The AI industry is quietly abandoning its obsession with speed. Instead of generating answers in milliseconds, a new wave of enterprise AI tools is spending hours reasoning through complex problems. Sakana AI, a Tokyo-based startup, just launched Marlin, a "Virtual Chief Strategy Officer" that deliberately takes up to eight hours to produce deeply researched, 100-page strategy reports complete with executive summaries and citations. This represents a fundamental shift in how businesses expect AI to work.

What Makes Marlin Different From Typical AI Chatbots?

For the past two years, the generative AI hype cycle has centered on speed. Companies raced to build systems that could generate poems, code snippets, or summaries in mere milliseconds. But Marlin takes the opposite approach. Instead of a quick back-and-forth conversation with a chatbot, users provide a research topic and then step away entirely. The system operates autonomously for hours, formulating hypotheses, gathering data from the web, cross-referencing sources, and mapping out complex business dynamics. Think of it less like a search engine and more like a junior strategy consultant locked in a room with a whiteboard and internet access.

The product is designed exclusively for enterprise use, targeting corporations, financial institutions, and think tanks. Sakana demonstrated Marlin's capabilities with real-world examples, including generating detailed resolution scenarios for a theoretical blockade of the Strait of Hormuz, mapping the fragmented global AI regulation landscape, and analyzing macroeconomic trends like the return of "bond vigilantes". Pricing starts with a pay-as-you-go tier and is available immediately through the company's website.

How Does Marlin Actually Reason Through Complex Problems?

Under the hood, Marlin relies on a technology called Adaptive Branching Monte Carlo Tree Search, or AB-MCTS, which Sakana AI developed and first introduced publicly in June 2025. The company released the underlying algorithm as open-source software called TreeQuest under the Apache 2.0 license to encourage developer experimentation. To understand how this works, consider how modern chess engines operate. Rather than simply looking at the board and guessing the best move, chess engines evaluate thousands of potential future moves before committing to an action. Marlin's AB-MCTS engine does something similar for research and strategy work.

Traditionally, when developers want higher-quality reasoning from large language models, or LLMs (AI systems trained on vast amounts of text), they rely on "repeated sampling," essentially running the model dozens of times in parallel and hoping one answer is correct. But repeated sampling operates blindly; it cannot evaluate its own intermediate steps or adjust based on feedback. AB-MCTS replaces this approach with a principled, multi-turn method driven by a Bayesian decision framework, which mathematically balances different search strategies.

Steps to Understanding How AB-MCTS Improves AI Reasoning

• Exploration vs. Exploitation: As the AI constructs a strategy report, the system treats the research process as a branching tree of possibilities. At each node, the algorithm dynamically balances spawning entirely new hypotheses when the current path yields diminishing returns, versus methodically refining an existing solution that shows high promise.
• Multi-Model Orchestration: Sakana AI's architecture introduces the ability to dynamically choose which AI model to invoke for specific sub-tasks, treating leading frontier models as a plug-and-play collective intelligence network. An orchestration model can delegate initial ideation to one LLM while using a reasoning-heavy model to audit and verify intermediate errors.
• Scaled Inference Compute: By scaling up computing power at inference time, leveraging the distinct strengths of multiple foundation models over thousands of automated cycles, AB-MCTS provides the mathematical guardrails needed to ensure 100-page strategy reports are highly vetted products of systemic, automated trial-and-error rather than long-winded AI generations.

This multi-LLM approach is what transforms AB-MCTS from a laboratory experiment into a commercial engine. Sakana did not disclose specific model names or providers powering Marlin, but the architecture allows the system to coordinate highly heterogeneous models, treating the industry's leading frontier models as interconnected components of a larger reasoning system.

Why Is Enterprise AI Shifting Away From Speed?

The enterprise frontier is rapidly shifting from shallow, rapid generation to deep, methodical reasoning. With Marlin, major businesses are no longer asking how fast an AI can answer, but how deeply it can think. This reflects a broader recognition that some problems require sustained analysis rather than instant responses. A strategy report on global AI regulation or macroeconomic trends cannot be properly researched in seconds; it requires systematic exploration of contradictions, verification of sources, and synthesis of complex causal relationships.

Marlin operates under strict enterprise-grade data handling terms, unlike many consumer-grade AI tools that silently harvest user inputs and proprietary data to train future models. For enterprises, licensing and data handling are often the determining factors in software adoption, making this distinction critical for corporate adoption.

The launch of Marlin signals that the next frontier of AI competition is not about millisecond response times, but about the depth and rigor of reasoning that AI systems can achieve when given time to think. As businesses increasingly rely on AI for strategic decision-making, the ability to produce thoroughly researched, well-cited analysis may prove more valuable than the ability to generate instant answers.