Sakana AI Launches First Commercial Product "Sakana Marlin"
Sakana AI ● Covered by 2 sources
Sakana AI just launched Marlin, an AI that does days of business research on its own. Give it a topic and it grinds for 8 hours, spitting out a full strategy report — no analyst team required.
Based on reporting by Sakana AI — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Sakana AI, the Tokyo-based lab known more for research papers than shipped products, has crossed into commercial territory with Marlin, an autonomous research assistant aimed squarely at corporate strategy teams. Point it at a topic, answer a few clarifying questions, and then walk away. Marlin will chew on the problem for up to eight hours, forming hypotheses, hunting down sources, checking its own work, and eventually handing back a set of executive summary slides plus a report that can run close to a hundred pages.
The pitch is specific: this isn't a chatbot that summarizes search results, it's meant to replace the multi-week slog a CSO and a small team would normally do by hand — mapping causal relationships in a market, then structuring findings into strategic options a leadership team can actually act on. Sakana frames the value proposition as freeing humans from the grunt work of comprehensive research so they can spend their time on the actual decision, not the digging.
Under the hood, Marlin is less a single model than a stitched-together product of Sakana's prior research bets. The company points to The AI Scientist, its Nature-published system for automating scientific discovery, AB-MCTS, a technique for coordinating multiple models to improve reasoning that picked up a NeurIPS 2025 Spotlight, and ALE-Agent, which automates algorithm engineering. All three feed into Marlin's long-horizon reasoning and multi-model orchestration.
This isn't a cold launch. Sakana ran a closed beta starting in April 2026 with roughly 300 professionals from banks, corporates, consulting firms and think tanks, who put Marlin to work on real strategy formulation, market research, risk analysis and competitive analysis tasks. Feedback apparently praised the depth of research compared to chat-based tools, but also flagged rough edges in output formatting and report structure — issues Sakana says it addressed before the public release.
Pricing follows a familiar SaaS ladder: a free-to-start pay-per-use tier, then Pro, Team and Enterprise plans, available immediately on a self-serve basis, though subject to regional availability. Sakana bills this as step one of a broader plan to move past chat interfaces entirely, betting that coordinating multiple models over long time horizons — not single bigger models — is where the real capability gains are hiding.
My take — AI-written commentary, not fact-checked reporting
An AI that works unsupervised for eight hours and hands you a hundred-page strategy report is exactly the kind of claim I want to see stress-tested by someone who isn't Sakana's PR team, because 'CSO-quality research, unattended' is a huge promise resting on a 300-person beta. That said, I like that this is a narrow, workflow-specific product rather than another everything-chatbot — that's the more honest way to sell autonomy right now, and if the reports genuinely hold up under scrutiny from actual strategy teams, this is a template other labs should be copying instead of shipping yet another general assistant.
Read more about this at: Sakana AI
