# How Andreas Kling actually uses coding agents

Ladybird browser, SerenityOS

**If you take one thing:** `Welfare checks`, on workers across 4 machines. the coordinator, not he, watches for stuck or crash-looping workers, so multi-day runs need no steering.

- **Agent:** Claude Code ([source, 2026-02-23](https://ladybird.org/posts/adopting-rust/))
- **Agent:** Codex ([source, 2026-02-23](https://ladybird.org/posts/adopting-rust/))
- **Model:** Frontier model ([source, 2026-09-21](https://x.com/awesomekling/status/2102089363631059199))
- **Review:** Adversarial review passes by different models ([source, 2026-02-23](https://ladybird.org/posts/adopting-rust/))
- **At once:** 4 computers ([source, 2026-09-21](https://x.com/awesomekling/status/2102089363631059199))
- **Harness:** Design doc, then a coordinator agent fanning work out to 4 computers over SSH, with welfare checks ([source, 2026-09-21](https://x.com/awesomekling/status/2102089363631059199))

## In their words

> I used Claude Code and Codex for the translation. This was human-directed, not autonomous code generation.

> I used Claude Code and Codex for the translation.

> 2. Hand it to a coordinator agent (frontier model with maxed-out reasoning)

> 3. Tell coordinator to parallelize the work over 4 separate computers via SSH (because I work on a browser and rebuilds/tests are very CPU+RAM intensive)

> After the initial translation, I ran multiple passes of adversarial review, asking different models to analyze the code for mistakes and bad patterns.


Archetype (our label): Orchestrator. Newest source: 2026-09-21.
Card: https://www.prehistoriq.com/stacks/awesomekling/ · Data: https://github.com/prehistoriq/stacks/blob/main/data/stacks/awesomekling.json