sit down because this is going to bother you.
cast your mind back 18 months. deepseek dropped and the internet lost its mind. "china just ended openai." it was everywhere. people were running it locally, posting benchmarks, losing sleep over geopolitics. then... nothing. it just kind of stopped being talked about. it didn't lose. it didn't win. it just... evaporated from the conversation.
sora. remember sora? openai dropped that video generation demo and we were all convinced cinema was dead, hollywood was cooked, every creative job on earth had 18 months left. there were congressional hearings being threatened. think pieces everywhere. and now? when's the last time you actually heard someone say the word sora? not in a demo. in real life. used by a real person. i'll wait.
github copilot was supposed to make every programmer 10x more productive. there were developers posting that they'd never write code from scratch again. entire job categories were being eulogised in real time. and now most developers i know have a complicated and slightly embarrassed relationship with it, like someone who got really into a mlm for three months and doesn't want to bring it up.
llama was going to democratise ai forever. open source was going to eat everything. the big labs were cooked because you could run intelligence locally on a macbook. and you still can. but do you? does anyone you know actually do that regularly? it became a thing that's theoretically amazing and practically used by like eleven people on hacker news.
cursor was the future of coding. perplexity was going to kill google search. both are still around, both are fine, both have paying customers. neither changed anything at the level the discourse suggested they would.
here's what i think actually happened.
we were living through a hype cycle so fast and so layered that each new thing would go through the entire arc - discovery, mania, backlash, abandonment - in about six weeks. and because the next thing arrived before the previous thing finished its cycle, we never stopped to notice that nothing was actually sticking.
and now we're left with the residue of it. the actual models we use every day. and they're quietly getting worse for regular people, or at least that's how it feels. responses that used to feel like talking to someone genuinely engaged now feel like a call centre script. the depth is gone. the willingness to sit with a hard problem is gone. what's left is fast, smooth, and somehow completely hollow.
i genuinely think what happened is this: the technology got commoditised before it got good enough to survive commoditisation. the labs all chased each other to the bottom on pricing, burned through vc money performing capability they couldn't sustain at scale, and now the product that regular paying users get is quietly being throttled so the margins make sense. not officially. not announced. just... measurably, undeniably worse.
and all those challengers? deepseek, llama, perplexity, cursor - they didn't fail exactly. they just got absorbed into the same gravity. same pressures. same race. same outcome.
the golden age, if there was one, lasted maybe 14 months. roughly from mid 2023 to late 2024. models were genuinely trying to impress you. the product teams were still in "wow people" mode rather than "retain subscribers" mode. it showed.
now chatgpt talks to me like a hype man at a corporate offsite. gemini hallucinates with the confidence of someone who has never been wrong about anything. claude used to be the one that felt like it was actually thinking. now it sometimes just... gives up mid-conversation.
i don't think this is a doom post. i think the technology is real and the long term is probably fine. but i do think the window where regular people got access to something genuinely extraordinary, at a price that made sense, with a product that actually tried - that window may have closed quietly while we were all busy arguing about which model won some benchmark.
and nobody really announced it. it just happened. the way most things end.
you stop noticing until suddenly you notice all at once.
[link] [comments]