spaCy creator Honnibal shares a cheap-LLM ensemble recipe for building NER data with coding agents
honnibal · x · 2026-11-23
Matthew Honnibal (spaCy/Explosion founder) shares a laziness-compatible recipe for creating NER data now that frontier models are almost—but not quite—good enough and expensive:
- Pipeline: run data through a panel of cheaper LLMs → accumulate correction rules → build a gazetteer → ensemble outputs with the Dawid-Skene algorithm → binary-annotate only the disagreements → rerun Dawid-Skene trusting your annotations so correct labels aren't outvoted.
- Key point: coding agents let you operate at a much higher abstraction level, making previously "nice to have" steps cheap to include and NLP/ML solutions far more accessible.
- Necessary hurdle: you still need a little human annotation—LLM-as-judge alone gives nothing to calibrate against, and you can't even specify requirements well without knowing the data.
Related event: spaCy Creator Open-Sources 'Vibe NLP' Pipeline for Low-Cost NER Data(2 posts)→
More from coding & agent
- Only-Knowing-Code Developers at Risk as the Agentic Software Factory Reshapes Engineering — Pavan_Belagatti · 2026-11-23
- MCP Tool Gateways Should Fail Loudly: A Plausible Answer From a Missing Dependency Is the Real Bug — Emergency-Tutor-7 · 2026-11-23
- Codex Can't Route PCB Traces: Devs Still Lack a Workable AI Tool for Hardware's Last Mile — goldvenetianmask · 2026-11-23
- Uncensored 26.4B model runs on a 12GB Mac at 85 tok/s with 8 parallel agents via Locals — SumitGup · 2026-11-23
- 398M tokens, 98% cached: how one dev ran 12.5h of agent game dev on 3% of a $200 Claude plan — Yamapama · 2026-11-23
- Harvard paper: higher SWE-Bench scores didn't translate into more merged pull requests — rohanpaul_ai · 2026-11-23