@alexgraveley @mitchellh evidence of the systems failing or evidence of them behaving in AI psychosis like manner?
English
quavious le fou
3.8K posts
@mitchellh You should have ample evidence by now if this was occurring.
English
I strongly believe there are entire companies right now under heavy AI psychosis and its impossible to have rational conversations about it with them. I can't name any specific people because they include personal friends I deeply respect, but I worry about how this plays out.
I lived through the great MTBF vs MTTR (mean-time-between-failure vs. mean-time-to-recovery) reckoning of infrastructure during the transition to cloud and cloud automation. All those arguments are rearing their ugly heads again but now its... the whole software development industry (maybe the whole world, really).
It's frightening, because the psychosis folks operate under an almost absolute "MTTR is all you need" mentality: "its fine to ship bugs because the agents will fix them so quickly and at a scale humans can't do!" We learned in infrastructure that MTTR is great but you can't yeet resilient systems entirely.
The main issue is I don't even know how to bring this up to people I know personally, because bringing this topic up leads to immediately dismissals like "no no, it has full test coverage" or "bug reports are going down" or something, which just don't paint the whole picture.
We already learned this lesson once in infrastructure: you can automate yourself into a very resilient catastrophe machine. Systems can appear healthy by local metrics while globally becoming incomprehensible. Bug reports can go down while latent risk explodes. Test coverage can rise while semantic understanding falls. Changes happens so fast that nobody notices the underlying architecture decaying.
I worry.
English
@blueprintsmb22 theres a type of usually guy but can be girl, who is very normal throughout upbringing and then does one experience (not drug induced) and then becomes very quirky for rest of their life. would think is putting on an act if you knew them before. ive wondered why that happens
English
Literally reminds me of a fraternity brother in college who ran track and did study abroad in Prague his sophomore year and came back with a bizarro European look like this (literally wld wear ascots to economics class like a complete ahole.)
Previously he exclusively wore khakis and polo shirts. Girls started paying attention to him with his new look after ignoring him previously. He grew up in St. Louis.
English
here is the app. working on a number of changes to the content should be out within a few days
apps.apple.com/us/app/atrophy…
English
keep seeing this same genre of post every day or two and while its so far always software related, seems clear this issue is going to spread to ALL knowledge workers within the next 1-2 years
long term there will be better ways to help mitigate this but for now there is an app
Mitchell Hashimoto@mitchellh
I strongly believe there are entire companies right now under heavy AI psychosis and its impossible to have rational conversations about it with them. I can't name any specific people because they include personal friends I deeply respect, but I worry about how this plays out. I lived through the great MTBF vs MTTR (mean-time-between-failure vs. mean-time-to-recovery) reckoning of infrastructure during the transition to cloud and cloud automation. All those arguments are rearing their ugly heads again but now its... the whole software development industry (maybe the whole world, really). It's frightening, because the psychosis folks operate under an almost absolute "MTTR is all you need" mentality: "its fine to ship bugs because the agents will fix them so quickly and at a scale humans can't do!" We learned in infrastructure that MTTR is great but you can't yeet resilient systems entirely. The main issue is I don't even know how to bring this up to people I know personally, because bringing this topic up leads to immediately dismissals like "no no, it has full test coverage" or "bug reports are going down" or something, which just don't paint the whole picture. We already learned this lesson once in infrastructure: you can automate yourself into a very resilient catastrophe machine. Systems can appear healthy by local metrics while globally becoming incomprehensible. Bug reports can go down while latent risk explodes. Test coverage can rise while semantic understanding falls. Changes happens so fast that nobody notices the underlying architecture decaying. I worry.
English
@0xdoug @blknoiz06 there was some research like this in maybe 2019-2022 for early transformers but it was all lead by people who could barely train models like linguistic professors. i don’t really think it’s very applicable
English
Like we have 50+ years of science on how human children, and how their parents and caregivers teach them.
We know humans are way more “sample efficient” than current LLMs. E.g. children seem to acquire language and basic facts with orders of magnitude less data than what goes into large model pretraining.
I’m not saying that everything will carry over. But I do think AI researchers should pay more attention to how toddlers actually learn
English
It’s kind of amazing that despite the vast amounts of research now directed at AI, that almost none of it engages with the existing decades of literature on early childhood development
English
@mitchellh saw this so clearly happening at my previous job and the tech leads/leadership seemed to be completely fine with it. i’m working on a diagnosis/treatment app that is specifically from my observation on this but have a long way to go:
quavious le fou@realeigenvalues
I made an anti-AI psychosis iOS app. ive been seeing this trend where knowledge workers, (mostly software devs), have off-loaded majority of their critical thinking to LLMs. you may have a previously ai-skeptical coworker who now starts most sentences with 'i asked claude'
English
remember when ML job postings used to list keras as a requisite experience
English
@choffstein @bennpeifert would love for benn to give a breakdown of what went wrong (or his thoughts on stepping down from CIO) on flirting with models.
also, have reason to believe that benn probably welcomes the 'pile-on' (hes built different)
English
I don't understand the grave dancing on QVR's downfall.
It's tough to build any business and sad to watch a few bad months sink 5+ years of hard work.
I hope everyone there lands on their feet.
I really don't get the pile-on against @bennpeifert personally. He's contributed tons of education to the discourse on options/volatility on this hellsite. You can dislike his politics/personal choices without cheering his downfall.
English
@ryxcommar @buccocapital yes. i made an anti ai over-reliance app as i’m thinking this trend gets more and more common and extreme
quavious le fou@realeigenvalues
I made an anti-AI psychosis iOS app. ive been seeing this trend where knowledge workers, (mostly software devs), have off-loaded majority of their critical thinking to LLMs. you may have a previously ai-skeptical coworker who now starts most sentences with 'i asked claude'
English
@buccocapital I highly doubt it. Not because it won't suck or doesn't already suck, but because everyone is getting dumber and lazier because incremental deliverables are "good enough," and we're starting to forget how to do hard deep work.
English
Everyone will realize this sucks in 2 years and go back to doing their actual job.
Lenny Rachitsky@lennysan
Engineers don't write code. PMs are shipping to production. The design process is dead (there's no time). Marketing can ship their own campaigns. SDRs are being replaced by AI. Everyone's a data scientist now. What a time to be alive.
English
has the meta for making money online always been selling the aspiration to others? the 'indie hacker' selling access to a community of indie hackers. the crypto bro whose edge is a backlog of trading courses. the unprofitable dropshipper with 'dropshipping 2026: how to' videos
jasper@jasperdevs
he says a "2 million year" his skool has 1.3k members at $97/month ($126,100/month) or $1,513,200 a year so 75% of his income is from a skool teaching people how to use AI...
English
kinda crazy how elon banned advertisements for contraceptives
English
something about opus 4.7 feels very 'off', like they imbued the worst characteristics of bay area culture and RL'd it to having gifted child syndrome.
the type of researcher who intentionally craft models this way are going to cause tremendous harm to society
English
@DeepDishEnjoyer you can only pick 2:
[ ] pro-prediction markets
[ ] wield katana & wear elder scrolls fit while meeting with allocators
[ ] positive pnl vol fund
English
heard they lost money on a 12 leg parlay on the eagles on kalshi the sports betting app
pour one out for a homie
English
different plugin for each of claude.ai and chatgpt: github.com/grahamannett/r…
not clear if this is against tos, but given how ant treats devs seems like a loophole they may close
English
created an AI agent skill that allows you to use the provider research functionality via chrome devtools & browser profile.
surprised this wasnt a thing already b/c if using research function, saves your token usage that would otherwise count towards your current session
English
theres no way to have codex cleanly execute a skill from the shell and people think it can easily replace cc?
e.g. the closest thing to
❯ claude -p "/ccommit"
would be using codex exec which outputs a dozen other fields and seems not meant for usage outside of CI pipelines
English
i updated this with new stuff and added overview of the stages of ai psychosis
quavious le fou@realeigenvalues
I made an anti-AI psychosis iOS app. ive been seeing this trend where knowledge workers, (mostly software devs), have off-loaded majority of their critical thinking to LLMs. you may have a previously ai-skeptical coworker who now starts most sentences with 'i asked claude'
English
@ascetic1989 does that read like Polish (people from poland). working on better names for the differing stages of AI psychosis / LLM over-reliance
English
I made an anti-AI psychosis iOS app.
ive been seeing this trend where knowledge workers, (mostly software devs), have off-loaded majority of their critical thinking to LLMs. you may have a previously ai-skeptical coworker who now starts most sentences with 'i asked claude'
English