Pinned Tweet
Iโm excited to announce an ambitious project: โจUniversal NER โจ. Multilingual NLP is missing a gold standard multilingual NER dataset, and this project aims to fill that gap.
This is a community effort: I need your help!
See: universalner.org
Or read more in the ๐งต
Stephen Mayhew retweeted
the fact that grass can become a sweater through the intermediary of a sheep ought to astonish us more than it does.
the biggest promise of AI is allowing humans to give instruction to machines -- a natural interface to a useful tool. I like to call this "automation in the Information Age"
These products do the inverse: giving machines an interface to interact with humans.
After Google acquired my AR startup, I started thinking about the next interface, this time for AI. Text and voice work fine, just like DOS did before the Mac. But nothingโs more intuitive than face to face.
After two years of research, today weโre launching Atmee v2: a real-time avatar model that makes live video agents feel truly life-like.
โข Any character from 1 image + 10s of audio
โข Contextual expressions, gestures, even interacting with the environment
โข Live video calls, voice only, or video generation
At its core, Atmee v2 is a video foundation model, it can render anything, but itโs built to excel at characters, which is what video models like Seedance 2.5 and Wan 3.0 are mostly used for. In comparison:
Atmee v2
โข Up to 50x faster
โข 10-100x cheaper
Build tutors, sales coaches, support agents, companions, game characters, even whole living worlds. The interface after text and voice starts today.
> listened to a track (my own recording) on Soundcloud
> a new track comes on
> pleasant, soft, piano jazz
> I like this, what is it?
> it's AI
> RIP music recommendations, was fun while it lasted
Stephen Mayhew retweeted
i've reproduced the results from the MIT blog post today on weight-space programming. but one additional result i have to add: this is a new, step-change emerging capability that's appeared in just the latest set of models.
Stephen Mayhew retweeted
the class of things we (i mean i) consider as "meh, of course an AI can do *that* " these days is kinda insane tbh.
there's no way this can be real
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! ๐งต
Blog: rikiyat.github.io/blog/weighโฆ
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-spโฆ
github.com/RikiyaT/lm_undโฆ
huggingface.co/Qwen/Qwen3.5-0โฆ
huggingface.co/RikiyaT/Qwen3.โฆ
Stephen Mayhew retweeted
I'm on the industry job market this year!
I'm a PhD student at CMU LTI. My work spans post-training, evals, alignment and multimodality!
Please reach out (DM) if you are hiring and think I'd be a good fit, I'd love to chat :)
Website: simran-khanuja.github.io/
CV: simran-khanuja.github.io/assโฆ
Stephen Mayhew retweeted
I love that scene in Good Will Hunting where Ben Affleck is talking to Matt Damon's titular character Will Hunting and he's like โyou've got a gift man and you've gotta get outta this town and make something of yourselfโ and the gift in question is being really good at math and the town he needs to get out of is Boston, Massachusetts.
Stephen Mayhew retweeted
There's an old joke about a scientist who tells God the creation of the world wasn't any great shakes. He could do just as well himself. Whereupon he picks up a handful of dirt. "Not so fast," God says. "Get your own dirt."
Claude, get your own dirt. /End
my new measure of techno-modernity is: how many battery-powered devices do I interact with each day?
yesterday: at least 9 (scale, phone, laptop, smartwatch, air-pods, over-ear headphones, metronome, lawnmower, iPad)
Stephen Mayhew retweeted
Oh, my, @nytimes .... again?!? ๐๐๐
























