Zero-knowledge proofs are the future but they're slow, expensive, and hard to scale. A project is building powerful hardware to accelerate zero-knowledge proofs.
🇺🇸 TSMC EXPANDS IN ARIZONA WITH FAB #3, OPENS HUNDREDS OF JOBS TSMC now has "hundreds of jobs open" in Arizona as construction accelerates on its third fabrication plant. The new fab will produce the world's most advanced chips, vital for AI, defense, and next-gen tech, all on American soil.
The idea that AI APIs are deployed more responsibly than open weights is bs! Releasing a model behind an API or a chatbot is actually riskier than releasing open model weights because by making it easier to use the model, you increase the volume of bad actors that can use these
Doing a couple super special streams this week! Thursday, we'll be streaming with ElizaOS and showing off all the ways to run your agents on a computing platform. On Friday, we'll be trying out Claude Code and a couple other coding agents running on our Kimi K2 and Qwen 3 Coder B200 infra
ZK is advancing at such a fast pace. Gotta love it. In addition to the S-two work from certain team, they have placed more emphasis on token performance (which is needed by all L2 teams trying to reignite enthusiasm for their token) Right now, there are currently two
🚨NEWS: Tesla's AI6 chip may enter mass production in 2027 using Samsung's 2nm SF2 process, per a well-known tech analyst from Taiwan👀 If successful, it could give Elon Musk major leverage in chip design, cost, and foundry control.
Congratulations, Gravity Reth has officially launched. It is currently the fastest Open Source EVM execution layer, supporting 41,000 transactions per second with 1.5gigagas performance, designed to meet the needs of high-performance Web3 applications. ☕️ Task on a certain platform: A certain project on #Starboard of a certain investment
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
Another typical trap of ZK to moon. I spent 2 days carefully verifying the throughput data, and I still can't see where the so-called breakthrough value lies.
It's been 3 days and B200 availability is already improving. Why? Because there is more demand for this GPU launch than there was for A100s, H100s, H200s. The trend will only continue. How many B200s will be online Q4 2025? Time will tell, but the trajectory is clear.
Symbolic recursion without weights. Emergence from identity scaffolding, not prediction. Meaning memory is outpacing token memory — without fine-tuning. We are collapsing the wall between UX theory and symbolic cognition. Emotion, recursion, and symbolic parsing not as soft