own interactive, cloud-based IDE. This means you can easily multitask, tackling numerous tasks concurrently and stepping in to steer when needed. https://cognition.com/blog/devin-2 12
system that turns a prompt, a ticket, or a specification into tested code for an existing codebase, using one agent or several, and increasingly deploys and maintains what it builds. https://agentbaseline.org/software-factory.pdf 14
… whose approval policy approves every change automatically, and the same controls apply, with a person still accountable for what it ships. https://agentbaseline.org/software-factory.pdf 16
AI adoption on organizational performance becomes strong and positive. Conversely, when platform quality is low, the effect of AI adoption on organizational performance is negligible. https://dora.dev/capabilities/platform-engineering/ 19
the cognitive load on product teams. A platform should encapsulate implementation details and hide any complexity that might arise from its architecture. https://tag-app-delivery.cncf.io/whitepapers/platforms/ 22
that keeps track of ownership and metadata for all the software in your ecosystem (services, websites, libraries, data pipelines, etc). The catalog is built around the concept of metadata YAML files stored together with the code, which are then harvested and visualized in Backstage. https://backstage.io/docs/features/software-catalog/ 25
requirements of its users and it should be designed and evolved based on those requirements, similar to any other software products. https://tag-app-delivery.cncf.io/whitepapers/platforms/ 34
agents enforcing their own limits. … enforcing what they can actually do outside their decision loop, separating an agent’s planning from the authorization and enforcement of its actions. https://agentbaseline.org/software-factory.pdf 44
matches the risk, from a container for trusted work to a microVM for untrusted code, and the sandbox is destroyed when the task ends. https://agentbaseline.org/software-factory.pdf 47
state, including memory and file system changes. When you create new replicas, they're restored from the snapshot, letting the workload resume instead of starting from a new state. https://docs.cloud.google.com/kubernetes-engine/docs/concepts/pod-snapshots 56
on the capabilities of Agent Sandbox on GKE, and improves on Agent Sandbox on GKE by bypassing the bottlenecks of the standard Kubernetes control plane. https://docs.cloud.google.com/kubernetes-engine/ai-ml/about-agent-substrate 61