Cost breakdown: 100 LLM calls in 8.5 seconds for roughly a quarter of a cent
narphorium · x · 2026-09-19
Follow-up on the English-logic interpreter: since Jev doesn't generate text, the author uses Claude Haiku to translate the question into a logic query first. That query then took 100 Jev calls in 8.5 seconds, costing roughly a quarter of a cent.
This fills in the economics of the pipeline — LLM query translation plus many cheap model calls plus proof search in code — showing the externalized-reasoning approach is cost-viable.
More from coding & agent
- LLM-powered tweet classification at scale: near-instant and essentially free — ivan_bezdomny · 2026-09-19
- SELF-INDEX: a framework for retrieval indexes that self-evolve without humans — Sangam Lee · 2026-09-19
- Factory launches Private deployment for VPC, on-prem and air-gapped networks — matanSF · 2026-09-19
- Jev returns choice probabilities at a fraction of LLM cost for game AI — tristanbob · 2026-09-19
- Automated Evals Recover 87% of Human-Flagged Failures, Parlance Labs Study Finds — HamelHusain · 2026-09-19
- Gary Marcus Warns LLM-Generated Code Is Piling Up a Security Nightmare — GaryMarcus · 2026-09-19