ResearchPublished All news from that day
Sakana AI released The Agentic Scientist, an open-source pipeline of coding agents for research, under Apache-2.0.
Sakana AI's 88-page paper and Apache-2.0 repository describe a pipeline that runs research from task draft to written paper, and the authors report one paper accepted at ALIFE 2026.
Check our sources · 7 facts from 2 sourcesSakana AI posted the paper "The Agentic Scientist: Orchestrating Coding Agents for Automated Scientific (Co-)Discovery" on arXiv on 13 October 2026. Its authors are Robert Tjarko Lange, Luca Grillotti, Yutaro Yamada, Rujikorn Charakorn and David Ha. The code is in the SakanaAI/AI-Scientist-v3 repository under the Apache-2.0 licence.
The pipeline splits research into stages: task drafting, idea generation, planning, sandboxed experiments, analysis and report writing. A meta-controller coordinates parallel research agents and applies rubric checks at each stage.
The authors report that the system was used to write a paper accepted at ALIFE 2026 in a co-scientist setting, and that it was also run fully autonomously on open machine learning problems. The repository README reports no benchmark numbers. It warns that agents run with full tool permissions and recommends running inside Docker.
Underlined sentences link to the facts we checked. Dotted words open a short explainer.
Fact 2 of 7 · from arXiv, 13 Oct 2026
task drafting, idea generation, planning, sandboxed experimentation
Fact 3 of 7 · from arXiv, 13 Oct 2026
meta-controller that manages parallel research agents and enforces rubric verification at each stage
Fact 4 of 7 · from arXiv, 13 Oct 2026
got accepted for publication at ALIFE 2026
Fact 6 of 7 · from Sakana AI
README reports no benchmark results
What is Docker?
Docker is an open platform for developing, shipping and running applications. It packages an application and what it needs into a container, a lightweight, loosely isolated environment. The same container can run on a laptop, in a data centre or in the cloud.
Background from ai notis, not part of the news.
Docker's documentation says containers hold everything an application needs to run, which makes them portable between systems, and several can run on one machine at the same time. A container is created from an image, a read-only template built from layered instructions in a file called a Dockerfile.
The Docker client talks to a background service, the Docker daemon, which manages images, containers, networks and volumes.
Sources: Docker.
Your reaction
We count reactions per story and day, never who reacted. The counts help us choose what goes in the monthly issue. If you are signed in, your own page shows yours too.
We count reactions per story and day, never who reacted. They help choose what goes in the monthly issue.
Check our sources
Every sentence above is checked against these 2 sources.
1 The Agentic Scientist: Orchestrating Coding Agents for Automated Scientific (Co-)Discovery
Open the source archived copy-
The arXiv paper 2610.18933, 'The Agentic Scientist: Orchestrating Coding Agents for Automated Scientific (Co-)Discovery', was submitted on 13 October 2026 by Robert Tjarko Lange, Luca Grillotti, Yutaro Yamada, Rujikorn Charakorn and David Ha, and is 88 pages.
citeThe Agentic Scientist: Orchestrating Coding Agents for Automated Scientific (Co-)Discovery; submitted 13 October 2026; 88 pages
-
The abstract says the pipeline splits research into stages such as task drafting, idea generation, planning, sandboxed experiments, analysis and report writing, with each stage using agentic self-verification.
citetask drafting, idea generation, planning, sandboxed experimentation
-
The abstract says a meta-controller coordinates parallel research agents and applies rubric checks at each stage.
citemeta-controller that manages parallel research agents and enforces rubric verification at each stage
-
The abstract says the system was used to write a paper accepted at ALIFE 2026 in a co-scientist setting and was also run fully autonomously on open machine learning problems. the company says
citegot accepted for publication at ALIFE 2026
2 SakanaAI/AI-Scientist-v3
Open the source archived copy-
The SakanaAI/AI-Scientist-v3 repository is licensed Apache-2.0.
citeApache-2.0 license
-
The repository README reports no benchmark results or performance numbers.
citeREADME reports no benchmark results
-
The README says agents run with full tool permissions (--allow yolo) and can run arbitrary commands, and recommends running inside Docker.
citeAgentic Scientist launches coding agents with full tool permissions (Headless `--allow yolo`). They can run arbitrary commands and inherit the API keys you give them
Every sentence above is checked against these 2 sources. Each fact has its own address you can cite.
Source 1 · arXiv · 13 Oct 2026
The Agentic Scientist: Orchestrating Coding Agents for Automated Scientific (Co-)Discovery
The arXiv paper 2610.18933, 'The Agentic Scientist: Orchestrating Coding Agents for Automated Scientific (Co-)Discovery', was submitted on 13 October 2026 by Robert Tjarko Lange, Luca Grillotti, Yutaro Yamada, Rujikorn Charakorn and David Ha, and is 88 pages.
The Agentic Scientist: Orchestrating Coding Agents for Automated Scientific (Co-)Discovery; submitted 13 October 2026; 88 pages
The abstract says the pipeline splits research into stages such as task drafting, idea generation, planning, sandboxed experiments, analysis and report writing, with each stage using agentic self-verification.
task drafting, idea generation, planning, sandboxed experimentation
The abstract says a meta-controller coordinates parallel research agents and applies rubric checks at each stage.
meta-controller that manages parallel research agents and enforces rubric verification at each stage
Show all 4 factsShow fewer facts
The abstract says the system was used to write a paper accepted at ALIFE 2026 in a co-scientist setting and was also run fully autonomously on open machine learning problems. the company says
got accepted for publication at ALIFE 2026
Source 2 · Sakana AI
SakanaAI/AI-Scientist-v3
The repository README reports no benchmark results or performance numbers.
README reports no benchmark results
The README says agents run with full tool permissions (--allow yolo) and can run arbitrary commands, and recommends running inside Docker.
Agentic Scientist launches coding agents with full tool permissions (Headless `--allow yolo`). They can run arbitrary commands and inherit the API keys you give them
Topics
The morning email
On the mornings we publish: the three top stories and up to four short ones. Free.
We email you a link to confirm. An issue may include one sponsor, always labelled Sponsored · Advertisement. Our emails count opens and clicks, not who made them. Unsubscribe in one click. What we keep