Context Window Problems

Technical discussion of how AI agents lose coherence as context grows, with compaction causing confusion and requiring human redirection

Current AI agents often struggle with "big picture" tasks, requiring users to decompose complex projects into smaller, tightly scoped units to prevent the models from becoming overwhelmed and losing coherence. Even as context windows expand, the process of information compaction frequently leads to a "stupid zone" where agents forget basic instructions or misconfigure themselves, necessitating constant human redirection to stay on track. This persistent friction suggests a need for new architectural standards, as users find that maintaining shorter file sizes and modular code is essential for both machine performance and human readability.

View on HN · Topics

The thing I've learned is that it doesn't do well at the big things (yet).

I have to break large tasks into smaller tasks, and limit the context and scope.

This is the thing that both Superpowers and Ralph [0] do well when they're orchestrating; the plans are broken down enough so that the actual coding agent instance doesn't get overwhelmed and lost.

It'll be interesting to see what Claude Code's new 1m token limit does to this. I'm not sure if the "stupid zone" is due to approaching token limits, or to inherent growth in complexity in the context.

[0] these are the two that I've experimented with, there are others.

View on HN · Topics

ah, so cool. Yeah that is definitely bigger than what I ask for. I'd say the bigger risk I'm dealing with right now is that while it passes all my very strict linting and static analysis toolsets, I neglected to put detailed layered-architecture guidelines in place, so my code files are approaching several hundred lines now. I don't actually know if the "most efficient file size" for an agent is the same as for a human, but I'd like them to be shorter so I can understand them more easily.

View on HN · Topics

I like it because it lets me shoot off a text about making a plot I think about on the bus connecting some random data together. It’s nice having Claude code essentially anywhere. I do think that this is a nice big increment because of that. But also it suffers the large code base problems everyone else complains about. Tbh I think if its context window was ten times bigger this would be less of an issue. Usually compacting seems to be when it starts losing the thread and I have to redirect it.

View on HN · Topics

> There is no code, there are no tools, there is no configuration, and there are no projects.

To add to this, OpenClaw is incapable of doing anything meaningful. The context management is horrible, the bot constantly forgets basic instructions, and often misconfigures itself to the point of crashing.

Summarizer