Comment by ijidak
Yeah the interesting question regarding Anthropic is: do they build a software moat around their models?Most of us, even myself, focus on the reality that models are likely to become a commodity and that means the big providers will be unable to maintain profitability off of their weights alone (ignoring whether or not they serve the infrastructure).
But what I have to give Anthropic credit for is that they bring to market products that work well enough incredibly fast and in that sense they remind me of Microsoft in the early '90s.
I use both Anthropic and OpenAI products and OpenAI products are so bad. Just incredibly clunky changing in strange and unnecessary ways. While problems that they should have fixed months ago, years ago, are still unfixed. No matter how fast you can build, you still have to build the right things.
It reminds me of a quote from Napoleon, which may be apocryphal and that I will likely butcher.
It was something along the lines of how Napoleon staffs his military. You have people that are smart and lazy, people that are smart and hardworking. You have people that are dumb and lazy and you have people that are dumb and hardworking.
He makes the lazy-and-smart generals, because they come up with good strategy but they're not excessively aggressive.
He makes the dumb and hardworking captains because they'll follow orders and lead their men.
He makes the dumb and lazy foot soldiers because they follow orders and they can be driven.
But the dumb and hardworking, he shoots because they both don't know what they're doing and they do a lot of it.
And I kind of feel that way regarding AI. AI makes everyone "hardworking." Now, with AI, everyone can get a lot of stuff done. But when it comes to product, if you're "dumb" regarding what to get done, you just do a lot of damage instead of producing results.
And between OpenAI, Google, and Anthropic... and I guess Meta.. Anthropic seems to be the only one of the big model providers that is both smart and "hardworking" (let's replace "hardworking" with "productive.").
Anthropic has a higher batting average on a lot of their productization ideas and does so quickly.
And ultimately that might be Anthropic's moat -- just as Apple's moat in smartphones was not that no one else could make a similar smartphone, but that no one else executed as well end-to-end for the customer during the Jobs era.
Even Anthropic's ability to get to an IPO before OpenAI is another example of them just doing things better.
It could be that this generation's Netscape is OpenAI and ChatGPT. While Anthropic ends up being this generation's early -90's Microsoft.