Primer Wants AI to Keep Learning After Training

Field notes

The project organizes its thesis around a staged progression: acquire knowledge after training, retain it through later learning, use it as a building block, and eventually improve the learning process itself. Its current evidence centers on a 1.3B-parameter system, including a reported 0.852 Novel Knowledge Acquisition Rate and perfect scores on two listed internal evaluations. The site separates demonstrated capabilities from active and future research, and notes that external validation remains ongoing.