most of it. She ranks where to spend effort as an inverse pyramid: code and API design first, then lint rules, types and tests, then runtime and formal verification, with prose rules last.

She treats the codebase as a garden that agents copy from, so workarounds and old patterns spread unless they are removed. She reports that her team ships thousands of pull requests a month with a revert rate near half a percent, while the app has become faster.

## Section notes

- [0:00](https://www.youtube.com/watch?v=O_zcc6Xl12g\&t=0s): The talk asks how to ship thousands of pull requests without losing quality.
- [2:00](https://www.youtube.com/watch?v=O_zcc6Xl12g\&t=120s): An inverse pyramid ranks code, checks, verification and prose by where to spend effort.
- [3:00](https://www.youtube.com/watch?v=O_zcc6Xl12g\&t=180s): The codebase is the largest input, so each correction should become a deterministic check.
- [5:00](https://www.youtube.com/watch?v=O_zcc6Xl12g\&t=300s): Her framework bans a React hook agents misuse and blocks imports across directories.
- [7:00](https://www.youtube.com/watch?v=O_zcc6Xl12g\&t=420s): Agents use comments and docs to justify workarounds, so most comments are banned.
- [9:00](https://www.youtube.com/watch?v=O_zcc6Xl12g\&t=540s): Videos, screenshots and a feature map replace most code review.
- [10:40](https://www.youtube.com/watch?v=O_zcc6Xl12g\&t=640s): Formal verification and UI property tests with Bombadil catch more bugs.
- [13:00](https://www.youtube.com/watch?v=O_zcc6Xl12g\&t=780s): Her team's numbers: more pull requests, few reverts and a faster app.
- [15:00](https://www.youtube.com/watch?v=O_zcc6Xl12g\&t=900s): Fix the inner loop before automating it, or automation only makes more slop.

## Key concepts

- [No comments in code](/lore/no-comments)
- [The fence](/lore/the-fence)
- [Property-based testing](/lore/property-based-testing)
- [Lauren Tan](/lore/lauren-tan)

## Our take

Lauren Tan argues that every correction you give an agent should become a lint rule, a type, a test or a better API, not another line of prose. She shows rules that ban patterns agents misuse, a ban on most code comments, and verification that replaces most code review. It is the thinking behind rat-stack's own fence.

## Related videos

- [Stop agent slop with Effect |  Mattia Manzati | Effect Milan 2026 🇮🇹](/tube/stop-agent-slop-with-effect-mattia)
- [Code Mode: Let the Code do the Talking - Sunil Pai, Cloudflare](/tube/code-mode-let-code-do-talking)
- [John Hughes - Testing the Hard Stuff and Staying Sane](/tube/testing-hard-stuff-john-hughes)
- [How We Build Effective Agents: Barry Zhang, Anthropic](/tube/building-effective-agents-barry-zhang)

Summary based on the transcript. Written 2026-11-16.
