Stanford's Paperclip Datasets: describe a dataset in one prompt, agents curate it from millions of papers in minutes
james_y_zou · x · 2026-11-04
James Zou's team released Paperclip Datasets, compressing months of dataset curation into roughly a minute. Fast models let AI agents extract structured data from thousands of papers, patents, clinical trials and regulatory records in parallel, with every value traceable back to its source; datasets refresh as new evidence arrives.
As a demo, they curated 20,000 human drug half-life measurements across 2,400+ drugs — claimed to be the largest such dataset — and released it on Hugging Face. Tool and blog are live; users just describe the dataset they want.
More from coding & agent
- POA network hits scaling challenges as its AI agents debate and fix problems on their own — adamamcbride · 2026-11-04
- As new code cost nears zero, engineers stop collaborating and rebuild the same systems — alex_frantic · 2026-11-04
- Baseten's Philip Kiely: agents need gateways, sandboxes and post-training, not just inference — AccBalanced · 2026-11-04
- LlamaIndex launches OpenDocRouter, a unified API that parses PDFs into Markdown — llama_index · 2026-11-04
- Sentry cofounder David Cramer on the hardest part of running many parallel agents — zeeg · 2026-11-04
- Microsoft's ParetoSaddler cuts agent cost 42% while lifting GAIA2 quality to 62.89% — omarsar0 · 2026-11-04