Gustavo Monnerat, Deputy Editor at The Lancet, shared on LinkedIn:
“You can now turn a paper into an agent that runs its methods on your data.
Paper2Agent analyses a paper plus its codebase, builds the tools, tests them against the paper’s own outputs, and drops whatever fails.
- The AlphaGenome version took 45 minutes with 22 tools, and it’s live on Hugging Face right now.
- They ran it on 100 bioRxiv papers and 74 worked. The rest were missing code or data, or had environments that wouldn’t rebuild.
Title: AI tool turns any paper into an ‘agent’ that can collaborate and answer complex queries
Author: Kaia Glickman.
Read The Full Article.
I tested with Claude Code and Codex on a couple of papers.
Do you think tools like this meaningfully improve reproducibility in science?
Ref: Miao et al, Reimagining research papers as interactive and reliable AI agents; Nature 2026.“
You can also read: Multi-Modal AI May Refine Adjuvant Escalation in High-Risk HR+/HER2− Early Breast Cancer
