Classifier, not LLM: how one user sorted 80k emails with JEV plus an LLM
LinusEkenstam · x · 2026-09-23
Responding to Ole Lehmann's claim that 90% of shared JEV use cases are larping, LinusEkenstam argues it's a skill issue: most vibe coders treat JEV like an LLM when it's actually a classifier outputting probabilities, scores and confidence. He used JEV to classify all 80k+ of his emails, then fed that data to an LLM, making sorting and cleaning 100x more efficient. Most failed replications, he says, come from not knowing how to use it.
More from Apps
- Leonardo launches its Agent: one prompt flow from character to album art to ads — aziz4ai · 2026-09-23
- NYT Reporter Hands His Life to Meta's Muse Agent: It Called His Insurer and Booked Dinner — firstadopter · 2026-09-23
- NYT Raves About Meta's Muse Agent, Which Canceled Duplicate Subs and Saved $44.99 — firstadopter · 2026-09-23
- User says AI agent Muse negotiated and helped him buy a car end-to-end — danielrock · 2026-09-23
- HF Tutorial: Real-Time Multi-Speaker AI with NVIDIA Nemotron 3 Diarization — Hugging Face Blog · 2026-09-23
- Amazon launches AI agents that manage listings and inventory for third-party sellers — Polymarket · 2026-09-23