Codex user says his $200 plan now lasts two days
calle bought a $200 Claude plan out of frustration and says Opus 5.5 feels like five times the usage.
calle, who describes himself as a heavy Codex user, says his $200 OpenAI plan now barely lasts two days. He says the same plan used to be incredible value, that he bought a $200 Claude plan out of frustration, and that Opus 5.5 feels like 5x more usage than he gets with Astra at the same price.
then, openai got so greedy that it now barely lasts 2 days anymore.
That is one user's read, not a benchmark. There is no token count, no log, no side by side test in the post. What there is, and what makes it worth reading, is the shape of the complaint. Two subscriptions carry the same $200 sticker and the buyer cannot tell from the price how much work either one will actually absorb before the limits start biting.
Same price, different meter
For a solo founder running an agent most of the working day, the headline number on a coding plan is close to useless on its own. The thing you are buying is a month of runway. calle's claim is that on one plan that runway ran out in 48 hours and on the other it did not, which is a gap big enough that no amount of feature comparison closes it.
He is also blunt about why he switched, and the word he uses is greedy. That is a customer reacting to a change in what he gets for money he was already spending, which is the kind of churn (customers leaving) that does not show up in a feature roadmap.
The mood around it is moving too. levelsio posted that Claude Code is cool again, and added that what is cool and what is not keeps flipping every six months in AI. Two posts is not a migration, and nobody in the sources has published usage data to settle it.
The business read is that if your own build depends on a coding subscription, the number to watch is not the monthly price, it is the date each month when you start rationing.
I love what's cool and what's not keeps flipping every 6 months in AI, it's so fast