Swamp = Drained

We're close to a huge moment in the UK car market: Chinese EV firm BYD is about to overtake Tesla

They solved environmental consistency with Genie 3, and this was an emergent capability. You can see the trees remain the same even after being out of line of sight. Visual memory extends back one minute now. Google is on a steady path to a real world simulator.

Services #ISM now screaming LATE-CYCLE #STAGFLATION. Prices going up and New Orders going down. Not seen in ~18 years since before the GFC.

5. Foucault interessierte sich für die Frage was Macht mit Erkenntnis und Wahrheit zu tun hat. Er untersuchte dafür, zunächst als klinischer Psychopathologe, die Bereiche, in denen Macht und Wissen problematische Beziehungen eingehen – Grenzbereiche wie die Psychiatrie, …

… den Strafvollzug, die Klinik usw. All das tat er als Wissenschaftler, mit Thesen und Argumenten, die einen Wahrheitsanspruch verteidigen. Dass es Foucault um die Frage „Wer hat die Macht?“ gegangen sei, zeigt, dass Vince Ebert nicht weiß, wovon er redet und Foucault nicht …

… gelesen hat. 6. Die „Suche nach Wahrheit, Wissen und Erkenntnis“ ist – buchstäblich – Foucaults zentrales Forschungsinteresse. Seine Lehrer waren ja auch angesehene Wissenschaftstheoretiker. Foucault hat sich politisch engagiert, ja – aber das gegen sein umfangreiches…

Today we release gpt-oss-120b and gpt-oss-20b—two open-weight LLMs that deliver strong performance and agentic tool use. Before release, we ran a first of its kind safety analysis where we fine-tuned the models to intentionally maximize their bio and cyber capabilities 🧵

We know that increasingly capable LLMs, while generally beneficial, can also accelerate frontier risks in areas such as biology and cybersecurity. We took extra care to study these risks before releasing open-weight models permanently to the world.

I’ve been fortunate to lead the infra and inference work that brings gpt-oss to life. A year ago, I joined OpenAI after building vLLM from scratch — It’s deeply meaningful to now be on the other side of the release, helping share models back with the open-source community.

I told you OpenAI’s open-source model wouldn’t just be o3-mini tier. It’s literally frontier o3 level at a tiny size you can run at home. If they’re releasing this for free, open source… Just imagine what GPT-5 will be. Your heads will melt.

On benchmarks and "vibe", the top GPT open source model is better than the best LLM in the world one year ago. On compute: someone (Nvidia?) will make a killing with a high-RAM, GPU, plug-in "LLM box" you connect to your computer. On safety: there are some real risks. 1/2

Open source *will* be jailbroken. Assume no safety limits at all. These models are persuasive enough that beyond, e.g., writing adult content (don't tell the UK govt!), their use for scams and so on is obvious. Not sure it's responsible to open source anything more powerful...2/2

GPT-oss is blazing fast on Groq

Exports have fallen right off a cliff since March. Countries around the world are responding to the U.S. trade war by boycotting American goods and services. Another penalty that Americans are paying for Trump’s tariff crusade.

My initial impression on OpenAI's OSS model is aligned with what they advertised. It does feel closer to o3 than to other open models, except it is much faster and cheaper. Some providers offer it at 3000 tokens/s, which is insane. It is definitely smarter than Kimi K2, R1 and

this is the only language Elon understands.

Underrated tweet

🧐

"…Although we believe most people will want to use a convenient service like ChatGPT, people should be able to directly control and modify their own AI when they need to, and the privacy benefits are obvious."