The startup Adaption has launched a new tool called AutoScientist , an artificial intelligence system designed to automate one of the most complex and costly processes in model development: the training and optimization of other AIs. The news was published by TechCrunch .
The proposal aims for an ambitious goal: enabling AI models to partially improve themselves autonomously, by testing configurations, adjusting parameters, and evaluating results without constant human intervention.
According to the company, AutoScientist functions as a kind of "automated researcher" capable of executing experiments, comparing results, and selecting the most efficient iterations during model training. Instead of engineers manually adjusting each variable, the system aims to accelerate the entire development cycle using advanced automation.
The announcement reflects a growing trend within the industry: using AI to build and optimize new AI models. Large technology companies are already experimenting with similar approaches to reduce computational costs and accelerate research on foundational models.
The competition is intensifying particularly because training modern models requires enormous amounts of resources, energy, and processing power. Even small improvements in efficiency can represent millions of dollars in savings for companies operating large-scale AI infrastructure.
The launch of AutoScientist also comes at a time when the sector is experiencing extreme acceleration. In recent weeks, companies like Microsoft have announced AI-powered autonomous vulnerability search tools, while Anthropic continues to rapidly expand its business presence in the face of OpenAI.
For many analysts, the real change is not just in more powerful models, but in the increasing automation of the scientific and technical process behind artificial intelligence itself. Tools like AutoScientist are precisely aimed at this scenario: systems capable of experimenting, evaluating, and optimizing other systems in a semi-autonomous manner.
This advance also raises new concerns within the technological and academic world. As AI begins to directly intervene in the design and training of future generations of models, the debate surrounding control, transparency, and human oversight intensifies.
Although we are still far from fully autonomous systems capable of developing advanced artificial intelligence without human intervention, projects like AutoScientist demonstrate the direction the industry is heading: an infrastructure where machines not only perform tasks, but also actively participate in improving their own learning processes.