@vitalikbuterin

Doing a bit of a self-experiment. Goal: use my personal health and travel data to provide personalized diet and exercise recommendations for me, using frontier models but in a way that avoids leaking to them any private information. Strategy: use a local model (Qwen 3.8 Flash Next) to orchestrate, and use powerful remote models as a tool call to benefit from higher-level thinking and knowledge that the local model does not have. Three-layer approach to privacy: Avoid leaking personally identifiable information, or leaking my identity through writing style -> local model writes the queries to the frontier model, not me Avoid leaking who I am through the payment channel -> zkAPI Avoid leaking…

Also this thing is done now

Testing out some of the mobile offline local knowledge apps that people have been trying to build (see here ) Definitely getting much better than the one I tried to build myself 2 months ago. But also still much slower and less effective at difficult questions than the models that can run on a laptop. It's weakest at specialized travel-related queries (eg. my eval is "Tell me the best vegan restaurants in [city I am currently in]", unfortunately none of these performed well on that) Looking forward to seeing these continue to improve! I hope we can soon get to the point where you can comfortably look up any facts about the world that you care about without needing to…

d/acc: one year later

@JJCloud777 @AITweep d/acc is more object-level, plurality is more meta-level imo. They're complementary.

"Proof of Stake", the (physical and digital) book compiling various writings I've made over the last ~10 years, will be out in a month! You can get a signed digital copy and NFT here: My share of the proceeds all goes to @gitcoin grants public goods!