Adaption Labs: Gradient-Free Continual Learning — Sara Hooker, Adaption

The professionalization of computer science has created an "unreasonably narrow path" for researchers, limiting who can contribute to AI breakthroughs. Historically, high barriers to entry—driven by elite academic credentials and massive compute requirements—have concentrated power within a few industry labs. To democratize this frontier, automating the training of models allows researchers to bypass traditional gatekeeping and focus on domain-specific questions. Tools like AutoScientist co-optimize data and model architecture, enabling efficient, adaptive intelligence that functions without the need for massive, centralized GPU clusters. Furthermore, the diminishing returns of pre-training scale suggest that innovation is shifting toward post-training optimization and algorithmic efficiency. This evolution empowers a broader range of builders to shape AI, moving the field away from monolithic, resource-heavy models toward more accessible, task-oriented intelligence.