The Best Model Loses

2025-12-17 · Source: The Leverage · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Novice, medium

Summary

The article explores several facets of the evolving AI landscape, from personal empowerment to market dynamics and regulatory responses. It highlights Keith, a non-programmer, who built a personal AI-driven software suite on a Mac mini, Raspberry Pi, and Google Pixel, demonstrating AI's capacity to enable individual creation. The piece also examines Thinking Machines' Inkling, a 975B-parameter open-weights model, which, despite a \$10M-\$20M pretraining cost, is positioned as an advertisement for its Tinker fine-tuning platform, quickly depreciating against new models like Moonshot's Kimi K3. Furthermore, Meta's Hyperion data center in Louisiana is expanding to 5 gigawatts at a cost exceeding \$50 billion, with Meta entering the compute-selling market, including a potential \$10 billion deal with Anthropic, though these large contracts often feature short cancellation terms. Finally, Major League Baseball recently banned generative AI in dugouts, where about a third of teams used chatbots for strategic decisions, illustrating AI's pervasive impact on competitive environments.

Key takeaway

For AI developers and strategists evaluating market opportunities, recognize that open-weight models face rapid depreciation, making sustained competitive advantage difficult without continuous, costly retraining. If you are investing in AI infrastructure, be wary of the short-term commitment structures prevalent in large compute contracts, as they introduce significant revenue instability. Consider AI's potential to empower non-programmers for personal projects, but also anticipate its pervasive impact on competitive fields like sports, requiring adaptive strategies.

Key insights

AI democratizes creation, but its commercial models face rapid depreciation and unstable compute markets.

Principles

• Open-weight AI models depreciate rapidly.
• Compute contracts often lack long-term commitment.
• AI enables personal, independent creation.

Method

The article suggests testing new AI models like Anthropic's Fable or GPT-5.6 Sol using prompt guides to personally assess their capabilities.

In practice

• Build personal AI tools on consumer hardware.
• Test new frontier models with custom prompts.
• Evaluate AI model longevity for investment.

Topics

• AI Democratization
• Open-Weight Models
• AI Compute Infrastructure
• Data Center Economics
• Generative AI Applications
• AI Market Dynamics

Best for: CTO, VP of Engineering/Data, Director of AI/ML, Entrepreneur, Investor, General Interest

Related on AIssential

• Elon Musk's xAI reportedly trained its coding models on Claude outputs for months before getting cut off — xAI reportedly trained its coding models on competitor outputs, facing internal turmoil and now renting out its acquired compute.
• Together AI positions open-weight AI models as the enterprise moat for cost, control and IP — Open-weight AI models offer enterprises superior cost, control, and IP protection for agentic AI workloads.
• Capital, Not Compute, is the Real AI Bottleneck — Capital, not compute, is the primary bottleneck in AI infrastructure expansion, driven by financing and build-out timelines.
• The AI Data Center Debt Trap: Why the Buildout Is Doomed and the Bubble Is About to Pop — The AI data center buildout is unsustainable due to diminishing returns, open-source competition, and infrastructure delays.
• Why Are Tech Giants Spending $700 Billion on AI Infrastructure If the Race Is About Models? — The AI competitive advantage is shifting from models to foundational compute infrastructure, which scales like heavy industry.
• China catches up — The "no moat" dynamic, driven by inefficient and replicable AI, leads to price wars and unsustainable profits.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Leverage.