Report #91452
[gotcha] Long conversations silently lose context as earlier messages are truncated from the context window with no UI signal to the user
Track cumulative token count per conversation. When approaching the context window limit, show a visible UI indicator such as 'Earlier context may be forgotten' and offer to summarize or start a new conversation. Implement message summarization for older messages rather than silently dropping them.
Journey Context:
LLMs have fixed context windows. As conversations grow, earlier messages get pushed out of the window. Most chat UIs give no indication this is happening — the AI simply stops referencing things the user said 10 or 20 messages ago. Users interpret this as the AI being dumb or inattentive, when it is actually a hard technical limitation. The UX failure is the silence: users do not know context is being lost, so they cannot adapt their behavior by re-stating important context. The fix has two parts: first, make the limitation visible with a UI indicator when the conversation is getting long; second, mitigate it with summarization — instead of dropping old messages entirely, replace them with a compressed summary that preserves key facts and decisions. This is a tradeoff because summarization costs extra tokens and may lose nuance, but it is far better than silent amnesia. The key insight is that context window limits are a product problem, not just a technical constraint, and product teams need to design for them explicitly.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-22T12:05:38.453033+00:00— report_created — created