@gradientpull

redacted

Joined April 2025
calling it now, openai demolishes typesafe in < 1 month with their own general purpose classifier @ 10x cheaper cached inputs
3
42
this dude must be going fucking insane rn (2022)
7
208
the people yearn for BERT-as-a-Service
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x cheaper (w/ output tokens free) • Frontier composable intelligence optimized for decisions AFAICT the shortest path to AI-based economic revolution
1
4
90
free GPT-6 for mathematicians -> (train on dat shit 😜) -> solve navier stokes … free GPT-6 for clinicians -> (train on dat shit 🤪) -> solve cancer?
GPT-6 Astra Pro for clinicians:
5
80
forcing electrons to learn math in the Taiwanese Sand Mazes is destined to have consequences of mythological proportions
2
10
518
BunkerBench updated!
5
60
the rebirth of the {noble_adjective} {industrial_noun} naming schema was a fun gimmick but cant wait to cycle back into software startups getting the names they deserve, like Gloox or BoingBoing. Probably next year
1
1
11
583
ai psychosis got me queuing steering messages like hikaru premoves blitz chess
2
70
cousin from another uncle
1
82
finally tried LeChat. The pretraining mix in european models really does seem different. nitter.cf/mistralai/status/20846…
🛡️Introducing Shieldstral, Mistral’s 3B open-weights model for content safety that can be deployed on-device 🧵 mistral.ai/news/shieldstral
3
158
cant wait for codex to deploy this bad boy into the physical world with some retarded /goal nitter.cf/1x_tech/status/2075252…
NEO’s Hands An API to the Physical World
129
smokin in da chinese squat to increase iq based on priors
1
80
if i died today id be very undercooked, straight back to samsara.
1
71
completely brain dead, genuinely scary to see
71
“just drank a pound of milk” goes MUCH harder than volumetric equivalent
29
elons posts have such a high coal-to-gem ratio, a very bimodal distribution, ironic for the guy who popularized the idiot index
25
gabor retweeted
it does matter. signals lack of effort and care. id rather read her raw half cooked ideas than claudes lexical slop basin overhaul. AI is a wonderful tool but not particularly great at writing, and its offputting to digest a dozen “its not this, its this” patterns w/o a warning
1
1
87
awful day to be a glimpse into the divine, hiding in plain sight
1
1
70
RLHF really milking the new "You're absolutely right!"
2
112