Use Claude Code and Codex for more than Code

They’re great for writing, strategy, and ops too!

Yes, “Code” is literally in the name. But the real feature is the filesystem.

If you’ve been working with AI, you have definitely been hit by a context window curveball at some point.

Context window is the number of tokens the model can hold in memory and use to generate the response to your prompt.

When it’s exceeded, the oldest tokens get dropped and your AI friend starts losing memory mid-conversation.

Newer models have implemented a feature where they compact the old context before purging the window. It’s better than nothing, I guess , but you don’t know nor can control what it saved.

Coding agents leverage a much better solution: files!

You see it in action when using Plan mode: as the agent reads through the codebase, it writes a file with findings and its own recommendations for implementation.

Well, you should leverage that exact feature even if you’re not coding.

Give your preferred coding agent your brief and ask it to write its findings to a file as it goes.

When compaction hits, it will have far less impact on the subsequent performance because the chat history becomes irrelevant.

Side benefit: you can also view what is generated and catch errors or gaps.