The AI Scientist, from Sakana AI, is an open-source system that automates the research loop: generating ideas, running experiments and writing full papers.
The AI Scientist takes a research direction and executes the full scientific process: brainstorming novel ideas, implementing experiments, analyzing results, and drafting papers with figures and citations. It represents a provocative step toward end-to-end automated research.
Released openly by Sakana AI alongside innovations like model merging and evolutionary model discovery, the system sparked wide discussion about the future of scientific work. Its codebase is extensible for different domains, and it produces complete, reviewable paper drafts that human researchers can build on. Watching the system propose, test and write up ideas reframes what automated research might look like within a few years. Extensibility means labs can adapt the loop to their own domains and experimental infrastructure.
The AI Scientist is free and open source. Costs come from the model APIs used during its research loops, which can be reduced with local or cheaper models. Free code pairs with model costs that vary enormously depending on which providers you use for the research loop.
The AI Scientist is for researchers exploring automated science, labs prototyping AI-driven discovery pipelines and anyone interested in the frontier of machine-generated research. It also suits philosophy and policy groups studying the implications of machine-generated science.
Yes, the AI Scientist is completely free and open source. Model API usage during runs is the only cost, which varies with the models you choose.
It is a codebase rather than a hosted service, so researchers run and extend it themselves, integrating it into their own experimental setups.
The AI Scientist is unique in automating the full research loop, where assistants like Elicit support human-led research. It is experimental and uneven, but its ambition makes it essential reading for the field. The AI Scientist is experimental rather than dependable; for rigorous production research support, established assistants remain safer. Output quality is uneven and human oversight remains essential, but as a working prototype of automated research it offers a concrete glimpse of how science tooling is evolving.
SciSpace is an AI research and writing platform that lets you search 200 million papers, chat with PDFs, and draft literature reviews…
Jenni AI is an academic writing assistant that helps students and researchers draft essays and papers with accurate citations as they type.…
Paperpal is an AI academic writing and editing tool trained on millions of scholarly manuscripts to make research papers submission-ready. OverviewPaperpal is…