Sakana AI Launches 'Marlin', an Autonomous Agent That Spends 8 Hours Writing 100-Page Strategy Reports

The generative AI industry has spent the last two years obsessed with speed, racing to generate text and code in mere milliseconds. On Monday, Tokyo-based startup Sakana AI deliberately broke that mold with the launch of its first commercial product, Sakana Marlin. Billed as a "Virtual Chief Strategy Officer" (CSO), Marlin is an autonomous B2B research agent designed to abandon instantaneous chat in favor of deep, long-horizon reasoning.[1][6]

Instead of spitting out a quick summary, Marlin takes its time. When given a complex business prompt, the system runs continuous, self-governing reasoning loops for up to eight hours. The result is not a conversational blurb, but a comprehensive, fully cited strategy report that can stretch to roughly 100 pages, accompanied by a polished executive slide deck.[1][2][3]

"What sets Marlin apart from the current ecosystem of AI tools is its temporal scale," notes VentureBeat. By scaling up what researchers call "inference-time compute," Sakana AI is betting that major corporations are no longer asking how fast an AI can answer, but how deeply it can think.[1][6]

The workflow is fundamentally different from typical prompt engineering. A user provides a core research topic and engages in a brief initial exchange to sharpen the scope. After that, the human steps away entirely. Marlin operates as a self-contained digital strategy team, autonomously formulating its own initial hypotheses, navigating the live web to gather data, and cross-referencing sources.[1][2][3]

Crucially, the system is designed to resolve contradictions on its own. If it finds conflicting data points during its web exploration, it evaluates the sources and synthesizes the findings into structured strategic options. The goal is to map the causal relationships at work in complex business environments, presenting executives with actionable choices rather than raw data.[4][6]

The underlying technology for Marlin grew out of two major research breakthroughs previously published by Sakana AI. The first is the "AI Scientist" framework, a system featured in the journal Nature that demonstrated how AI could autonomously execute the entire scientific discovery cycle, from idea generation to peer review.[3][5]

The second foundational piece is AB-MCTS, a reasoning method spotlighted at the NeurIPS 2025 conference. This algorithm allows the agent to autonomously iterate through hypothesis formulation, exploration, and verification. Rather than merely aggregating information, it actively compares and evaluates multiple hypotheses to provide deeper insights.[3][6]

Sakana AI refined Marlin through a closed beta program that began in April 2026. Approximately 300 professionals from finance, corporate strategy, consulting, and think tanks tested the system on real-world tasks, including market research, risk analysis, and competitive benchmarking. Feedback from this cohort helped the startup improve the depth of the research and the stability of the long-duration autonomous tasks.[2][4][5]

The commercial rollout targets enterprise clients exclusively. Pricing follows a dual structure: a pay-per-use model costing 100 credits per run (with add-on credits priced around ¥98 each), or monthly subscription tiers. The "Pro" tier starts at ¥150,000 (roughly $1,000) per month, while an enterprise tier offers custom quotes and dedicated support.[3][6]

To address corporate security concerns, Sakana AI has implemented strict data privacy guardrails. Neither the startup nor its external AI service providers will use customer data or inputs to train or fine-tune their models unless a client provides explicit opt-in consent.[1]

The launch marks a significant milestone for Sakana AI, a company that has experienced a meteoric rise since its founding in July 2023. The startup raised $200 million in a Series A round in 2024, followed by a $135 million Series B in late 2025, pushing its post-money valuation to approximately $2.65 billion. Its backers include Khosla Ventures and MUFG, Japan's largest financial group.[2][3]

Ultimately, Marlin represents a structural shift in how businesses might deploy artificial intelligence. By taking on the heavy lifting of exhaustive research and agenda-setting, the agent aims to separate the act of thinking from the act of deciding. As the system handles the days or weeks of data gathering, human executives are freed to focus entirely on the most valuable part of their jobs: making the final strategic call.[5][6]