Open-Science: open-source workbench turns coding agents into auditable research assistants
rohanpaul_ai · x · 2026-10-20
AIPOCH released Open-Science, a free Apache-2.0 desktop workbench that turns coding agents into research assistants with a full provenance trail. Agents work in Python/R notebooks with scientific skills (AlphaFold2, ESM2) and connectors to PubChem, GTEx and cBioPortal, with tiered autonomy. Every output is stored as an immutable checksummed version; a Provenance view shows producing code, inputs, environment and execution history, and built-in replay reruns any version in isolation for comparison. Sessions export as portable .science packages. Engines are pluggable — Claude Code, Codex, OpenCode or CodeBuddy via cloud APIs, Ollama/vLLM local models, or existing subscriptions — projects stay local by default, and big jobs can go to your own server over SSH/Slurm. AIPOCH reports 79.05 with gpt-5.6-sol on BiomniBench-DA public tasks, first in its comparison.
Related event: AIPOCH Open-Sources Open-Science Workbench for Traceable AI Research(2 posts)→
More from Apps
- CueForge v3.4.0 brings ComfyUI 0.42.1 parity to mobile with 3D previews and low-resource mode — galactic_lobster · 2026-10-20
- Don't ask two AIs the same question—make one grade the other's answer — Ok_Negotiation_2587 · 2026-10-20
- Papers with Code Upgrades PDF Reader with RAG-style Citations via Chandra-OCR 2 — NielsRogge · 2026-10-20
- User Builds Automated Meal-Planning Workflow from YouTube Influencers with Muse Agents — leveragedupside · 2026-10-20
- Adobe Lightroom mobile can now open closed eyes in photos with one tap — The Verge AI · 2026-10-20
- Lovable launches Enterprise: deploy apps in your own AWS/Azure cloud with EU inference — AlexandrePesant · 2026-10-20