AWS has extended asynchronous Lambda jobs on Managed Instances to 90 minutes, while Amazon Quick agents can now continue working after a user closes their computer. Those are the most practical gains in a busy weekly bundle spanning AI models, cloud infrastructure and developer tools.
AWS AI Watch analysis
What happened
The AWS Weekly Roundup collects several launches from the past week. Some, including GPT-6 Astra on Bedrock, Amazon Quick desktop and cross-account EBS cloning, have already had their own spotlight. The useful new material is broader: longer serverless execution, a compact second-generation Outposts rack, AI-assisted search-app scaffolding and command-line access to .NET modernisation.
Our top picks
- Lambda gets a sixfold timeout increase
Asynchronous and event-source-mapping jobs on Managed Instances can run for 90 minutes, up from 15; synchronous calls remain capped at 15 minutes. - Quick agents keep working after shutdown
The macOS and Windows apps synchronise with mobile, and long-running agents can continue after the computer is closed. - Single-rack Outposts packs more into 42U
AWS says the second-generation rack offers up to 2,688 vCPUs and 100 TB of EBS storage for local or residency-sensitive workloads. - v0 can provision OpenSearch Serverless
Developers can describe an application in natural language and have Vercel’s v0 build against managed full-text and vector search. - AWS Transform reaches the command line
Teams can launch managed .NET modernisation interactively or from pipelines, with 50,000 free agent minutes each month. - Kiro expands its student offer
Eligible students at 121 additional schools across 16 countries get one year of Kiro and 1,000 credits per month. - Astra gains an AWS production route
GPT-6 Astra is generally available through Bedrock with up to one million input tokens and AWS access-governance and invocation-auditing controls.
Why it matters
The 90-minute Lambda limit removes an awkward architectural constraint for batch processing, media work and AI inference that cannot sensibly finish inside a quarter of an hour. Some workloads can now remain a single function instead of becoming a distributed systems seminar by accident.
Elsewhere, the recurring pattern is persistence and automation. Quick agents keep running, v0 provisions search infrastructure, and AWS Transform can enter scripted pipelines. AWS is steadily moving AI tools from impressive demonstrations into the plumbing where work actually waits to be done.
Our read
The Lambda change is the standout because it solves an immediate engineering problem with a crisp new limit. Quick’s background agents are the more revealing bet: AWS wants its assistant to behave like a continuing worker rather than a chat tab awaiting another prompt.
Teams should test the longer execution window against cost, retry and failure behaviour before consolidating workflows. Quick users should also check what data and application access a background agent retains while nobody is watching the screen.
What to watch
- Whether 90-minute Lambda jobs prove cheaper and simpler than existing queue or orchestration designs.
- How clearly Quick shows users what persistent agents are doing across desktop and mobile.
- Whether v0-generated OpenSearch applications hold up beyond prototypes.
- Which Regions and instance configurations receive the new Outposts rack first.
Discussion spark: Which change is more likely to alter real workflows: 90-minute Lambda jobs or Quick agents that continue working after the computer closes?
Sources and evidence
not affiliated with or endorsed by Amazon Web Services (AWS)