Do Not Standardize on an AI Yet.

We’re at a step-change in technology, and I have been thinking real hard with as much empathy as I can about the current state of the AI race. And I’ve come to one conclusion…

This is not the time to standardize on an AI.

I’m sure all of you are wrestling with the competing narratives around AI. From “it’s going to save us all!” to “It’s going to kill us all!” Somewhere in the middle towards “save us all” is a good spot if you can find it.

During massive technology overhauls, I tend to go back to history to read its tea leaves. For example, in the early to mid 90s, there were a couple of powerful PC companies that had amazing technology: Sun Microsystems and Silicon Graphics (SGI). Both had faster chips, better graphics (Jurassic Park, anyone?), and crushed Apple and the other PCs in performance.

But…where are they? Where did they go? 

Most of the time, tech winners don’t go to the best but to those that can scale. I don’t think anyone argues that DOS was ever better than IOS. DOS and Microsoft won because they went after the enterprise building enterprise applications. DOS won the desktop while Apple won hearts. Apple did indeed go on to become the most valuable company on Earth because they changed the form factor from PC to phone.

We’ve seen the same in search. In the late 90s when Ciceron was a wee little thing, we had to navigate all sorts of search engines, from Yahoo to Alta Vista to an upstart called Google. Google was better as a search engine, but then in 2000 they launched AdWords, making search a pay-to-play model which was predictable and performance-rooted. Advertisers loved it because SEO was (and is) unpredictable and labor-intensive, and now you could do simple economic math on what you were willing to pay for a new customer.

Who’s going to win AI? We have no idea. The Chinese are changing the model with DeepSeek because, ironically, it’s open source and requires significantly less compute, a major barrier to development for all models leading up to it. The winners were supposed to be the ones who could afford the compute, hence why Open AI and xAI are fighting for data center dominance. Meta’s Llama models are also open source. Now everyone’s shifting, because of DeepSeek, to lower compute technology, so you have chipmakers like Nvidia scrambling to create a new narrative. Oh, how the mighty drop during disruption! In the blink of an eye!

So will open source win? Will the Chinese? What about Open AI with its dominant market share? They’ve been winning because they were first and not necessarily best.

In the short term, I believe the advantage of scale — or “network effect” — will continue to be a strong part of the narrative, leaving Google Gemini and Microsoft in a strong position simply because they own the enterprise, yet very few people in the world of AI believe those two companies’ AI are the best. Adopting these models because they integrate doesn’t mean you’re getting great AI, and you need to be aware of the tradeoffs.

You have the option today to work alongside all of the models or a select few. Learn to use these models to solve real world problems for yourself or your business. Take advantage of the uncertainty by experimenting with many competing models while you can. Because who knows how long the best will be around. Early adopters not models do indeed win races!