Tokens, infra, debug, monitoring, false positives. Here is the full bill of a SaaS AI app for a client, including always-forgotten line items.
"How much will the API cost?" Real answer: API is only 40-50% of the total.
LLM tokens ~$20k, embeddings ~$80, vector DB $350, app infra $810, monitoring $630, human time $2k, hidden costs $700. Total ~$24k/month for 2500 MAU.
Prompt caching, smart routing, batch API, per-user caps.
Prompt injection, data leakage, jailbreaks, runaway cost. The pragmatic guide to AI-specific risks and their countermeasures.
LLMs will never fully stop hallucinating. But 80% of hallucinations come from 4 avoidable mistakes. Here they are.
No more "please return valid JSON" with fragile parsing. Here is how to get 99.9% valid JSON in production.
I design and ship tailor-made applications. 9 years of experience. Let's talk over a coffee (or a call).
Start a project