Remix.run Logo
▲ chubot a day ago

It is pretty wild. Did the revenue explosion really start approximately in February 2025 with Claude Code?

AI learned how to use Unix shell? We don't need MCP?

All types of software companies, small medium big, just loved it for writing software? And just shoveled money into Anthropic?

Is it really as simple as that? I may be out of the loop.

(FWIW I like Claude and use it every day, but I use Claude Code just a little bit. I actually like AI more for non-coding things at the moment, rather than coding)

---

Informally I believe that (according to the market) Anthropic must be the biggest hit company since Google. Google started in 1998, and IPO'd in 2004.

Even Meta, Uber, and all the rest took longer than that.

▲chubot a day ago | parent | next [-]

Meta: I pasted my comment into Claude, Gemini, and DeepSeek. And actually all 3 models pointed out what should have been obvious: Google was wildly profitable at IPO, while Anthropic is not!

But yeah that's why I said "according to the market", which is a big caveat.

(Gemini was the most sycophantic, as usual. It kinda restates everything I say in flattering terms. Claude and DeepSeek pushed back on the shell vs MCP thing -- it's not either-or. But Claude agreed Unix shell was a big surprise.)

▲segfaultex a day ago | parent | prev [-]

Claude Code is exploding because it's incredibly subsidized IMO. ANT is selling $3 bills for $1.

A fully-utilized Claude Pro plan ($20/mo) equates to ~$400/mo in API value.

I could wind up being very wrong, but I don't see a net revenue increase if ANT cranks up the price. IE: I think they'll lose more in value than they gain from the cost increase to the remaining users.

▲orangecat a day ago | parent [-]

ANT is selling $3 bills for $1.

No. You may be able to get more than $20 of tokens from your individual subscription, but the large majority of their revenue comes from enterprises paying API rates, which are highly profitable.

▲gonzalohm a day ago | parent | next [-]

And how long until mediocre models that are good enough can run properly on cheap hardware? Then you would need to be stupid to pay Anthropic for a slightly better AI model

▲segfaultex 12 hours ago | parent | prev [-]

Yes. Enterprise token spend is falling as companies are reigning in budgets. Even with the margin on enterprise inference, they’re bleeding money.