Hola, I’m Hieu, a Software Engineer

Every team shipping an AI feature knows their monthly API bill. Almost none of them know what a single request costs. Those are different numbers, and only the second one is actionable. The bill is a total. It cannot tell you whether the total is large because each request is expensive or because there are far more requests than you planned for. Those two problems have opposite fixes — one is an architecture problem, the other is a packaging problem — so a team that only has the total tends to respond the same way regardless: someone spends a week trimming the system prompt. As you’ll see below, that is usually the smallest lever available, and it’s the first one everyone reaches for. The bill is also a single scalar for an account that might serve six features. If summarization is cheap and the new agentic workflow is expensive, the invoice averages them into one meaningless number that goes up. What follows builds the missing number from scratch: what one user action costs, end to...

Most people try to cut their AI bill by picking a cheaper model. That’s the smallest of the levers available, and usually the one that costs the most in quality. The bill is set by three things, in this order: What you pay for that’s actually free. Several providers give away real capacity with no card. What you pay for twice. The same 20,000-token system prompt, resent on every request, at full price. What you pay frontier rates for that doesn’t need them. Classification, extraction, renaming, formatting. Fix them in that order. The first is free money, the second is usually a 5–10x reduction on the same model, and only then does model choice matter. Here’s what each one is actually worth right now, with numbers from the providers’ own docs rather than from memory. Tier 0: what’s genuinely free today Free tiers are real, and they’re bigger than most people assume. But the numbers that matter aren’t the daily totals everyone quotes — they’re the per-minute ones, because th...