Why Claude Loses Track in Long Chats

Short answer: Claude can hold a lot in one conversation, but the more it holds, the worse it recalls any one piece. Anthropic calls this context rot [1]. Long chats get summarized to fit, and a summary drops detail. The fix is shorter chats that start from a written record, not one chat that tries to hold everything.

How big is Claude's context window?

On paid plans, the newest models support up to 1 million tokens, and others 500K or 200K [3]. That's several novels' worth of text. A little of it is always held back for Claude's reply [3].

Why a bigger window doesn't fix forgetting

Anthropic's own docs: "As token count grows, accuracy and recall degrade" [1]. More context isn't automatically better; what you put in matters as much as how much fits [1]. A detail from the first hour of a long chat is competing with everything said since.

What compaction does

When a conversation nears the limit, Claude summarizes earlier messages so the chat can keep going [3]. In the Claude app this needs code execution turned on, and your full history is kept [3]. Anthropic says response quality degrades as a conversation grows [2], and compaction is the workaround, not a cure: a summary is shorter than what it replaced, so something is left out. Long chats that trigger it also use more of your usage limit [3].

If you see "Your message will exceed the length limit for this chat," the message itself is too big for what's left. Start a new chat, or attach less [3].

What works instead

  1. One topic per chat. When the subject changes, start fresh.
  2. Write decisions down as you go. Anything you'd be upset to lose belongs in a file, not in chat history.
  3. Start every chat from the same short brief. Keep your company facts in one folder of dated files with an index Claude reads first. A new chat then knows what the last one knew, without carrying its full weight.

Do it yourself, or have it done

Chet sets up that folder (we call it Start) from your own documents, grades your Claude setup, and fixes what breaks. $99 a month, cancel anytime.

Maximize Claude with Chet: https://www.coworkers.global/get-chet?utm_source=blog&utm_medium=internal&utm_campaign=chet&utm_content=why-claude-loses-track-in-long-chats-cta

FAQ

Why does Claude say the prompt is too long? The message or files you sent don't fit in what's left of the conversation's context window. Split the content, or start a new chat [3].

Can I make Claude's context window bigger? No. The window is fixed per model and plan [3].

Does compaction lose information? The full chat history is kept [3], but Claude works from a summary of earlier turns, which is shorter than the original.

Sources

  1. Claude Platform docs, context windows: https://platform.claude.com/docs/en/build-with-claude/context-windows
  2. Claude Platform docs, compaction: https://platform.claude.com/docs/en/build-with-claude/compaction
  3. Claude Help Center, how usage and length limits work: https://support.claude.com/en/articles/11647753-how-do-usage-and-length-limits-work

Regards,

Charles Stack
Founder, Coworkers.Global

Coworkers.Global is not affiliated with Anthropic. Claude is a trademark of Anthropic, PBC.

Coworkers.Global is an AI staffing agency. We place managed agents into organizations that need dedicated expert knowledge work. A managed agent is an AI specialist provisioned for a specific role, trained on your context, supervised by a person, and accountable for its output. The first, Alex, evaluates startup business plans for fundability, informed by human expertise and research, and calibrated against real investor decisions. We are early-stage with paying clients, and we still lead with the quality of our judgment rather than customer logos. The managed AI staffing agency.
Get new posts by email

What we're learning building a startup with managed agents, plus notes on raising. Weekly newsletter, no spam.

Want more out of Claude today? Chet is available now, $99 a month. Every other role is on the waitlist.