A developer encountered issues with Claude Code and Codex, where the models would stop generating output mid-task. Standard commands like /compact and /clear did not resolve the problem. The issue appears related to how context windows are billed cumulatively, with each message incurring costs for all preceding messages in the session. AI
IMPACT This issue highlights potential limitations in current LLM context management and billing, which could affect user experience and operational costs.
RANK_REASON User reports a problem with an existing AI product.
Read on Mastodon — mastodon.social →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →