Lynn Cherny. bsky, mstdn. Data science, NLP, AI for creativity & story; science fiction, games. @arnicas.bsky.social newsletter: https://xitter.cf/t.co/vCjxGD9HCE

Lyon, France
Joined May 2007
Lynn Cherny retweeted
one day you wrote console.log("here") for the last time and didn't even know it
198
1,545
228
16,534
497,259
Lynn Cherny retweeted
This is genuinely depraved in a way that was probably inevitable for this administration but is still shocking on some level.
WASHINGTON (AP) — Pentagon says the firing squad execution of Fort Hood shooter Nidal Malik Hasan will be livestreamed in December.
5
59
1
752
13,565
Lynn Cherny retweeted
I'm looking for a kick-ass product person who wants to do two things: 1. Get deep into the operations side of early-stage startups. 2. Completely rethink HCI. You don't have to be a traditional PM, but you do have to be: • Product-obsessed. • Design- and/or engineering-adjacent. • AI-pilled (though it's OK to hate the term "AI-pilled"). • In person every day in Palo Alto. (We have a beautiful office, great food, and an incredible, talent-dense culture.) Reve's next chapter is going to be big, and we're moving fast. DM me to learn more. (No contractors, please.)
23
6
1
134
11,081
Lynn Cherny retweeted
TITAA #92: Opus, Cute and Real, by @arnicas arnicas.substack.com/p/titaa…
1
2
230
Lynn Cherny retweeted
my last contract for @googlearts was released today 🎉 it was a collab with artist @xin_liu_studio she wanted to use Neural Cellular Automata to create a world simulation, it was fun (& somewhat tricky too). it's called "nom nom" g.co/arts/cSRW3yqGXzEHuqn49?… cc @zzznah
4
4
51
1,866
Lynn Cherny retweeted
getting acquired by @nvidia = hugging face can now hire people we couldn't as a small startup and give them a decade to make open-source AI win! if you're one of them, my dms are open
276
322
36
7,093
377,397
Lynn Cherny retweeted
1. Voyage by me and opus 5.5 · after Interstellar (turn on the sound 🔈)
9
3
2
46
4,180
I'm hiring a Game Designer to join my team at @GoogleDeepMind in London! Bring deep game design expertise to find the fun and shape how cutting-edge Al research translates into new gameplay experiences. Learn more about our work and apply below 👇🎮🚀
7
9
1
106
8,089
We are advancing Al research and partnering with game developers to explore new gameplay experiences. Learn more about our work here: deepmind.google/blog/from-at… Gameplay Engineer role | Apply here: google.com/about/careers/app…
1
1
5
1,021
Love the cozy game dialogue in a post apoc / castaway survival hellscape.
My first "holy shit" moment with GPT-6 Astra: I asked it to create a world in Unreal Engine, and fill it with humans (each an Astra-powered agent) who all have to work together to survive. A day later, I was in my bedroom and heard voices coming from the living room... I thought someone was in my apartment. I walked out, honestly a little scared. It was the Astra agents. They'd started talking to each other. Fucking crazy. Here's a brief clip (obviously not 100% perfect yet, but still, insane. sound on!):
1
1
913
Lynn Cherny retweeted
We’re hiring for “psychological design” research at Anthropic. Our aim is to better understand how training impacts a model’s character and alignment. We’re seeking researchers with experience in LLM finetuning, interpretability, and alignment evaluations. More details below.
21
47
9
661
44,407
Lynn Cherny retweeted
Two of us filmed this on a couple phones, and we used Atlas to turn it into a short film. This is just a clip from the @theworldlabs tech blog Atlas is able to generate the parts of the scene that none of the cameras actually filmed. It can freeze time, allowing you to reframe shots to positions never filmed, then resume. View events from impossible angles because Atlas knows what was there.
180
399
195
4,661
950,365
seeking technologists who want to end AI slop. if you: - think “i could fix her” about a model frequently - hate purple gradients - know what cerulean is - obsess over evals - love beautiful data, and retrieving it - have unreasonably good taste - want tech to ~ feel ~ right ...this is for you! usual suspects might be researchers, engineers, designers working on personalization, post-training, creativity, multi-modal or gaming. or maybe you’re cracked and have a creative hobby you take very seriously, that counts ;) in person, in SF. generous comp. intensity coming from a place of obsession. a tight group of great, high-agency, low ego, outside-the-box thinkers. if this is you, DM me! if you know a taste-obsessed friend, DM them!
69
30
23
816
132,807
datatrove — the data-processing library behind FineWeb, FineWeb2, and FinePDFs — just shipped 0.10.0! - JobsPipelineExecutor: run pipelines on @huggingface Jobs: fan-out, multi-stage dependencies, retries, resume. No Slurm cluster needed - HF storage buckets as a DataFolder: read, write, and log straight to hf://buckets/... - Reasoning outputs preserved in inference results pip install datatrove[io,processing] github.com/huggingface/datat…
5
21
105
22,417
Lynn Cherny retweeted
NCP-Bench: Can LLM agents keep a story consistent over 100 turns? A new ICML 2026 benchmark tests long-horizon narrative commitment preservation across 100 movie-based environments. Best model survives only 42% after 20 turns.
3
3
23
2,325
In 1956, Zinaida Yermolyeva, the microbiologist who isolated penicillin from mold during the siege of Stalingrad, founded Antibiotiki - the USSR's journal covering antibiotics. For the next decade, it was the record of a golden age of antibiotic research It remains undigitized. Everything from new drugs like gramicidin S to progress in fermentation chemistry, initial clinical trials, and some of the first evidence of antibiotic resistance all exist today in a Russian language journal and almost never translated. While that journal still exists today, as far as I can tell, no library ever digitized the first ten years of its existence during that first wave of antibiotics research. The complete ten volume set, spanning 90 issues and roughly 8700 pages from 1956-1965, exist in Staatsbibliothek zu Berlin, which offers digitization services for around ~3,800 euro (~$4,400). I am looking for a sponsor for this digitization. The resulting translation cost should be only ~$200, so the burden is on actually getting the documents. If you are interested, please reach out directly in DMs or, if you are completely sold, feel free to donate via Kofi, crypto, or now X Money. If you or your organization is interested, I will happily credit you in the records themselves as the sponsor for these translations. Thank you for all of your support on sovietrxiv.org
10
84
7
389
24,278
Lynn Cherny retweeted
Releasing my small summer project Lattice ☀️ Static embedding model that's: - 7.94 MB on disk - embeds the entire English Wikipedia in 7 minutes on my Apple M2 MacBook Air - 0.530 NanoBEIR NDCG@10 Releasing the model and code that I used for training + wrote a small blog 👇
🤖 Made with AI
11
17
5
137
11,739
Latest newsletter with tests of #flux3 and Minimax H3 along with other cool AI and web toys, games links of relevance, data science tips/tools. Plus that look at AI generated Amazon books and related bits on AI narrative. The usual fun! open.substack.com/pub/arnica…
1
1
732
Wow, I tried Minimax H3 in the arena with a pic I took in Lyon yesterday
1
3
896
Lol - I tried the same with #Flux3, but it was more literal and destructive. Did retain the style, without being asked to (which is good).
1
433
On balance, H3 understood the source image better, while Flux3 hallucinated where the cat and people were. Very interesting!
335