Amazon's engineering leaders sent revised software engineering guidelines to developers earlier this year, accompanied by a letter urging them to learn AI and stay close to their customers and products. The update emphasizes measuring engineers by outcomes rather than inputs. Brooker said AI agents now generate almost all implementation code for the vast majority of Amazon's developers. A previously unreported team, Velocity Labs, released an open-source AI model less than a week after its concept.
The group, whose existence was not previously reported, formed in April and has about 35 people, most based in Seattle. It is designed to think beyond Amazon's normal product-oriented approach, prototyping quickly in areas of AI the company sees as underexplored.
Marc Brooker, the AWS vice president and distinguished engineer who leads the team, said the group is an example of creating organizations suited to AI tools. He described a complete change of tools: two years ago, most Amazon developers were still writing code by hand; now, for the vast majority of the developer population, almost all implementation code is agent-generated.
The revised guidelines, sent earlier this year, were accompanied by a letter from a vice president and distinguished engineer in Amazon's stores business. The letter advises engineers to learn AI now and to stay close to their customers. It emphasizes measuring engineers by outcomes rather than inputs such as tokens spent or lines of code committed, which Brooker called anti-patterns. The letter says Amazon has found that usage metrics correlate with important outputs, but says they should be used with care.
Asked whether AI use is a direct factor in performance reviews, Brooker said it is not in a metrics sense, but could factor in to the extent it makes someone a stronger performer.
The transition has had challenges. In March, Amazon acknowledged that an engineer who followed inaccurate AI agent advice contributed to one of several outages on its retail site. The Financial Times reported that in May Amazon shut down an internal leaderboard called KiroRank after finding that some users were using the AI coding tool excessively just to climb its rankings.
The guidelines tell engineers that if their work does not benefit from current AI tools, that is a finding, not a failing, as long as they can explain why.
On jobs, Amazon has cut roughly 30,000 corporate roles since late 2025. CEO Andy Jassy has said he expects AI to shrink the corporate workforce over time. AWS CEO Matt Garman said Amazon is hiring as many software developers as ever. An Amazon spokesperson told Business Insider that more than 11,000 interns and early-career software engineers would join the company globally this year. Brooker said he does not think AI changes the number of engineers at the company.
Analysis
Why This Matters
- The guidelines from one of the world's largest employers of software engineers signal how the entire industry may evaluate and manage engineering work in an AI-driven environment.
- The creation of Velocity Labs demonstrates how major tech firms are restructuring to match the speed of open source and startup innovation in AI.
- The tension between maximizing AI usage and preserving engineering judgment is a challenge facing the entire technology sector.
Background
Amazon has been developing and deploying internal AI coding tools for years. The formation of the Velocity Labs team in April 2026 and the release of revised engineering role guidelines represent an effort to formalize what the company has learned from a period that included AI-related outages and the gaming of internal AI usage metrics. The company has cut around 30,000 corporate jobs since late 2025 while maintaining that it is hiring software developers at a consistent rate.
Key Perspectives
Amazon Engineering Leadership: The new guidelines reorient performance around customer outcomes rather than coding input metrics. They acknowledge that AI tools are not always the right answer and direct engineers to document where they fall short.
Marc Brooker (AWS VP, Velocity Labs lead): Sees the transition to agent-generated code as a complete change of tools but argues it does not reduce the number of engineers, because the company values ownership and judgment.
Amazon Engineers (via New York Times, Financial Times reports): Some have described managers exerting pressure to adopt AI, with increased output goals and less forgiving deadlines.
What to Watch
- The balance between Amazon's corporate headcount reductions and its stated steady rate of software developer hiring.
- Whether the new outcomes based evaluation system prevents the type of metric gaming that led Amazon to scrap the KiroRank leaderboard.
- What specific agentic AI products emerge from the Velocity Labs team and how quickly they reach the market.