Post

for the haters - bad news, i am afraid. literally every Fortune 500 company is using AI to develop their products and services. in many cases they're no longer even doing manual code review. i hate to break it to you, but if you won't use services built using AI you just checked out of modern life.

17 reposts100 likes

enterprise customers have been paying full price for a while now, and it seems to be going ok. and the ceiling keeps rising while the floor keeps falling; glm-5.3 flash (320B-A16B) scores about as well as opus 4.8 did in may, at a fraction of the price. the debt thing is uhhhh 😳😳😳

At current pricing Anthropic is making money on their inference, with excellent margins. My work enterprise plan works at API pricing levels, no discount, and it's $25/month. It's the most profitable way Anthropic sells Claude and everybody's making money. Seems sustainable to me.

Except they need to keep making their models better to avoid being caught by cheaper open source rivals, which means they *have to keep paying a fortune to train frontier models* so the margin on inference isn't particularly relevant.

1,000% agree. Private equity DD now often includes building out the product using AI as a way to test the defensive moat. Anyone saying they won’t use any product that uses it is delusional and far removed from the industry.

IMO, scolding especially free / open source software developers for daring to touch AI is actively harmful. Microsoft is showing us how not to use or integrate AI (with Windows/Office) and now as the communities we have a stronger fighting chance with making/improving our own software alternatives.

Individual developers who are visible and honest about their AI use get the most vitriol, even by people on a social media using AI, on their browser using AI, on their OS using AI. Why develop for a community that wants you to conform or risk being thrown aside thanklessly (and with harassment)?

(Open source alternatives existed before, yes, and I've been using them for a decade. Being realistic, not everything works, and the learning curve puts off many non-technical people. With LLMs, I've fixed wine bugs and ported windows tools that didn't run under wine, now I never boot my windows VM)

Peronally I wouldn’t be half as irritated if I didn’t expect most of AI use to be done because every company expects it to be a wonderful accountability sink when things will go wrong.

Yeah Capitalism has many flaws. "It was the AI that did it" will be a thing for sure. We have a whole mess of legal and regulatory stuff to work out here.

i mean I appreciate the desire to do that. Late stage capitalism fucking sucks, in many many ways. The theft of IP is gross, the environmental arguments against AI are quite reasonable. We need jobs to live. But just saying I won't ever use anything that touched AI doesn't seem realistic.

i won't be an apologist for the excesses of the AI bro revolution, but I am a pragmatist and a realist. We have to deal with the reality of the situation as it is, not just how we'd like it to be. It's my job to track software and I fucking love software developers and developer culture.

Even for the folks who want to shut it all down and put the genie back in the bottle, I’m not sure they’re fluent in the current state of the industry. Just having a clear-eyed look at where things stand is pretty clarifying. I’ve seen if shift people into harm reduction vs abolition.

it's got to be reduction. we need regulation. we need to think carefully about the social safety net, and the role of creativity and creatives as social glue.

your pain and feelings shouldn't be discounted. many devs are feeling grief right now, for lost craft and agency, while others are high as kites on LLM dopamine and stock options. but yeah, I don't want to leave the tech future to the fascists, so I feel like we kind of need to engage.

I hate AI slop. I hate AI art. I hate AI music. Writing is definitely best left to humans. But in the field of software development, we've hit a revolutionary tipping point. 2026 is the year when the models became good enough to solve hard problems autonomously. this is the new reality.