← sources

Hot take: with the new usage limits, your 80-message ChatGPT thread is now your most expensive habit

source: reddit (reddit:ChatGPT)published 2026-11-03 14:28ingested 2026-11-04 02:37

retained text

Everyone's annoyed about usage being measured in tokens now, and fair enough. But I think a lot of us are going to burn through it way faster than we need to, because of how we use the chat, not just because of the limits.

As far as I understand how these models work, every reply is generated with the conversation so far in front of it. So message 80 in a thread costs a lot more to answer than message 3, even if you only typed one line.

The habits I think cost the most now:

**The endless thread.** One chat for a whole project, running for days. When it starts drifting, ask it for a 10-line summary of what you decided, open a new chat, paste the summary. Cheaper, and the answers usually get better too.

**Correcting instead of editing.** "No, I meant the other file" adds a whole new turn on top of the wrong answer. Editing your original message replaces it instead.

**Pasting the whole document for one paragraph.** If you need help with section 3, paste section 3.

**Asking for the final version in one shot.** Ask for an outline first. Fixing a 6-line outline is cheap. Regenerating a long draft three times is not.

**Using the heaviest model for everything.** Rewording an email doesn't need the strongest model you have.

None of this makes the metering okay if it's priced badly. But a lot of the "I hit my limit in an hour" posts this week read like thread-length problems to me.

What's your most token-hungry habit? I'll go first: the never-ending project chat.

document fetched 2026-11-04 00:20 · canonical https://reddit.com/r/ChatGPT/comments/1xmuddx/hot_take_with_the_new_usage_limits_your_80message