bakebook · cost & profit

When does bakebook make money?

A what-if model for a subscription app with an AI feature that costs real money to run. Drag the inputs; the numbers update live.

Living doc · rebuilt 2026-08-20 · numbers tracked to the real caps and prices

Inputs

bakebook+ members100
Free users1,000
Price / month$4.99
annual plan is $44.99 (~$3.75/mo)
Apple's cut15%
15% under the Small Business Program
butter messages / user / month10
most free users send only a handful a month; the app still caps them at 4/day
Claude cost / message$0.030

Per month

Subscription revenue$499
After Apple's cut$424
Claude cost (all users)$378
Profit$46
Break-even
covered at paid members

How to read it

The one dynamic that runs the whole model.

The free tier is the cost driver. Free users pay nothing but still run butter (capped at 4 chats a day), so every paid member has to cover their own usage and chip in toward the free users. Break-even is the number of paid members where the subscriptions finally cover the whole free tier. The two inputs that move it most are Apple's cut and the cost per chat (which is really the model and the reply length), which is why both are sliders above.

What keeps the AI cost from running away

The reason a free tier doesn't sink the whole thing.
Free cap
4 butter chats a day for free users. Enough to feel the value, not enough to bleed cost.
Quick-mode step-down
Paid members get full-depth Sonnet, then quietly step down to a cheaper model after 20/day.
Premium fair-use
A 40/day ceiling no real baker hits, so worst-case cost is bounded.
Token ceiling
Every reply is capped, so a single chat can't run up a huge bill.
Apple takes 15% here, not 30%, because bakebook is enrolled in the Small Business Program. That single number moves break-even a lot, which is why it's an input above.