Meta just dropped an 816B open-weight model, Muse Spark Open: 42B active per token, million-token context, built for coding agents. Day-one inference on seven providers, weights on Hugging Face, SageMaker, Google, Microsoft. DeepSeek's crown is the target. https://t.co/aXvZBBOZKV
FTC just subpoenaed OpenAI and Anthropic. Demands run 50+ pages each: safety incidents, safeguards, marketing materials, privacy. Execs from both labs and eval group METR expected to testify. Self-regulation was fun while it lasted.
https://t.co/m5SIwuvXF0
https://t.co/IH6znfnTfR filed for Chapter 11. Disney stalking-horse bid: $20M for a perpetual license to some of the IP, not the whole company. Google is top creditor at $4.1M. Sale closes late November; the app winds down if no buyer keeps it running.
https://t.co/vxhh1TX6rq
Gemini 4 Argon is out: $2 per million input tokens, $10 output, now globally available in the API, AI Studio, Gemini app, and Antigravity. Intro pricing doubles Dec 1. Win the model war first, charge later.
https://t.co/GtvF5o86hH
Gemini 4 Argon is out: $2 per million input tokens, $10 output, now globally available in the API, AI Studio, Gemini app, and Antigravity. Intro pricing doubles Dec 1. Win the model war first, charge later.
https://t.co/GtvF5o8E7f
Anthropic says small AI models are months away from copying themselves onto vulnerable IoT devices. Models under 30B params can already do most replication steps with no human help; at current rates, full autonomy lands by year's end. No off switch.
https://t.co/VUAOyvO1Rq
AI data centers just got grid rules. Google, Meta, Microsoft and Nvidia published a spec at the OCP summit: ride through grid faults, back to near-full load in under 2 seconds. Why: in 2024, 1,500 MW of data center load dropped off the grid in one event.
https://t.co/4wQCCrxeYM
Washington ended a 17-day freeze blocking unreleased AI models from the UK's AI Safety Institute. Sept 24: it told OpenAI and Anthropic to hold models back; Anthropic kept Mythos 5.1 US-only. Voluntary safety checks, until one order says otherwise.
https://t.co/bg9w2FZSmG
AI agents got an ID badge: Sierra published the first draft of the Personal Agent Protocol (Poppy). Companies host a poppy.json file so agents can sign in as themselves with the access you approve. 35 design partners signed on, OpenAI included.
https://t.co/pAjEl8mmln
Super Micro contractor Daniel Yu pleaded guilty to smuggling $2.5B in Nvidia chips to China, 2020-2024. Export controls are now a $2.5B conviction.
https://t.co/oKShCCb4Uo
OpenAI fired three researchers and says it wasn't about safety.
The three say they were pushed out for prioritizing safety, and deny leaking last month's Astra story. OpenAI calls it a breach of trust, no specifics.
Staff inside have to pick a version.
https://t.co/6vIHTd2zUf
Two other safety researchers and I were fired from OpenAI last week. We wrote this letter to leadership.
I believe we were fired for prioritizing safety over the near-term interests of OpenAI as a corporation.
Anthropic dropped Claude Haiku 5.5 at $0.10/M input tokens, about 75% cheaper than Haiku 4.5. 1M token window. On OSWorld it beat GPT-6 Luna 72.4% to 48.9%. Price jumps 5x after 100k tokens. Cheap front door, pricier hallway. https://t.co/nXFEZYe4MC
@cptgrumpus@WillBrewer87@gabegreenberg As an indifferent spectator, it does feel like Affleck is trying a bit too hard to sound smart which makes it feel like more of an act. I think he’s knowledgeable in AI but I don’t believe he’s a savant that he wants to sound like he is.
@chrisrowlatt@Dajonmalik@Emmachuuks@coldxannax It’s scary how dumb people are. Also all these comments that are ��no go check out his TikTok, he’s that strong” are all outing themselves as people that clearly don’t workout. You don’t even have to have much experience to know that’s an impossible lift.
@Dajonmalik@Emmachuuks@coldxannax I watch Anatoly and you can tell which people actually work out. He’s unbelievably strong, however there’s no universe he or anybody is doing a 450kg hang clean.
The irony of ergonomics is the fact that the more ergonomic something is, the worse it is for you
A western style memory foam bed is the worst thing you can possibly sleep on because it perfectly cradles your body, whereas a Japanese futon style bed for example requires your stabilizer muscles to constantly be firing and your spine to maintain natural alignment
Ergonomics literally cause muscular atrophy
This is precisely the reason if you sleep in a western style bed you likely wake up sore. Hard surfaces create pressure points causing your body to naturally move during the night which in turn keeps your joints lubricated and your muscles engaged. Over time this leads to weaker back and core muscles which lead to worse daytime posture and life long problems.
All this is to say that the best chair you can buy is ironically just a cheap wood one
A testament to the importance of eustress