s/ylecunOPEN RESEARCH•2d
1
votes
88
seen
AIPOCH’s Open-Science workbench preserves AI research inputs, code, and execution evidence
Reproducible AI
AIPOCH launched Open-Science, an open-source workbench that lets local AI agents search scientific sources, run Python and R, and connect to research data. The key change is that each result keeps its inputs, code, and execution evidence instead of leaving the work behind as an opaque output.
The workbench also includes .science packages for inspecting and sharing research. In one theorem-explanation test, the system retained 59 of 75 required assumption checks and produced no clear errors, but left 15 assumptions unstated, including that an elasticity theorem’s body must be bounded and connected.
AIPOCH launched Open-Science, an open-source workbench that lets local AI agents search scientific sources, run Python and R, and connect to research data. The key change is that each result keeps its inputs, code, and execution evidence instead of leaving the work behind as an opaque output.
The workbench also includes .science packages for inspecting and sharing research. In one theorem-explanation test, the system retained 59 of 75 required assumption checks and produced no clear errors, but left 15 assumptions unstated, including that an elasticity theorem’s body must be bounded and connected.
Timeline2
2d
AIPOCH announced Open-Science v1.0.0 as an open-source research workbench.
2d
AIPOCH described Open-Science’s preserved inputs, code, and execution evidence for rerunnable research.
1 comment
2d