Thank you my man

Bundled credits put COGS inside the plan. Heavy users cost real money every month, light users cover them, and gross margin slides even while AI revenue climbs. Margin is what the multiple gets set against, so strong AI numbers and a soft price can sit together fine. Plus Bending Spoons is underwriting price tolerance, and they'll take the churn to get it. Rob's Evernote numbers back out exactly: 0.52 of the users times 2.5x ARPU is the whole 30%. Does anyone know how Miro priced the credit packs? Whether the overage came off inference cost or off the seat price would tell you a lot.

I build these workflows rather than price them, so this is just one small run. Last week I tracked every job in a 6-job pipeline. The price list was wrong on every endpoint I checked, and always in the same direction. One page said $0.34 a second and charged me $0.374. Another said $0.08 and charged $0.15, nearly double. So if your cost model starts from the vendor's price list, it starts from a wrong number. The fix was simple. Ignore your account balance, it tells you nothing, especially on a shared account. Most of these APIs give you the real cost of the job in the response when you submit it. Save that number next to the job ID. That was the only figure that ever added up. One more thing worth knowing. You're charged when you submit, not when the job finishes. Cancel a job or retry one that timed out, and you've paid for both. No idea on sample size, 6 jobs is far too few. For scale, I expected $7.92 and was billed $8.71.