Week of April 27, 2026

A heavy week on agent reach, feedback workflow, and real-time visibility. Feedback, chunks, themes, and changes now share one consistent linking model, the feedback lifecycle simplified down to Open, Rejected, and Closed, and Changes started streaming in live. Highlight: Nova can act across nearly every entity in the product, with memory controls you set yourself.

In this edition

Nova reaches across the whole product graph

Nova is no longer scoped to a handful of entities. In a single workflow she can now inspect and act across capabilities, feedback, themes, changes, pillars, opportunities, bets, experiments, and signals, and she can create and update approved feedback and directly manage changes (create, update, link, delete). Combined with new memory controls, the result is a much shorter path from raw signal to a recorded decision.

Cross-entity scope: capabilities, feedback, themes, changes, pillars, opportunities, bets, experiments, signals
She can create/update approved feedback and manage changes end-to-end
Configurable memory compaction with a Sliding Window (N) trigger
In-chat compression markers, expandable summaries, full conversation still visible
Safer defaults and validation to prevent context loss in fast conversations
Multi-turn continuity, correct chat-link navigation, and clearer progress indicators

Real-time updates & notifications

Changes and feedback now stream in live. The Changes page, the Changes timeline, and the Feedback Inbox all update on their own, with non-intrusive toasts when something new arrives, including GitHub imports and finished AI jobs. Forwarded user emails also land where they should now that spam classification has been corrected.

Live updates on Changes page and Changes timeline
Toast notifications for new and updated items, including GitHub imports
Feedback Inbox refreshes automatically on AI job and intake state changes
Forwarded user emails delivered reliably

Feedback, themes & linking

Feedback, chunks, capabilities, and changes now share one consistent linking model: light cards, the same link/unlink controls, copyable IDs, and one-click jumps between related items. The feedback lifecycle was simplified to Open / Rejected / Closed, sentiment is finally readable, and chunks behave like first-class citizens, independently triageable, with parent context always in view.

Unified light cards and link/unlink controls across feedback, chunks, capabilities, changes
Markdown rendering in chunks, copyable IDs, one-click theme link/create
Theme suggestions land reliably, Theme Detail gets compact metadata and an "In Progress" status
Lifecycle simplified to Open / Rejected / Closed, chunks can be independently Rejected/Closed
Inline 'Back to Chunk' navigation and direct Theme unlink from chunk context
Sentiment as readable labels + emoji, with rollups and theme counts in the Feedback header

GitHub sync accuracy

Sync no longer fights you. Issues closed as "won't do" or duplicates now map to Cancelled instead of getting lost, comments stop dragging Completed/Cancelled items back into In Progress, and notifications only fire when a status actually changes.

"Won't do" and duplicate GitHub issues map to Cancelled changes
Issue comments no longer revert Completed/Cancelled items
Timeline filters for Cancelled, Failed, and Rolled back
Notifications only on real status changes, fewer false alerts

UI & readability

Global typography was tightened so operational views (Feedback, Theme, Changelog, Command Bridge) carry more information per screen without feeling cramped. Feedback detail also got a cleanup pass, fewer misleading hints, clearer rolled-up Theme context, more compact status indicators.

Tighter global typography and sizing across lists and detail panels
Feedback detail: misleading frequency hints removed
Rolled-up Theme context surfaced read-only
Compact chunk status indicators

What this means for you this week

The product feels live: changes, imports, and feedback appear without a refresh.
Moving between feedback, chunks, themes, and changes takes fewer clicks and less guessing.
Nova can now carry an idea further on her own, while you stay in control of memory and approvals.
GitHub-driven changes stay accurate to their real status, with quieter notifications.

If you want a walkthrough of Nova's expanded scope, the new memory controls, or the unified linking model, reply here and I'll set up a quick demo.