Jev agent demo: classifies bribes, threats and pleas with no keyword matching
chongdashu · x · 2026-09-20
A demo of the agent "Jev": type anything and it classifies the message as bribe, threat, plea or small talk (even politeness level) — no keyword matching, the model judges it — then picks the character's line accordingly. The author frames agentic work with a vivid analogy about a dog let in for crumbs leaving mud down the hallway: with 15 minutes before Mom gets home, you can't fix everything, you have to choose.
More from coding & agent
- Vibe coder can't read AI-generated code: black-box testing alone hits a wall — Ryan_yii · 2026-09-20
- Projects keep breaking: vibe coder asks for a full practical workflow — Ryan_yii · 2026-09-20
- Open-source testsprite CLI lets coding agents self-test live backend APIs and auto-fix failures — KARYTON618 · 2026-09-20
- OpenEnv sees active community work on infra providers, MCP and WebSocket fixes — ben_burtenshaw · 2026-09-20
- Gemma 26B A4B Aces a C++ Coding Test Locally but Fumbles Tool Calls — HyperWinX · 2026-09-20
- mitsuhiko on the Shared Frustrations of Agentic Software Engineering — mitsuhiko · 2026-09-20