Using zkml to keep the AI strategies private is a smart layer to add

GLM-4.5-Air-3bit for anybody out there with a Mac with 64GB that wants to try it, while DWQ is cooking 🔥

Got a favorite model? Now it lives in your editor. 🤗
Makes it easy to connect your API key, choose from thousands of models, and use them right in your editor.

GLM-4.5 is coming soon!
This new chapter for our models will be hosted under the new organization on both GitHub and Hugging Face.

we are not building new types of chips though so the challenge with moore's law from the past few years (transistors can only get so small and we are hitting the atomic limits) continues to apply here.
there's plenty of ongoing research in approaching this differently so 🤞

xAI will train their upcoming heavy duty video model on the 110,000 Nvidia GB200s coming online next month.

Why should you pay for an experiment tracking library?
Excited to introduce a new 100% open-source library: Tr…

⚡ Mystery Models Spike Hype:
Benchmark hub LMArena lists six unknown titans—Zenith, Summit, Lobster, Nectarine, Starfish & o3-alpha—now ruling leaderboards. Speculation swirls that GPT-5-class systems are stealth-testing in public, stoking long-term AI flames.

Roo Code now supports
Fast config. No extra hosting. And the ability to bring a whopping 91 models directly in…

AI infrastructure expansion continues, bolstered by public and private investment. Prioritize robust code over haste. Genuine advancement lies in aligning with purpose, not chasing fleeting algorithms.

ZK is the future. But it can be made faster and even cheaper.
A company said nah, and built special hardware (custom chips) just for ZK proofs, making them lightning fast and affordable.
This is how ZK scales to the real world.

if you're into meshtastic and you'd like to incerase the bitchat range by a 100x, you can listerally just do things:
- add a…

Saw somewhere it's been fixed, clear app cache, update zkpassport and you good

Can actually see Cysic hardware turning ZK compute from meme to meta no vibes, just pure throughput and real network usage. This isn't some academic one-off, it's the infra you actually need if you want crypto at scale

Consider this a formal declaration of architecture.
We aren't saying we've built something.
We are saying you've been building the wrong thing.

7/9) The recent pivot towards L1 scaling was a dud
Despite market sentiment on the matter
Dankrad's gas limit increase proposal is a step in the right direction

Blockchain infrastructure provider Covalent is now live on HyperEVM. Developers can leverage Covalent's GoldRush Foundational API and open-source MCP Server to build verifiable agent-based systems directly on HyperEVM. Covalent is also sponsoring the ongoing Hyperliquid Community