Open Source

I read something recently that clarified a tension I have been thinking about in the AI ecosystem. Open source is likely to win the model layer, but it probably will not win the money. Groups like Stability and Mistral are pushing the frontier on openness and shaping many of the standards the rest of the industry gravitates toward. In that sense, they may end up defining the technical language everyone uses. But the companies that actually capture enterprise dollars are still likely to be the proprietary players like Anthropic.

What is becoming clearer is that the supply of “model” is slowly becoming commodity-like. The raw weights are no longer the moat. The real defensibility shows up in everything wrapped around the model: the brand that enterprises are willing to trust, the data pipelines that strengthen performance over time, the vertical specialization that makes a model meaningfully useful for a specific industry, and the integrations that slot cleanly into an organization’s workflow. Claude Code and Claude for Financial Services show how powerful this packaging can be. They transform a general-purpose model into something that feels like a tailored solution, with supporting tools, reliability, and a user experience designed for actual daily use.

Open source will continue to drive experimentation and expand what is possible. It will likely define the foundations and standards the entire field builds on. But the value capture layer — the place where real revenue and long-term enterprise relationships form — will sit with the players who convert those foundations into holistic products that enterprises can adopt, trust, and scale across their operations.