I just found out something quite interesting in the world of AI. It turns out that Jed McCaleb, the founder of Ripple and Stellar, is moving serious money into AGI research. Through his non-profit institute, Astera Institute, he committed one billion dollars to develop artificial intelligence systems that truly replicate how the human brain works.

What catches my attention is his critique of the now-dominant Transformer architecture. McCaleb argues that these models only do one thing well: prediction. They lack real planning, genuine decision-making, motivation. Basically, they are highly sophisticated autocomplete machines, nothing more. That’s why he believes we need to revisit how the brain actually learns.

They have already hired Dileep George, who came from DeepMind, to lead the project. The plan is quite ambitious: expand the team to 30 researchers this year and publish everything openly. It’s not the typical secret-keeping move you see in Silicon Valley.

But wait, there’s more. Jed McCaleb also invested an additional 600 million specifically for neuroscience research. We’re talking about implanting brain-machine interfaces in mice, primates, and eventually humans to map neural activity patterns. Those patterns would become new AI architectures. It’s ambitious, sounds like science fiction, but it’s what’s happening.

The vision is clear: if we want true AGI, we need to understand the brain first. McCaleb strongly bets that the solution isn’t in making larger Transformers, but in completely changing how we think about AI architecture.