Yoav Goldberg: For Recurring Tasks, Tune Bespoke Predictors Instead of Always Using Reasoning LLMs
yoavgo · x · 2026-09-23
AI researcher yoavgo (Yoav Goldberg) sketches an engineering idea in a short thread: many tasks don't need a reasoning LLM — defining a task in text and executing it directly (à la Jev) is already neat. But for recurring tasks, he suggests going further: collect examples and tune a tailored predictor of your own. He adds that you could even design bespoke feature extractors for the variables the predictor should consider. The takeaway: instead of routing every task through heavy LLM reasoning, distill high-frequency tasks into small specialized models.
More from coding & agent
- Rivet launches SQLite Forking: 20ms agent-thread forks with zero data copying — samgoodwin89 · 2026-11-11
- YouArt Desktop launches: AI agent edits video in CapCut, DaVinci and Premiere from client feedback — mhdfaran · 2026-11-11
- Boubou: one-command local voice AI that serves your phone and robots over Wi-Fi — LuozhuZhang · 2026-11-11
- Python 3.15 makes AttributeError actionable: suggests .append for .push mix-ups — KhuyenTran16 · 2026-11-11
- sentence-transformers maintainer releases new embedding pooling training method, code & checkpoints — tomaarsen · 2026-11-11
- droid-mcp turns your Android phone into an MCP server with 160+ tools for Claude Code — zvone1122 · 2026-11-11