The bill is 20x the estimate a week after launch

• Projected LLM spend was ~$400/month; the first week alone cost $2,100.
• Traffic is only slightly above the estimate — request volume does not explain the gap.
• Average tokens per request are far higher than in the demo.

demo avg: 3,200 prompt tokens/request, 1 model call
prod avg: 28,500 prompt tokens/request, 4.1 model calls/request
top cost driver: full document corpus (all 40 chunks) injected every request, no cap
model: using the largest model for every call, including classification