I’ve come to think—maybe a bit later than I should have—that the hardest problem in AI isn’t making models smarter. It’s building systems where AI can actually operate in ways humans are willing to trust. The market still seems overly focused on reasoning ability, when the real bottleneck may be elsewhere: incentives, accountability, and the friction of real-world implementation.
An AI that can make decisions but can’t verify its own actions or carry responsibility for outcomes is still, in many ways, just an interface layer. That’s why the more important question isn’t how advanced the model is, but what kind of abstraction layer allows it to interact with the real world without creating even more noise.
That’s also what made me pay attention to Newton Protocol. At least from my perspective, it doesn’t look like a project built from the usual mindset of “let’s build AI.” It looks more like an attempt to design an environment where AI agents can operate under rules that are transparent, enforceable, and verifiable.
Of course, that only matters if the actual behavior of agents on the network ends up aligning with the incentives the system is trying to create. I still have doubts about whether the design is robust enough to handle unexpected behavior over time. In the end, the more interesting question may not be whether AI replaces humans, but what kind of system we’re quietly constructing for AI to live and act inside.
#newt $NEWT @NewtonProtocol
An AI that can make decisions but can’t verify its own actions or carry responsibility for outcomes is still, in many ways, just an interface layer. That’s why the more important question isn’t how advanced the model is, but what kind of abstraction layer allows it to interact with the real world without creating even more noise.
That’s also what made me pay attention to Newton Protocol. At least from my perspective, it doesn’t look like a project built from the usual mindset of “let’s build AI.” It looks more like an attempt to design an environment where AI agents can operate under rules that are transparent, enforceable, and verifiable.
Of course, that only matters if the actual behavior of agents on the network ends up aligning with the incentives the system is trying to create. I still have doubts about whether the design is robust enough to handle unexpected behavior over time. In the end, the more interesting question may not be whether AI replaces humans, but what kind of system we’re quietly constructing for AI to live and act inside.
#newt $NEWT @NewtonProtocol