Route cheap decisions to small models, save big models for real work
Arindam_1729 · x · 2026-10-15
The author argues most agents don't need a smarter model — they need a cheaper way to decide what happens next. Route trivial decisions (follow up on a lead? reply to this email? read this doc? escalate to a specialist agent?) to a small decision model like Jev, which returns structured verdicts, while Claude handles research, coding and writing.
His wiring recipe: define a fixed answer set; give the model only needed context; set a confidence threshold in code; route uncertain cases to Claude or a human; log decisions and measure accuracy, cost and latency. Don't use AI for decisions a simple rule handles. Per the linked article, Jev launched Sept 15 and within two weeks OpenAI and AWS shipped their own decision models.
Related event: Cut Agent Costs: Cheap Models for Decisions, Big Models for Writing(2 posts)→
More from coding & agent
- a16z partner: the key skill with coding agents is obsessing over the right details — trq212 · 2026-10-15
- trq212: being technical really means going deep on the right details at the right level — trq212 · 2026-10-15
- Magicare AI cuts nursing admissions to under a minute, slices inference costs 33% on Baseten — baseten · 2026-10-15
- Stanford and Together AI's HybridAgents reason in latent space, gaining 10.4% with 2.7-3.7x cost savings — omarsar0 · 2026-10-15
- Codex tip: you can ask it to find a specific past conversation in a new conversation — burkov · 2026-10-15
- Watchdog blocks all 609 prompt injections by checking every agent tool call against the mission — jagga_tech · 2026-10-15