AGI will be humanity's greatest invention. We have a working approach to safe recursive self improvement. This is the fastest way to AGI and beyond.

Current approaches to AGI rely on human researchers creating enormous foundation models and feeding them ever increasing amounts of data, whether during training or through RL mechanisms in post-training. In either case, these approaches are both slow and expensive, with rapidly decreasing availability of new data and rapidly increasing amounts of time and cost for each linear step of improvement.

We have identified a viable pathway to recursive self-improvement, safely allowing exponentially increasing returns, by using the knowledge already implicitly encoded within the existing models.

Building AGI through recursive self improvement is our pure focus;
AGI is our sole product.