A brake, not a wall.
Every AI answer costs real money, so there has to be a limit. I wanted the limit to feel like slowing down, not like being locked out.
There's a basic problem with an AI product that people use all day: the heaviest users cost far more than the lightest, and nobody should be able to use up a month's worth of budget in one afternoon. Most apps solve that with a wall. You hit the limit and you're told no.
I didn't want a wall. On August 30th we put a different design in place.
How the brake works
Neeos keeps a running tally of what it's spending on your behalf. As that tally approaches the limit, it gives things up in order, starting with the least important. First it turns off niceties like the upgraded voice and automatic naming of your chats. Then it eases back on web searching and how long the AI thinks. Then answers get shorter and it uses fewer tools. Everything still works at every step. It just gets a bit plainer.
The reason I like this: you can lose some polish without losing the ability to ask a question. And at most steps, you won't be able to tell much changed.
Where this came from
Earlier that week I'd replaced a guess about how much a typical person uses Neeos in a day with a measurement from my own account. The guess had been off by roughly a factor of three. Real usage is a better teacher.
The honest part
This is new. It's switched on, but I haven't proven yet that it behaves well for a large number of people, and I'm not going to promise you a limit that never changes. What I can say is that its design goal is to slow down, not to say no.
Written October 21, 2026 from the build log, and dated to when this happened.