As CEO of Pydantic, the company behind one of the most widely used frameworks in AI development, Samuel Colvin sits at the center of the action.
Pydantic works closely with leading frontier-model labs and AI developers, giving Colvin a front-row seat to the rapid evolution of models, agents, and coding tools. Investors are paying attention too: Silicon Valley powerhouse Sequoia Capital led Pydantic’s latest $12.5 million funding round.
In a recent interview, Colvin shared where he thinks AI development is headed next, and what that means for the rest of us. This Q&A has been edited for length and clarity.
Q: You work with the big frontier labs, including Anthropic and OpenAI. How do you see their strategies evolving?
Samuel Colvin: “A year ago, what they cared about was revenue. So anything you can do to use our inference is great. Now when one assumes they’re both trying to IPO, their profit margin becomes really important. And if you want a margin, then what you don’t want is to compete just on model quality because at that point you need to spend masses of money training the best model and you need to provide inference on that model as cheaply as possible. So they are doing their very best to find ways of locking people in that are not related to model quality. That’s where I think Claude Code and Codex and all that work is coming from.”
Q: How are OpenAI and Anthropic moving beyond model performance as a route to success?
Samuel Colvin: “The reason we have Codex and Claude Code discounts — $200 a month subscriptions when you are actually spending maybe thousands on inference with that subscription — is obvious. They’re trying to grow market share. They’re trying to get as much usage as possible. There’s perhaps a more profound thing they may be trying to do, though. Once customers have these enormous code bases, which would be basically written AI, you get to a point where you can’t maintain them as a human. If I’ve used AI to generate 20,000 lines of code overnight, I can use a model to go and fix that, but as a human, I can’t ever go and maintain that code.”
With such huge code bases, corporate customers will have to keep using these AI coding services from Anthropic and OpenAI. Then, once that usage is locked in, these companies will likely raise prices, Colvin told me.
Q: How are these coding offerings from OpenAI and Anthropic changing?
Samuel Colvin: “My personal guess is that, fairly soon, they’re going to say, ‘if you use us through a corporate subscription, we don’t just do the coding generation and coding agent stuff, we’ll also store the traces or trajectories of the full exchange between users and the model as it’s writing code.’ And then you will have a database where you can look up, for any line of code in your code base, the intent when that code was written. The argument will be that it’s really useful, and that it makes coding agents even better. And then they’ll probably say, ‘we give you that for free, but you can’t export it.’ So now you’re locked into whoever you’re using for that across the whole business.”
Q: Explain this a little more. How would this work?
Samuel Colvin: “Imagine you have a software bug. You have some odd behavior on some line of code. Maybe the developer, human or AI, who wrote that line of code left a helpful explaining comment about what was going on there. Now you can have too many comments. You can have five lines of comments for every line of code and it gets impossible to read. But imagine if I could click on that line of code and see the full exchange that my colleague had with the AI model to write that line of code, along with all of the reasoning, including the reasoning from the model, the input from the human, therefore a full explanation. Now I have this much richer understanding of the intent behind my code base. And so going in and changing that line of code becomes lower risk because I can know what someone was trying to do when they wrote it and work out whether that’s really a bug or whether that’s some intended piece of behavior.
“I think this idea of basically ‘we store your trajectories and we give you some database of your trajectories’ is attractive and valuable. Those two things are not necessarily the same, but on this occasion, it will be actually valuable as well as being attractive.”
Sign up for BI’s Tech Memo newsletter here. Reach out to me via email at abarr@businessinsider.com.
