The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery
What this work does
Orchestrates idea generation, novelty checking, code edits, executed experiments, plotting, manuscript writing and automated reviewing around supplied research templates.
Connection to discovery
An end-to-end harness baseline and a source of candidate trajectories for studying research-stage failures.
Evidence
Inspected the pipeline, reviewer calibration and a generated code/manuscript case study.
Scope and limitations
Automated reviewer scores and completed manuscripts do not independently establish scientific correctness or novelty. Initial templates and editable evaluation code are important experimental controls. The expanded Nature 2026 version is separately downloaded and body-screened in the version audit; it explicitly reports human filtering for workshop submissions and distinguishes workshop from main-conference acceptance.
THE RESEARCH CONVERSATION
Discussion & reactions
Share a question, an insight, or a replication experience. Sign in with GitHub to join the conversation.