COST gone through the roof.
It does ot seem what model I am using the token burn has gone through the roof - before for $20 I used to be abel to do days of work and now it lasts maybe 10 prompts - its gone mad - I just was using Hiku and it used $5 in one promot to add some text to a page - something is wrong with the billing - I think you should check and maybe think of crediting some toekn spend ??
Pending
Hi Gavin, Thank you so much for your patience while we dug into this properly, we wanted to give you a real answer rather than a quick guess. We pulled a full week of your usage on HowHot, and here's exactly where the $80.08 went: • Claude Sonnet: 565 calls, $28.03 • GPT-5.6: 332 calls, $30.65 • Claude Opus: 77 calls, $9.87 • iOS simulator preview time: 15 sessions, $10.40 • Grok and Haiku combined: about $1.13 The honest headline is this: your prompt volume this week is roughly 15 to 20 times higher than your usual 2 to 3 prompts a week, you sent around 40. Your spend only rose 3 to 5 times over that, which is actually a good sign, it means a lot of these were small, cheap tweaks rather than big, expensive builds, and the cost per prompt is behaving normally. In other words, this looks like a busy, productive stretch, the kind you'd expect right around a launch, rather than something wrong with billing, pricing, or which model gets used. We know that's still a real cost to you regardless of the reason, and we don't take that lightly. We really appreciate you flagging this and giving us the chance to look closely rather than just assume. If the pace keeps up and you'd like to talk through ways to manage cost during a crunch like this, we're genuinely happy to help think that through with you. Thank you again for building with us, and congratulations on pushing hard toward launch. Warmly, Pranav