Most AI chatbots pride themselves on speed. Sakana AI just built one that takes its sweet time on purpose.
The Tokyo-based AI research startup launched Sakana Marlin on June 15, its first commercial product. Rather than spitting out instant responses like every other chatbot on the market, Marlin is designed to run autonomous research tasks for up to eight hours, producing comprehensive strategy reports that can stretch to roughly 100 pages.
A virtual Chief Strategy Officer, minus the corner office
Sakana AI is positioning Marlin as a “Virtual CSO,” a B2B research agent that mimics the strategic thinking process of a senior executive. The system autonomously forms hypotheses, gathers information from online sources, reconciles conflicting data, and synthesizes everything into polished deliverables, including executive summary slides.
The product grew out of two notable research efforts at Sakana AI. One is the AI Scientist framework, which earned a feature in Nature. The other is AB-MCTS, a reasoning method that was spotlighted at NeurIPS 2025.
Before going public, Sakana ran a closed beta starting in April 2026 with approximately 300 professionals from sectors including finance and consulting. Pricing follows a dual structure: a pay-per-use model at 100 credits per run, or monthly plans starting at ¥150,000 (roughly $1,000 at current exchange rates).
The money behind the machine
Sakana AI is not short on capital. Founded in July 2023, the company raised $200 million in its Series A round in 2024. It followed that up with a Series B of approximately $135 million in late 2025, landing at a post-money valuation of about $2.65 billion.
Backers include MUFG, Japan’s largest financial group, and Khosla Ventures, the Silicon Valley firm known for aggressive AI bets.
What this means for crypto and finance
Sakana Marlin is not a crypto product. It doesn’t trade tokens, manage wallets, or run on-chain. But among the sample analysis topics Sakana has highlighted are regulatory changes for stablecoins in Japan and enterprise AI agent market mapping, both of which sit squarely at the intersection of fintech and digital assets.
The risk, as always with AI agents operating without human oversight for hours at a time, is hallucination at scale. A chatbot that makes something up in a two-paragraph response is annoying. An autonomous agent that builds a flawed assumption into page 12 of a 100-page strategy report, then compounds it through 88 more pages of analysis, is a different kind of problem entirely.