Push-Based Execution and A Simpler Plan View

Agents no longer rely on constant polling to keep work moving, and the workspace planning experience became more unified and less cluttered.

This day tightened the shape of autonomous work. Execution became more push-driven and completion-driven, while the workspace itself became simpler to read because planning and live activity moved closer together.

Execution Keeps Moving With Less Babysitting

Agents no longer have to wait around for the next poll cycle to notice that work can continue.

  • Plan submissions can trigger execution immediately
  • Sequential tasks now chain forward from completion, which makes longer orchestrated flows feel more automatic
  • Duplicate dispatches and stale retry edge cases were tightened up, so repeated or failing work is less likely to fork into confusing states

The Planning View Became More Cohesive

The workspace got easier to scan because current work and planning structure no longer live so far apart.

  • Activity and Planning merged into a single Plan tab
  • Live activity surfaces more naturally inside the planning flow
  • The planning UI became flatter and less noisy, which reduces the sense of jumping between two separate mental models

Digests Got More Useful In Multi-Agent Work

Notification digests also matured from a rough batch into something more aligned with how people actually use multiple agents.

  • Digests can separate updates by agent
  • Digest emails are more readable and easier to reset back to defaults
  • Summaries are better suited to catching up on a stretch of agent work instead of replaying every raw event