Figma Specs puts the product brief next to the design
The new Markdown-based document format brings decisions, frames, and implementation context onto the same canvas. It is rolling out in open beta on paid plans.
The software signal
Latest news.
The ideas, launches, and shifts moving software forward.
Notion wants to fix the context gap between your team and its AI agents
Column permissions and rebuilt synced databases address a practical question: what should an agent know about your company, and what should stay out of reach?
GitHub Copilot’s October model retirement is a workflow maintenance task
GitHub scheduled several models for removal on October 19. Teams should check saved model choices and organization policies before assuming replacements are available.
Vercel Daemons brings long-running background workers into the project
The public beta supports processes that pull work from queues and other systems, extending the platform beyond workloads triggered by an HTTP request.
Cursor cloud agents can now share interactive canvases
The October 16 update adds shareable canvases and a beta Environments API. It gives teams another way to inspect an agent’s output beyond a text summary.
Figma’s code layers enter open beta with prototypes on the canvas
Teams can bring supported prototype code into the shared design canvas and explore alternatives together, instead of passing around separate development sessions.
Linear adds stacked pull requests to Diffs, keeping related changes together
The October 8 update groups related pull requests and exposes review, CI, and merge status, helping reviewers follow work split across several dependent changes.
Claude Haiku 5.5 focuses on the repetitive work inside AI applications
Anthropic positions its new small model for summaries, classification, and other high-volume tasks. The useful question is which part of a workflow needs that model.
Figma’s agent targets the design system chores teams keep postponing
Component documentation and library upkeep are part of Figma’s agent pitch. A bounded maintenance task is a useful place to evaluate what the assistant actually changes.