Models, margins and the business decisions behind applied AI. Reporting and analysis from Fathom Tech.
Google has begun distributing Gemini 4 Argon, its latest frontier AI model, to vetted cybersecurity professionals through a restricted access program. The model surpasses competing systems from Anthropic and OpenAI across most of Google's internal performance tests.
As inference workloads overtake training in AI deployments, CoreWeave is moving beyond GPU capacity into networking, storage and software to optimize the full system's token production efficiency.
The heads of major technology companies have endorsed a voluntary safety framework for artificial intelligence development, with President Trump announcing plans for a dedicated federal AI safety committee.
As AI models demand ever-greater compute and infrastructure, Cerebras Systems CEO Andrew Feldman will discuss whether the industry can sustain this trajectory and what happens when conventional hardware hits its ceiling.