OpenAI has a lot of key challenges that makes this a do or die year for the company.

OpenAI has a lot of key challenges that makes this a do or die year for the company.

1. Their AI models are increasingly undifferentiated as too many labs have caught up.

2. Even with raising $100B they are still being outspent by most of FAANG and can’t keep up that fundraising pace annually.

3. They’ve spread themselves too thin by going after devices, web browsers, ad platforms and short form video at once. They don’t have the talent density or competitive advantage to do all these well.

OpenAI has a lot of key challenges that makes this a do or die year for the company.

1. Their AI models are increasingly undifferentiated as too many labs have caught up.

2. Even with raising $100B they are still being outspent by most of FAANG and can’t keep up that fundraising pace annually.

3. They’ve spread themselves too thin by going after devices, web browsers, ad platforms and short form video at once. They don’t have the talent density or competitive advantage to do all these well.

@carnage4life

I think your missing a subtle hidden challenge.

Companies start to realize that smaller more narrowly focused models with training data tailored to their problem are actually better solutions. Over humongous solve everything models.

Maybe contrarian here but I think this was always the biggest risk. Fundamentally if we look at four decades of AI solutions in production, they are all narrowly focused and this was not a accident.

The whole go LARGE and try to solve everything was always kind of crazy if you stopped and thought it through.

@carnage4life

I think your missing a subtle hidden challenge.

Companies start to realize that smaller more narrowly focused models with training data tailored to their problem are actually better solutions. Over humongous solve everything models.

Maybe contrarian here but I think this was always the biggest risk. Fundamentally if we look at four decades of AI solutions in production, they are all narrowly focused and this was not a accident.

The whole go LARGE and try to solve everything was always kind of crazy if you stopped and thought it through.

@shafik @carnage4life it's classic "crossing the chasm" by Geoffrey Moore.

Identify the product
Start with early adopters
Hit the "chasm" - the gap to mass adoption
Pursue verticals - smaller focused markets that provide cash and allow you to scale

Anthropic - focused on coding vertical - well regarded for their Claude Code solution

OpenAI - trying to go direct to consumer with no value proposition, again and again and again.

And flailing.

I thought Silicon Valley learned about this concept in the 2000's. I guess this generation forgot.

Either that, or OpenAI forgot what problem LLM's actually solve, and swallowed their own hype.

https://share.google/gLDQjOuFbVHQkoX7S

@leanlearnlead @shafik @carnage4life

Call back to excellent book. How many in the field have read it? Understood it?

OpenAI has a lot of key challenges that makes this a do or die year for the company.

1. Their AI models are increasingly undifferentiated as too many labs have caught up.

2. Even with raising $100B they are still being outspent by most of FAANG and can’t keep up that fundraising pace annually.

3. They’ve spread themselves too thin by going after devices, web browsers, ad platforms and short form video at once. They don’t have the talent density or competitive advantage to do all these well.

@mlevison @shafik @carnage4life when I worked in tech it was required reading. We built our product management strategies on it. Built a $19B brand at its peak.

Moore's approach isn't a panacea - he doesn't offer much once you get to mass market, and his approach doesn't work for mass consumer brands once mature. But when starting up, his insights are gold.

@leanlearnlead I hear you.

I bought two copies over the years. I first read it in the 90s. Loaned a copy to a coworker and had to repurchase.

I’ve been at two companies that failed to cross the chasm.