Things I have been thinking about: > I cannot see how AI agents will proliferate without strong verifiability & privacy. The Claude for Financial Services + GPT agent are in my eyes half-baked. > Companies should write their Culture statement in code/smart contract. Culture is the closest we have in Context Engineering. An agent operating inside Amazon vs an agents operating inside Netflix should not perform a task in the same way. These companies have different "operating models", culture and way of doing things. If not, then in the near future companies risk of losing any comparative advantage. > I am not into sports betting but if I were I'd be obsessively using Polymarket + Grok to find opportunities for easy money. This still seems to me an inefficient market that hasn't adjusted to the new reality. > Memory isn't retrieval. Memory is generation. As written before, the next big AI breakthrough will come from the field of cognitive science focusing on memory. > In my eyes, reinforcement learning=spaced repetition. If that's true, what other AI technique could inspire a new learning method for humans? > Blockchains and ZK are the most important political upgrade since the US constitution.