# Stop agent slop with Effect |  Mattia Manzati | Effect Milan 2026 🇮🇹

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

[Effect | TypeScript for the AI Era](https://www.youtube.com/@effect-ts) · Published 2026-06-25 · 33:52 · S tier · Effect

[Watch on YouTube](https://www.youtube.com/watch?v=88XERFAibQc) · [All RatTube videos](/tube)

## Summary

Mattia Manzati examines how to keep coding agents from producing code that ignores an application's rules. [Effect](/lore/effect-basics) makes success values, failures, and dependency requirements visible to the [type checker](/lore/the-fence), which can feed mistakes back to the agent.

He also connects [Effect](/lore/effect-basics) with runtime observation. An experimental side project called inspect collects telemetry through the JavaScript runtime's debug protocol. Agents can use it to debug a local reproduction.

## Section notes

- [0:55](https://www.youtube.com/watch?v=88XERFAibQc\&t=55s): The talk frames agent engineering around preventing low-quality generated code.
- [12:03](https://www.youtube.com/watch?v=88XERFAibQc\&t=723s): Effect's type parameters let the checker expose new failures and requirements.
- [18:16](https://www.youtube.com/watch?v=88XERFAibQc\&t=1096s): Lint and language service rules can steer agents from raw environment reads, logging, fetch, and JSON.parse toward Effect APIs.
- [24:20](https://www.youtube.com/watch?v=88XERFAibQc\&t=1460s): Logs, metrics, traces, and errors provide runtime evidence.
- [30:31](https://www.youtube.com/watch?v=88XERFAibQc\&t=1831s): The experimental inspect tool collects telemetry through the JavaScript runtime's debug protocol for debugging local reproductions.

## Key concepts

- [Effect](/lore/effect-basics)
- [schemas](/lore/schemas-define-the-boundary)
- [type checker](/lore/the-fence)

## Our take

Mattia Manzati shows how to keep agent-written code honest with Effect: the Effect LSP turns a missing requirement into a plain diagnostic where the compiler only reports an assignability error, then lint rules, tests, and watching the code run. It is the same kind of fence this stack runs.

## Related videos

- [Effectifying OpenCode // Kit Langton // Cause & Effect 8](/tube/effectifying-opencode-kit-langton-cause-effect)
- [Next-gen DevTools for Effect | Mattia Manzati (Effect Days 2025)](/tube/next-gen-devtools-for-effect-mattia)
- [Production-Grade TypeScript by Johannes Schickling (Effect Days 2024)](/tube/production-grade-typescript-by-johannes-schickling)
- [Effect TypeScript Library: Beginner / Intermediate Workshop (Effect Days 2024)](/tube/effect-typescript-library-beginner-intermediate-workshop)

Summary based on the transcript. Written 2026-11-15.
