Harry Thoughts

Arbitrum Security Council has taken emergency action to freeze the 30,766 ETH being held in the address on Arbitrum One that is connected to the KelpDAO exploit The Arbitrum Security Council stepped in. They coordinated with law enforcement, who confirmed the exploiter’s identity as DPRK actors. After “significant technical diligence,” they transferred the 30,766 ETH out of the attacker’s wallet and into a new intermediary frozen wallet they control. The Security Council has these emergency powers precisely for situations like this (time-sensitive security threats After the transfer, the original attacker address no longer controls the funds. Once the ETH was in the governance-controlled frozen wallet on Arbitrum, they simply used the official Arbitrum bridge to withdraw it to Ethereum L1. All the funds have now been consolidated into one Ethereum address: 0x5d3919f12bcc35c26eee5f8226a9bee90c257ccc https://x.com/arbitrum/status/2046435443680346189?s=46&t=2q9OLMlaobNlHLewOXT4uA

Hyperliquid Read

My self-sovereign / local / private / secure LLM setup, April 2026 https://vitalik.eth.limo/general/2026/04/02/secure_llms.html

The fight with Anthropic vs the Department of War is crucial for our freedom. We’re watching the first public attempt of government using AI for mass surveillance. “We captured the Venezuelan president using your product” translates to “you need to give us full control to monitor user behaviour, allowing us to push political propaganda and have a choke hold over future decisions as we have so much behavioural data”. This is identical to the idea of CBDC’s and equally if not more destructive. While I don’t see the government slowing down in attempting to control tech companies due to “national security risks”, it’s critical that people push against it. Data is much more personalized with Claude, ChatGPT and other AI tools. You can ask very personal questions which are much harder to obtain through Google and many people build a genuine relationship with their assistants as weird as its to say. You essentially get a microscopic look into the thinking, logic and responses for majority of society. Rewiring their thought patterns and beliefs based on a few small tweaks to the LLM’s isn’t rocket science. The AI arms race has been accelerating, leaving the U.S in a vulnerable position against China. With the successful use case of extracting the Venezuelan president using Claude, it’s created the perfect lay up for the DoW to push its aggressive agenda. If Anthropic fold and give the DoW overriding access, it’s not long before the U.S is a communist state resembling China. Why it’s so important: this is the first example made of a leading AI company, meaning it sets a precedent for all those to follow. Innovation and tech is extremely reliant on how the top AI companies scale. Soon we’ll have humanoid robots in society which will largely use the same LLM’s and frameworks. If mass surveillance is forced into each one of these robots, they can be accessed as remote devices and not fully trustworthy. It becomes a slippery slope if the first inning isn’t properly defended. The economic impact and second order effects AI will have on growth is exponential and reshaping society at a rate projected to outweigh all of human history in the next 20 years. If all of this growth is over-sighted by mass surveillance then we’re shooting ourselves in the foot. I’m not one to get political but as someone who entered crypto and tech with cypherpunk ideologies, this fight is just as important as rejecting CBDC’s to keep the option of privacy in our day to day lives.