Fabrizio Batista 🅰️➖

🖖

[Fabrizio Batista 🅰️➖ @FabrizioBatista]

I keep coming back to the same point.

Once you run the model on-premise, you stop handing your real IP to a hyperscaler.

For a lot of businesses that isn’t a nice-to-have. It is the entire reason to care.

Most of the conversation is still about which model is smarter or which cloud has more capacity. Meanwhile the companies that actually have sensitive knowledge and processes are quietly deciding they will never put that material on someone else’s servers.

Local by default. Cloud by exception. That framing is starting to feel more realistic every month.

Curious how many people are still treating this as a secondary concern.

[Jared Duker Lichtman @jdlichtman]

Impressive for sure, but not surprising.

If you found this news surprising, pause and reflect.

Seek out those who did not. You'll be less disoriented for what's next.

[Noam Brown @polynoamial]

An internal version of Astra, @OpenAI’s next major model family, solved 10 major open problems in mathematics, quantum complexity, and theoretical computer science.

We believe it will be a major step for scientific reasoning. https://t.co/iP6cyheZ7i

[Sundar Pichai @sundarpichai]

Very happy to support this on behalf of Google. We have long benefited from open source, are big contributors to open source and in fact have consistently made open weights models with Gemma available from @GoogleDeepMind @demishassabis . Onwards!

[Satya Nadella @satyanadella]

Open-weight models are essential to a healthy AI ecosystem. Together with others across our industry, we are outlining a path for open-weight models to strengthen American competitiveness and expand economic opportunity, while protecting national security. https://t.co/Tr0sAzAxTD

[Guillermo Rauch @rauchg]

Fable just found a 15-30% memory efficiency improvement in Turbopack / Next.js, nearly autonomously.

@tobi asked me today: what have been your “holy s***” moments with AI?

My answer was: it’s every single week. And it’s accelerating. In fact, “WTFs/day” might just be my favorite metric for AI progress.

It’s one thing to read benchmarks or stories online. It’s another to watch these machines pull engineering feats every day.

3 days ago? Sol helped us find novel vulnerabilities in some of the most audited code in the world. Today? Fable helps us ship this large optimization of a very complex Rust codebase. I just saw some results of work we’ve done to shrink binaries by 10-20x. List goes on.

[Elon Musk @elonmusk]

We are in the Singularity

[QM @qasimmith]

@jimcramer With all due respect, I don't think you understand what it means to run open weight models. If an American company runs a Chinese open weight model locally on AWS or on their GPUs (which they actually do), no data is going to china. Happy to go over this in as much detail.

[Chubby♨️ @kimmonismus]

Vercel CEO: Kimi K3 demonstrates true frontier performance in covert internal cybersecurity evaluations – likely not benchmark overfitting.

Sol is significantly more powerful, but more expensive; Fable, on the other hand, is practically unusable. Open-weight models have thus reached the forefront of cybersecurity.

Kimi K3 is that good.