Ask HN: How do you manage long running AI conversations?

[boh144]

ASK HN: I’m trying to understand how people handle longer running AI conversations that span days or weeks and involve exploring multiple approaches.

Personally, I keep running into issues like:

- losing track of where a useful insight actually came from - meandering threads where good ideas get buried in tangents - having to re-explain context when returning later - branching ideas across multiple chats

Curious how others deal with this today.

Do you keep everything in one thread? Split aggressively? Export to notes/docs? Something else entirely?

If you prefer to share your workflow privately would love to hear about it: [e-mail]

[uberman]

I assume you mean outside the boundaries of context window size or token thresholds right?

[boh144]

Both in a way. I'd say it gets worse outside the context window it seems to get worse due to compression and loss of detail. Even in a single context window it's hard to keep ideas and branches clean and remember important insights.