Author: Lin Qiao
Translation: Deep Tide TechFlow
Deep Tide Insight: The sudden shutdown of Mythos this week directly exposed a fatal risk overlooked by most founders: when your core capabilities are completely dependent on someone else's platform, your fate is no longer in your own hands. Who truly owns the intelligence that your product relies on to function?
Mythos was shut down this week. Whether you agree with the decision or not is almost besides the point.
A company built on intelligence it cannot control suddenly found itself exposed to decisions it could not influence. Upon seeing this, many founders asked themselves the same question: which parts of my business are essentially just renting?
For the past few years, discussions about open-source models have mainly revolved around cost. Can they really get the job done? If so, how much cheaper are they than calling cutting-edge APIs?
We now have a fairly clear answer. We have worked with companies like Ramp, Cursor, and Harvey, employing the same basic approach: start with a powerful open-source model, perform post-training for the work that truly matters to the business, and rigorously evaluate it against frontier models.
The results have consistently been surprising. On the tasks they care most about, fine-tuned open-source models can achieve frontier model quality at an extremely low cost. What happened this week made one thing clear: cost was never the most important issue.
The deeper question is control. Who owns the intelligence that your product relies on?
Much of the recent discussion has been framed as renting versus owning. It's not a perfect analogy, but it's useful.
Renting Intelligence
Renting is great until it isn't. Apartments are move-in ready. The lights work. The plumbing is connected. Someone else handles maintenance. That's why most companies start here.
Frontier APIs are incredible products. They allow startups to build things that seemed impossible just a few years ago.
But renting has limitations. Landlords can raise the rent. They can decide what modifications you can make. They can change the rules. Occasionally, for reasons that have nothing to do with you, they will tell you it's time to leave.
You didn't do anything wrong. You're just operating on someone else's property. This is why the Mythos story resonated with so many people. When your core capability is completely dependent on someone else's platform, you are exposed to decisions you cannot control.
Most of the time, this doesn't matter. Sometimes it suddenly matters a lot.
Owning Intelligence
The lesson is not that companies should stop using frontier models. Quite the opposite. Frontier labs have built extraordinary technology. Most products should use it. We use it too. In many ways, frontier models are becoming infrastructure. But infrastructure and ownership are two different things.
You can use public infrastructure while still owning what creates value for your business. In AI, ownership means starting with the most advanced open-source models and shaping them around the unique aspects of your company.
Your data.
Your workflows.
Your domain expertise.
Your edge cases.
Your evaluation criteria.
Your definition of "good".
Over time, the model becomes less generic and more reflective of the work your company does every day. This is where value is created.
Think of a house. Moving furniture is easy. Painting walls is easy. But if your future depends on the layout itself, eventually you'll want the ability to move walls. The same is true for intelligence.
When the intelligence belongs to you, no one can quietly pull the foundation out from under your product.
This is why we built Fireworks this way.
Training and inference under one roof, so companies can adopt the best open-source models, shape them around the problems that matter most to the business, and deploy them reliably in production.
Not just consuming intelligence. Owning it.
There Is No Single Frontier
An optimistic takeaway from this week is that the future of AI does not depend on the victory of a single model.
There is no single frontier. There are many frontiers.
Frontier models are one frontier.
Models post-trained on years of proprietary company knowledge are another.
Specialized models that solve a narrow problem better are another.
A router that maps requests to an ensemble of models, collectively outperforming any single model on many tasks, is another frontier.
The most interesting thing in AI is not that a single model is getting smarter. It's that intelligence is becoming increasingly customizable. The winning companies won't necessarily be the ones with the biggest models. They will be the ones who turn intelligence into unique, owned assets.
Looking Ahead
While everyone was reacting to the news this week, we were busy shipping products—Kimi Moonshot K2.7 Code, MiniMax M3, Alibaba Qwen 3.7 Plus.
The future I look forward to is not one where a single model quietly eats everything it sees. It's one where many teams own the part of the frontier that matters to them.
If the shutdown of Mythos has made you think differently about this trade-off, we'd love to chat.
