Are AI Models the Streaming Services of 2025 And Why I'm Not Losing Sleep Over the Latest Releases.

A while ago, I decided to set aside my initial fear of AI and go all in on it. I was one of those guys who swore I needed a physical QWERTY keyboard instead of the iPhone’s touchscreen—look where that got me? Now, I’m an Apple evangelist.

Fast forward to today, and I’m getting pretty good at prompt construction and learning how to build assistants and agents across different AI models. The challenge? The sheer speed at which new models are being released. It reminds me of when streaming services exploded to compete with Netflix. My thinking then was that very few people would subscribe to them all; instead, they’d gravitate toward the few that best aligned with their viewing preferences.

I see the same thing happening with AI models. Each has its own strengths and weaknesses, but at their core, they’re about 80% the same—they all train on public information (i.e., the internet). However, unlike streaming services, where content exclusivity was the differentiator, AI models will likely diverge based on proprietary data, integrations, and niche specializations.

A DataCamp blog comparing DeepSeek and ChatGPT put it this way:

"Each model shines in different areas. DeepSeek has shown impressive capabilities in technical tasks, particularly excelling in mathematics where it achieves a 90% accuracy rate—notably higher than many competitors. This makes it particularly valuable if you are working on technical problems. ChatGPT, however, demonstrates stronger capabilities in understanding context and providing more nuanced responses across a broader range of topics." (https://www.datacamp.com/blog/deepseek-vs-chatgpt)

The release of DeepSeek caused a stir, but from a practical (rather than political) perspective, it doesn’t feel like a world-shaking event. If data security is your main concern, you can already access DeepSeek through Perplexity, which claims to store data on U.S.-based servers—though that doesn’t eliminate potential biases within the model itself.

More importantly, AI models are in a constant feature war. The moment one model introduces something groundbreaking, competitors quickly roll out similar capabilities. For instance, Google recently launched its Deep Research model, and now other models are introducing their own research-focused tools. While these developments are exciting, they feel more like incremental improvements rather than paradigm shifts.

That’s why I find it hard to get too worked up over the release of any single "disruptive" model. Sure, each new AI might be a bit faster or slightly more accurate—but they all still have the potential to hallucinate. At some level, they start looking the same. So how do users decide which to adopt?

I think we’ll gravitate toward the models that best support our unique use cases. For me, ChatGPT is my workhorse—likely due to a first-mover advantage—supplemented by Google’s Deep Research for in-depth analysis. I’m also experimenting more with Perplexity. I like Claude, but I’m still figuring out where it fits into my workflow.

However, this landscape may not stay static.

• Proprietary data and training sources could become the true differentiators. OpenAI’s partnerships with major publishers, Google’s vast search dataset, and domain-specific models like BloombergGPT for finance or Med-PaLM for healthcare suggest that the future of AI might not be about speed or accuracy—but who has access to the most valuable private data.

• Enterprise and ecosystem integrations will drive stickiness. If Microsoft fully embeds an AI model into Office 365, or Apple develops a native AI assistant, many users will adopt those by default—regardless of whether another model is slightly “better.”

• Regulations and ethical concerns could reshape competition. AI models with questionable data sourcing or national security risks (like DeepSeek in the U.S.) may face adoption barriers, while others will gain traction due to compliance advantages.

For now, most models feel similar, and the arms race continues. But in the long run, differentiation may come less from what these models can do and more from what data they have access to, where they are embedded, and whether they align with corporate and governmental policies.

It’ll be fascinating to see how this all plays out…