# Code research projects with async coding agents like Claude Code and Codex

**URL:** <https://awg.osdr.space/t/code-research-projects-with-async-coding-agents-like-claude-code-and-codex/3605>\
**Category:** AI/ML AWG Topics\
**Tags:** coding, agent\
**Created:** [November 8, 2025, 6:27pm UTC](https://awg.osdr.space/t/code-research-projects-with-async-coding-agents-like-claude-code-and-codex/3605 "2025-11-08T18:27:45Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![rtscott2001](https://sea1.discourse-cdn.com/flex001/user_avatar/awg.osdr.space/rtscott2001/32/5_2.png) [@rtscott2001](https://awg.osdr.space/u/rtscott2001)\
**Post date:** [November 8, 2025, 6:27pm UTC](https://awg.osdr.space/t/code-research-projects-with-async-coding-agents-like-claude-code-and-codex/3605/1 "2025-11-08T18:27:45Z")

</div>

> **[Code research projects with async coding agents like Claude Code and Codex](https://simonwillison.net/2025/Nov/6/async-code-research/)**
>
> I’ve been experimenting with a pattern for LLM usage recently that’s working out really well: asynchronous code research tasks. Pick a research question, spin up an asynchronous coding agent and …

Real worth your time to read and consider especially for [@AIMLawg](https://awg.osdr.space/groups/aimlawg) projects

---

<div class="post-metadata">

**Author:** ![angel](https://sea1.discourse-cdn.com/flex001/user_avatar/awg.osdr.space/angel/32/49_2.png) [@angel](https://awg.osdr.space/u/angel)\
**Post date:** [November 8, 2025, 7:09pm UTC](https://awg.osdr.space/t/code-research-projects-with-async-coding-agents-like-claude-code-and-codex/3605/2 "2025-11-08T19:09:11Z")

</div>

These are extraordinary tools, but a single prompt can break the entire project. In fact, from the very beginning, if you ask them for certain types of mathematics (even without requesting a concrete application) they claim they cannot provide it due to national security reasons. If you combine many complex domains in a very dense text with too many premises, all you will get are hallucinations. They are impressive, but one must pay very close attention to what they generate.

---

<div class="post-metadata">

**Author:** ![Jose840602](https://sea1.discourse-cdn.com/flex001/user_avatar/awg.osdr.space/jose840602/32/2657_2.png) [@Jose840602](https://awg.osdr.space/u/Jose840602)\
**Post date:** [November 9, 2025, 1:26am UTC](https://awg.osdr.space/t/code-research-projects-with-async-coding-agents-like-claude-code-and-codex/3605/3 "2025-11-09T01:26:01Z")

</div>

Excellent, it is essential to consider prompt or context engineering. Not to ask too many things, or to combine with N8N for workflow automation.