Amazon Quick Desktop Credit: Amazon Web Services

This week I attended the AWS Summit in New York, which I have found to be a useful checkpoint on AWS progress — a read on the prior year’s re:Invent and a preview of the one coming this fall. A number of things stood out, and my net read is that AWS made real progress in security and the developer stack, while its ambitions for a general-purpose user-assistance agent and product messaging remain promising but unproven. Consider this Field Note a high-level macro view, with deeper dives to come in the next few weeks.

What really stood out

The infrastructure story remains strong. There’s a new Graviton 5 chip, but the bigger story was Continuum, a family of agents designed to deliver earlier, better security at AI scale. As we are learning with Claude Mythos and Project Glasswing, scaled security is becoming a defining problem — and AWS is right to push detection and remediation more to the left in the build process.

A much-improved development and code automation layer. The keynote’s baseline argument — that more productive coding does not equal faster or better releases — led to a more holistic approach across the entire SDLC. The developer story is no longer just the Kiro IDE, which now has an iOS app similar to Claude Dispatch and builds on the spec-driven approach I covered at re:Invent (Six Five On the Road). There are also new agents for Dev/Ops and release management, plus the next evolution of Transform as a continuous software improvement tool. Additionally, there were many additions such as a new Harness automation capability in the AgentCore foundation I detailed in my December research paper. AgentCore will get a deeper dive from me in the coming weeks as I research the future of AI Cloud Stacks. 

What to keep an eye on

Entering the general-purpose agent space with Amazon Quick. Quick is AWS’ enterprise answer to Claude Cowork and is similar to Microsoft Copilot Cowork. I like the embedded knowledge graph for context, which should improve results the more it is used and spare users from building their own second brain, as so many of us do with Claude today. But it is a new product and my first impression is that it needs more soak time with customers on usability. That said, I will be using Quick over the next week or two to see if my impressions change. Ultimately, the real test is whether the Quick team can wire in AWS’ core agentic differentiators — model choice, security, and evaluations in a future release. I have high expectations for Quick by the re:Invent timeframe.

Evolving AWS messaging and positioning. It was notable to see Continuum and the developer portfolio placed in a clear customer context — explaining specific challenges and issues — rather than buried in the usual avalanche of loosely connected product announcements. That is a welcome shift. The caution: marketing can quietly set the stage for a change in strategy, and I’d advise AWS not to stray too far from what it does best, which is letting builders build great stuff as opposed to the more integrated stack approach of AWS’ hyperscaler competition. 

Walking away from the event, AWS showed continued, credible progress on the agentic and AI fronts, and I respect its willingness to test new products and marketing approaches. But as with last year, several of these innovations will need time to prove themselves — and Quick and the new messaging are where I’ll be checking AWS’ work first.