# How Nick Baumann actually uses coding agents

Codex at OpenAI, ex-Cline

**If you take one thing:** `"Chief of staff" thread`, spawns every project thread. one persistent Codex thread runs his work. It starts each new project thread and checks in on them.

- **Agent:** Codex (app) ([source, 2026-05-29](https://x.com/nickbaumann_/status/2060480729830789376))
- **Model:** GPT-6 Astra ([source, 2026-09-20](https://x.com/nickbaumann_/status/2101756295577849940))
- **Model:** GPT-5.3-Codex-Spark ([source, 2026-02-12](https://x.com/nickbaumann_/status/2022019219588886739))
- **At once:** 3–6 threads ([source, 2026-02-12](https://x.com/nickbaumann_/status/2022019219588886739))
- **Harness:** Codex app: one teammate thread, subagent threads, thread automations ([source, 2026-04-16](https://x.com/nickbaumann_/status/2044847066728579493))
- **Skills:** cli-creator ([source, 2026-04-10](https://x.com/nickbaumann_/status/2042705384306336083))
- **MCP servers:** slack, linear, sentry ([source, 2026-04-10](https://x.com/nickbaumann_/status/2042705384306336083))

## In their words

> This has fundamentally changed how I use Codex

> The degree to which Astra is making my life easier at work right now is hard to understate

> 2. GPT-5.3-Codex-Spark when I am in the loop, this is the model I want.

> 1. Codex App: I now run 3-6 threads in parallel, usually on xhigh, and just jump in when input is needed.

> I run this as one main teammate thread plus a few long-lived subagent threads.


Archetype (our label): Orchestrator. Newest source: 2026-09-20.
Card: https://www.prehistoriq.com/stacks/nickbaumann98/ · Data: https://github.com/prehistoriq/stacks/blob/main/data/stacks/nickbaumann98.json