Today’s frontier LLM companies have the same product strategy and sell a slightly different encapsulation of same capabilities with switching costs still very low. They all seem to be converging on long context windows, tool use & APIs, chain-of-reasoning, summarizing, coding, translations and more with subtle differences.
What that means is that differentiation comes down to one thing and one thing only — marketing.
There’s no better way to express this than to point to the iconic scene from the Mad Men pilot when adman Don Draper pitches Lucky Strike’s owners on the ‘greatest advertising opportunity since the invention of cereal’:
[Don Draper]: We have six identical companies making six identical products, we can say anything we want. How do you make your cigarettes? [a clueless Lee Garner, Jr lol]: I don’t know. [Lee Garner, Sr]: Shame on you. We breed insect repellant, tobacco seeds, plant ‘em in the North Carolina sunshine, grow it, cut it, cure it, toast it. [Don Draper, as he writes “IT’S TOASTED” on the chalk board]: There you go. There you go. [Lee Garner, Jr]: But everybody else’s tobacco is toasted. [Don Draper]: No, everybody else’s tobacco is poisonous. Lucky Strike’s is toasted.
The whole 5-min scene from Mad Men S01E01 is worth watching, and for me, will always be one of television’s all-time greatest moments.
Veteran tech analyst Benedict Evans drives this point home in the latest episode of Another Podcast.
And there’s this sort of thing here of like, you have this tension between the general purpose model, around the general, how do you differentiate? […] What would it even mean to have different product strategies for something where the whole concept is it does everything? How do you have different kinds of everything?
If you’re into this stuff, I highly urge you to listen to the whole thing — it’s quite an insightful 30m discussion.
Sure, CEOs of OpenAI and Anthropic can ask their chief scientists to go “build a better model. here’s a billion more dollars” (of course, all in lowercase). And every few weeks or so, the newest Gemini or DeepSeek will surpass say, Llama or what have you, by some % on some aspect like accuracy or performance to make it to the top of the LLM leaderboards, and it’ll make for sensational “new leader in AI arms race” headlines.
But, consumers don’t care about that, and they never have. And Evans sheds light on the consumer perception of models with an interesting analogy to consumers being asked to compare mortgage rates from two banks:
But what you have as the model is completely undifferentiated, except at the level, as I said, of what color is the icon. It’s kind of challenging, provocative thing to say, perhaps. But yes, I’m sure if you use all of these things every day, you would say, oh, but you know, Anthropic is much better at code and ChatGPT is much better at multi-stage reasoning. Well, yes, but give this to a random person on the street. So you want to try and compare two mortgages, which one would you use? How would you know?
Precisely.
But, what consumers do care about is exciting new features. And that’s not coming from the model companies so much as it is from the products wrapping them by paying for API tokens.
The one large tech company whose moves people were anticipating here was naturally Apple. They had the right idea with Apple Intelligence but they overpromised and vastly underdelivered. And as John Gruber put it, it might take a while for them to recover. Tim Cook’s got that, EU regulations and now, the whole tariff crisis to deal with. So, I don’t really have big expectations from Apple around this for a couple of years.
OpenAI? Maybe. Sam Altman on his interview with Ben Thompson of Stratechery a few weeks ago hinted that ChatGPT is only their first product and what they’re looking to build is a portfolio of products reaching a billion DAUs similar to what Meta has done with their properties. But it’s anyone’s guess what the next couple of products could be.
It’s going to take a while to get answers to basic product questions like what does it mean to have websites when LLMs are the base layer? What AI product replaces the browser? And lots of questions like that.
But, until then, marketing is the name of the game. If everyone’s got the same thing to offer, you can say anything. So you might as well say, “it’s toasted”.