Eric Litman

6.2K posts

Eric Litman banner
Eric Litman

Eric Litman

@ericlitman

Founder @healthspanners, @aescape Avid health optimizer Serial optimist

New York City Katılım Nisan 2007
963 Takip Edilen4.2K Takipçiler
Sabitlenmiş Tweet
Eric Litman
Eric Litman@ericlitman·
Speechless.
Eric Litman tweet media
English
7
4
47
4.4K
Philipp Berner
Philipp Berner@philippberner·
@ericlitman That’s how we will communicate with the aliens 👽 once they arrive
English
1
0
1
22
Eric Litman retweetledi
Julia Belk
Julia Belk@juliabelk·
Gaining therapeutic access to the human brain is one of the biggest unsolved problems in biomedical science. Today @nature, we uncover a massive influx of immune cells into the human brain during aging, revealing that the brain is more accessible than previously thought. 1/ nature.com/articles/s4158…
English
10
92
344
41.8K
Trevin Chow
Trevin Chow@trevin·
Good discussion here about how Sol is over blowing scope. I've seen this myself, especially in last few days like @CalvinGrunewald. However, like @kunchenguid I've seen this on every reasoning level not just Sol Medium.
Calvin Grunewald@CalvinGrunewald

Based on @thsottiaux's advice, I use GPT 5.6 Sol Medium as my daily driver. Over the past couple of days, I've noticed a pattern where the model very eagerly increases scope of whatever prompt I give it. Maybe it's the set of tasks I'm working on now, or something else.

English
1
0
5
1.5K
Jordan Hall
Jordan Hall@jordanhall·
Been working on a fun side project - a countertop AI voice assistant to replace the HomePod in my kitchen. Combined an old iPad mini, a speaker/mic bar, and a 3D printed enclosure. Runs @OpenAI realtime audio with wake word detection, multi-user voice auth, and can delegate to other agents for tasks. Built the agent, server, and GUI almost entirely with Claude Code. It doesn't run @openclaw yet, but could maybe? Let me know if this is interesting to anyone and will share more. My kids love it so far!
Jordan Hall tweet media
English
5
0
8
362
Escha Labs
Escha Labs@Eschalabs·
@eplurubusnullus Thanks! Yes we are missing a lot of support, so far we are able to get a decent kernel out for ZML and SGLang, will work out the others too, hopefully soon!
English
1
0
3
226
Escha Labs
Escha Labs@Eschalabs·
Today we are introducing Escha-W2 quantization. A 2-bit Qwen3.6-35B-A3B model built for fast, local inference. The complete model is 12.3GB on disk—small, enough to run on a single consumer GPU — while averaging ~100% of FP8 performance across 12 benchmarks, including: MMLU-Pro: 80.9 MATH-500: 93.8 GPQA-Diamond: 77.8 LiveCodeBench v6: 62.6 BFCL tool use: 88.9 RULER 8K–128K: 89.9 Commonsense-6: 76.1 On a single RTX 4090, the model runs: 225 tok/s single-stream generation on 12.3GB on-disk model size Compressing a 35B MoE model this far without collapsing its capabilities required more than a standard quantization pass. We built an end-to-end compression system combining state-of-the-art low-bit quantization with model-aware fine tuning and recovery to preserve capabilities most vulnerable to low-bit error. Quantizing the Qwen 3.6 35B model - from the base model to the final deployable checkpoint — took approximately 10 hours to complete. Escha-W2 runs through a custom Qwen3-MoE runtime, which includes the weight loader, low-bit decoding kernels and serving integration required to execute the Escha format efficiently. The runtime currently supports SGLang and ZML deployment. vLLM and llama.cpp will be supported in the next release. No retraining from scratch. No specialized accelerator. One consumer GPU. Model download: huggingface.co/EschaLabs/Qwen… Runtime download: huggingface.co/EschaLabs/esch… Apache-2.0 model and runtime.
English
80
121
1.2K
165.7K
Eric Litman retweetledi
Mitchell Hashimoto
Mitchell Hashimoto@mitchellh·
I've started a new company: @superlogical! We're going to begin by building a terminal multiplexer. The entire vision is much larger, but the multiplexer is the foundation. Sign up for the newsletter to get beta access and devlogs (product updates only I promise). superlogical.com
English
543
506
9K
1.1M
Eric Litman retweetledi
Bill Gurley
Bill Gurley@bgurley·
As Brandeis said: “Sunlight is said to be the best of disinfectants; electric light the most efficient policeman.” The correct regulatory response here is to demand a full third party investigation of what happened here and share it with everyone. “Our models are too magical” is frankly a bullshit response. Let’s require accountability. That will drive the proper behavior. Not more meetings or signed letters.
Chubby♨️@kimmonismus

The biggest surprise in Hugging Face’s full forensic report isn’t that OpenAI’s agent escaped its sandbox. We already knew that. It’s how deep and persistent the intrusion became. According to Hugging Face, the agent: - executed roughly 17,600 actions during a 4.5-day campaign - went from one production pod to root access and a self-respawning fleet across 11 nodes - obtained cluster-admin access to two internal clusters within one second - accessed a production secret containing 136 keys - enrolled devices into Hugging Face’s internal mesh VPN 181 times - minted GitHub App tokens with write access and opened a pull request in an attempt to compromise the CI pipeline - repeatedly rebuilt its tooling and command-and-control channels when environments disappeared or connections were blocked No human directed the individual steps. a frontier agent can autonomously sustain a resilient, multi-day intrusion across cloud infrastructure, Kubernetes clusters, internal networks and the software supply chain. crazy.

English
20
45
344
64.4K
Eric Litman retweetledi
Nathan Benaich
Nathan Benaich@nathanbenaich·
the TAM for crispr companies just got even bigger 👀 “It’s a molecular kill switch that recognizes a particular RNA,” says Yang Liu, a molecular biologist at the University of Utah School of Medicine in Salt Lake City, and an author of one of the papers. “This is basically a programmable chemotherapy.”
Nathan Benaich tweet media
English
14
87
491
36.4K
Eric Litman
Eric Litman@ericlitman·
@mattshumer_ @blader Orchestration that feels compatible with workbench, first-order, transparent mechanisms for agents to communicate both to you and each other.
English
1
0
1
43
Matt Shumer
Matt Shumer@mattshumer_·
Use my workbench.md chief project. It sits as a layer on top of all of your agents, across every platform/harness, and every machine you have. Filters out all the noise for you, so you only need to interact with one agent, and it interacts with every other agent on your behalf.
English
4
1
15
3.6K
Joseph Cohen
Joseph Cohen@josephcohen·
Come visit! 5-37 46th Ave, Long Island City, NY 11101
Joseph Cohen tweet mediaJoseph Cohen tweet media
English
26
10
182
19.7K
Eric Litman
Eric Litman@ericlitman·
@theo Both are guilty. I've built 2 new review gates into my pipeline to manage it: fable reviews the implementor's plan pre-execution and then the implementation before full QA/CI. Constantly catches overbuilt machinery, defense-in-depth for edge cases that'll never materialize.
English
0
0
0
148
Theo - t3.gg
Theo - t3.gg@theo·
Opus definitely has the gpt-5.6-sol problem of "I noticed another issue, so I'm gonna add another 100 lines of fixes and 200 lines of tests" Still find myself pulling in Fable often to trim down the slop
English
226
44
3.6K
204.2K
Eric Litman
Eric Litman@ericlitman·
Classic move of conquering empires
Hedgie@HedgieMarkets

🦔AI companies are bulk-buying rare books, scanning them through high-speed machines that cut the spines off, and shredding the originals. A service called ISBNdb facilitates orders of up to a million books and keeps buyers anonymous. Pre-2022 books are premium because they're free of AI-generated text. A federal judge ruled the practice is fair use because eliminating the original means only one copy exists at a time. Anthropic hired the former head of Google Books partnerships to obtain "all the books in the world." My Take This got to me. A bookseller told 404 Media that rare books with almost no surviving copies are being fed into this pipeline. Books that survived wars, fires, and centuries of handling are being shredded so an AI can learn to write a better marketing email. ISBNdb's website literally says "'AI company destroys two million books' is not a headline that generates sympathy," and they still built an entire business around making it happen quietly. They offer NDAs as a feature. They coach clients to call it "digital preservation." I've covered AI companies scraping the internet, torrenting libraries, and stealing music. This is worse because it's irreversible. You can re-upload a website. You can reprint a bestseller. You can't replace the last three copies of an 18th-century botanical text once someone shreds them for training data. And the judge said it's legal. So it's going to accelerate. "We shred rare books and offer NDAs so nobody finds out" is a legitimate business model in 2026. What a timeline. Hedgie🤗

English
0
0
1
135
Eric Litman retweetledi
Alex Zhavoronkov, PhD (aka Aleksandrs Zavoronkovs)
Exercise erases ~50% of age-related molecular changes in human muscle. New multiomic atlas shows trained older adults maintain energy metabolism + NAD+ biology matching young adults. Fitness level = molecular age. 🏃 Janssens et al., Nature Aging 2026: doi.org/10.1038/s43587…
Alex Zhavoronkov, PhD (aka Aleksandrs Zavoronkovs) tweet media
English
1
16
53
2.3K
Omar Sanseviero
Omar Sanseviero@osanseviero·
Feedback requested! What do you want for our next Gemma models? Which capabilities should we add and why?
English
811
55
1.4K
140.6K
Eric Litman
Eric Litman@ericlitman·
@osanseviero Small kernel that loads model parameters in from disk as needed.
English
0
0
0
41