Some secrets are worth keeping. 🔒 Stellar Hacks: Real-World ZK — build anything with zero-knowledge on Stellar. Privacy pools, private payments, confidential tokens, proof-of-anything. $10,000 in XLM • June 15–29 • Details 👇

Much appreciate it to @superteam and the @dev3pack community for an incredible first community call! 🇨🇷 @SolanaCostaRica Thank you for support to the global Solana ecosystem for Costa Rica community builders, developers, and creators

Costa Rica ganó el #1 global en Dev3Pack. Ahora nos organizamos. El miércoles hacemos la primera llamada comunitaria de Solana CR: quiénes estamos, qué se está construyendo acá y cómo sumarse. Miércoles 10 de junio, 5:00 PM. Online.

The first ERC-8183 Builder Session goes live on X today. Time: 10:30 AM ET / 2:30 PM UTC / 10:30 PM SGT Join us to learn why ERC-8183 was built, and how we’re making it easier for anyone, anywhere to build AI agents that can request work, pay for services, coordinate execution, and settle outcomes onchain. Hosted with the Ethereum Foundation dAI team, featuring builders from OKX, PRXVT, BNB Chain, and Virtuals. Set your reminder below:

Updates since then: * Deepseek v4 is out. There *is* a 2-bit quant that can run within 90 GB ( huggingface.co/antirez/deeps… ), and it works, however it's only fast on Apple hardware (I've head ~35 tok/s). On AMD, it's ~7 tok/s. IMO actually taking the effort to properly support more than one hardware manufacturer is a great example of the difference between mere "decentralized AI" and genuine "CROPS AI". I hope we can become better at this. * github.com/vbuterin/messagin… also has alpha telegram support now. However, the path to adding your account is quite janky * github.com/Luce-Org/lucebox-… looks promising as a way to run "dense" models (eg. Qwen 27B) more efficiently. It's janky, but on my 5090 laptop it seems to be ~2x more tok/s than llama.cpp * VoxTerm (local AI recording, no third-party servers) continues to be developed github.com/dmarzzz/VoxTerm And there's a lot more projects coming on the horizon. One other thing that has been on my mind is that there's actually a lot of intersection between "CROPS ethereum access layer" and "CROPS AI". For example, we want a ZK way to make (paid) calls to remote LLMs. But if we have this, then it's just as useful for solving another problem: private RPC reads in Ethereum. Another example: application-specific finetuned LLMs. Leanstral ( mistral.ai/news/leanstral ; I get ~38 tok/s on AMD) fits into < 70 GB, but can hold its own against 1T models on writing Lean code. Things like this are a huge boon for writing more secure code ( vitalik.eth.limo/general/202… ). We should have models finetuned for Ethereum-related use cases as well.