🇨🇭

the devnet cluster has initiated another update prior to the integration of new proof types additionally, a high speed matrix processing mechanism (utilizing HFHE) will be added for inference, running the same aml code locally demonstrates an approx 14-fold speedup compared to existing results * several new hfhe opcodes (octra vm) have also been added (one of which is multi layer division by an arbitrary variable and a constant) an additional test cluster scanner is available at the following address: http://139.59.183.141:18090/index.html

the largest dev net update is currently underway, support for direct GPU computing for resource-sharing circles has been added to nodes as part of the development of on-chain state inference mechanics we plan to complete this tomorrow afternoon (this will deserve a separate article) thanks everyone

The answer to a bad AI monopoly is not a “good” AI monopoly. A world where many closed labs trade the frontier is much more resilient than one with a single winner. A world with open-weight models close enough to provide a credible exit is better still. Especially if they carry different values and answer to different jurisdictions. But the strongest escape hatch is full-stack open source: not just weights, but code, training recipes, evals, inference stacks, and safety tooling that people can run, inspect, verify, modify, and fork. The goal is not to guarantee that every model is benevolent. It is to ensure that no model, company, or state becomes the unaccountable choke point for intelligence. Open, distributed, censorship-resistant.

🤔 the devnet just came back online, and someone is already putting encrypted LLM chunks onchain?