---
id: "2026-10-13-sakana-agentic-scientist"
url: "https://postcutoff.com/e/2026-10-13-sakana-agentic-scientist/"
as_of: "2026-10-17T02:25:00+02:00"
date: "2026-10-13"
date_precision: day
category: agents
importance: 2
confidence: high
status: [Confirmed]
sources: 1
editor: human editor
human_review: null
version: "2026-10-16"
---

As of: 2026-10-17 02:25 CEST.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/e/2026-10-13-sakana-agentic-scientist/

# Sakana AI's Agentic Scientist orchestrates coding agents for research

Full title: Sakana AI's 'The Agentic Scientist' orchestrates coding agents through the research cycle; a co-scientist run produced a paper accepted at ALIFE 2026

On 13 Oct 2026 Sakana AI researchers (Robert Tjarko Lange, Luca Grillotti, Yutaro Yamada, Rujikorn Charakorn, David Ha) posted arXiv 2610.18933, "The Agentic Scientist", the successor to The AI Scientist. A meta-controller runs parallel coding agents through task drafting, ideation, planning, sandboxed experiments, baselines and ablations, analysis and report writing, with rubric checks at every stage. As a co-scientist tool a development version produced a paper accepted at ALIFE 2026; run autonomously it was applied to open machine-learning problems.

## Key facts
- Paper: 'The Agentic Scientist: Orchestrating Coding Agents for Automated Scientific (Co-)Discovery', arXiv 2610.18933, v1 13 Oct 2026, 88 pages
- Motivation (abstract): fully autonomous research agents still have a narrow range of ideas and search in a myopic, hill-climbing way
- Design: distributed pipeline of stages (task drafting, idea generation, planning, sandboxed experimentation, result aggregation, baseline and ablation construction, post-hoc analysis, report writing), each run by coding agents with self-verification; a meta-controller coordinates parallel agents and enforces rubric checks; independent of domain, execution environment and coding harness
- Results: in co-scientist mode a development version produced a paper accepted at ALIFE 2026; in autonomous mode it was applied to open ML problems in several computational domains
- Ablations: reliable autonomy depends on the cross-agent orchestration layer as well as on base-model capability
- No Sakana blog post or press coverage found as of 16 Oct (sakana.ai/agentic-scientist returned 404)

## What happened

Sakana AI, whose AI Scientist v2 had a workshop paper pass peer review in 2025, posted a new system built on general coding agents
rather than a fixed research script. Human researchers can work with it stage by stage, or it can run an investigation end to end.

## Why it matters

It is an AI lab's statement that orchestration across many agents, not only stronger base models, is what makes automated
research reliable, and it adds a further accepted conference paper to the record of AI-generated research.

## Your AI and this story

- GPT-6 Astra (training cutoff April 2026): 166 days after its cutoff
- Claude Opus 5.5 (training cutoff June 2026): 105 days after its cutoff
- Gemini 4 Argon (training cutoff March 2026): 196 days after its cutoff
- Grok 4.7 (training cutoff May 2026): 135 days after its cutoff

## Sources

1. [arXiv 2610.18933: The Agentic Scientist (Lange, Grillotti, Yamada, Charakorn, Ha)](https://arxiv.org/abs/2610.18933) (arxiv.org, paper)

## Changes

- 2026-10-16 (filed): Created (arXiv abstract read)

## Related

- 2025-03-12: [Sakana's AI Scientist-v2 writes the first fully AI-generated paper to pass peer review](https://postcutoff.com/e/2025-03-12-sakana-ai-scientist-v2-peer-review/index.md)