← sources

You probably don’t need Astra for half the stuff you’re using Astra for

source: reddit (reddit:ChatGPT)published 2026-09-30 10:34ingested 2026-10-01 02:38

retained text

guys, try Luna before spending your whole limit on Astra!!!

I see so many “Astra ate my weekly usage” posts. And yeah the limits are confusing, but why are we using the strongest model for everything before even knowing what we need?

First just discuss your idea in normal Chat. Ask questions, let it google stuff, figure out what you actually want. Change your mind there, not halfway through an expensive Work session lol. Chat has limits too, depending on what you use, but this doesn’t need to eat your Work/Codex budget.

Then bring the conclusions into Work. What you need, what should NOT change, how to check the result. Give it a proper prompt.

Start with Luna low. Not enough? More effort. Still stuck? Sol. Then Astra if needed. You don’t need to try every setting, but bro at least give the cheaper one a chance before going straight to max everything!!!!

Don’t make Astra do what Luna can do. And don’t make Luna do what you could sort out in normal Chat.

Also use MD files for project context. Seriously just ask the model to make them. Separate topics, small index with links, nothing complicated.

You fixed logging today? Save where the config is, which files matter, what you changed and why. Next time tell it to read those notes. Why pay again for it to explore the same shit? Keep them updated though, otherwise you just have very organised wrong information lol

And long chats usually mean more context to process. Caching helps, but keeping every unrelated task in one endless chat isn’t free. Those MD notes make it easier to start fresh without explaining the whole project again.

There is a BIG diff between finished task and price per token!

document fetched 2026-10-01 00:23 · canonical https://reddit.com/r/ChatGPT/comments/1wu1hko/you_probably_dont_need_astra_for_half_the_stuff