Amazon Web Services has formed a small internal team called Velocity Labs, which went from concept to the public release of a new open-source AI model in less than a week. The group, whose existence hasn’t been previously reported, was formed in April and has about 35 people, most of them in Seattle. It aims to build quick prototypes in underexplored AI areas and move faster on agentic AI, with scientists and engineers encouraged to cross over between research and engineering. Marc Brooker, an AWS vice president and distinguished engineer who leads the team, said AI agents now write almost all the code for the vast majority of Amazon’s developers that would previously have been written by hand. The team is part of Amazon’s broader rethinking of software development for the AI era, laid out in revised guidelines sent to developers earlier this year. A letter accompanying the guidelines, published last week under the byline of Michael “Terk” Terkowitz, a VP and distinguished engineer in Amazon’s stores business, advises engineers to learn AI now and stay close to customers and products, and acknowledges that some engineers and managers have become “less connected to their customers and business.” Brooker said the update emphasizes measuring engineers and teams by outcomes rather than inputs, citing anti-patterns of focusing on tokens spent or lines of code committed. Asked whether AI use is part of performance reviews or promotions, he said, “Not directly in a metrics sense,” though it could factor in if