Week of May 25, 2026

A heavy week on discovery and planning. The old Pipeline gave way to an interactive Validation Tree, experiments moved onto a five-stage Kanban, and a Methods Library of 31 templates made structured experiments quick to start. Work Items arrived too, and planning gained a persistent shared view. Highlight: the Validation Tree turns discovery into a graph of Opportunities, Bets, Hypotheses, Experiments, and Evidence.

In this edition

The Validation Tree replaces the Pipeline

Discovery used to live in a flat Pipeline. This week it became a tree. The new Validation Tree visualizes Bet > Hypothesis > Experiment > Evidence relationships as an interactive graph, with radial progress indicators on each node, click-to-open slide-out details, and drag-and-drop to rearrange the evidence flow. It centers on load, keeps your pan and zoom when you open a detail, and no longer rearranges itself out from under you.

Experiments got a real workflow underneath. The new Experiments Kanban moves work across five validation stages, and the Methods Library gives you a running start instead of a blank form.

Validation Tree with radial progress, interactive nodes, slide-out details, and drag-to-rearrange evidence
Experiments Kanban with five stages (Backlog > Designed > Running > Analyzing > Concluded), drag-and-drop status, and richer cards (objectives, method and type badges, linked bets and hypotheses, cost and duration, learning counters)
Many-to-many linking between Bets, Hypotheses, and Experiments for flexible modeling
Methods Library: searchable catalog of 31 experiment templates with previews and target-layer badges, plus a two-step creation wizard with a split-pane preview and editor
Compact floating Bets panel grouped by Opportunity, with collapsible groups, search, progress bars, and persisted selection across navigation
Standardized non-modal slide-out sheets across Opportunities, Bets, Experiments, Hypotheses, and Capabilities, with inline title editing and safe Unlink replacing destructive delete

Planning and delivery get steadier

Planning picked up a persistent collaborative view. Shared view state, timeline projection controls, and a visible Now boundary let a team orient ongoing versus future work and come back to the same perspective. Roadmap visuals now surface planned placeholders and learning context for delivery forecasting, so planning decisions have more to stand on.

Persistent Theme Planning view with shared state, timeline projection, and a visible Now boundary
Delivery and Changes graphs surface planned placeholders and learning context for forecasting
New Work Items context: track issue-like work separately from change records, link directly to Changes and Themes, and open, link, or unlink related artifacts with visible counts
Delivery workspace navigation aligned across Timeline, Work Items, and Changes

Feedback triage gets faster signals

Theme suggestions now reach further, automatically surfacing and grouping a much broader set of open feedback into candidate themes and linking the underlying items. Inbox cards carry severity at a glance, and linking failures tell you exactly what went wrong instead of failing quietly.

AI theme suggestions group a broader set of open feedback and link the underlying items automatically
Severity badges (S0-S3) on feedback inbox cards for instant prioritization
Explicit error toasts when a linking operation fails, and a clear AI-suggestion error in the Add Theme dialog so you can proceed manually

Imports that respect your permissions

GitHub and Jira imports got more honest. Imports now respect installed GitHub App permissions, report permission gaps without failing items that already imported, and keep future-themed items created from linked issues visible. Change and Work Item details render GitHub and Jira markdown, so descriptions read as formatted content for easier handoffs.

Imports respect installed GitHub App permissions and report gaps without failing already-imported items
Future-themed items created from linked issues stay visible
Change and Work Item details render GitHub and Jira markdown

Nova and the agent roster

The agent roster is discoverable now, with specialist profiles for Nova, Polaris, Vega, and Nexus, per-agent customization, and Polaris selectable directly from the launcher. Nova's tooling got faster and cheaper too, and agents can now manage a theme's lifecycle when asked.

Discoverable specialist profiles (Nova, Polaris, Vega, Nexus) with per-agent color and layout, plus Polaris selectable in the launcher
Large tool outputs compacted before storage and resend, with stable tool definitions and extended caching for snappier, more reliable conversations
Agents like Nova can update a theme's lifecycle status and disposition when instructed

Polish, performance, and platform

A cleanup pass across visuals, speed, and stability. Saved planning settings and graph coordinates now survive stale or invalid values, the feedback capability graph ships with a lower-latency default, and the usage widget got more useful for watching costs.

Resilient saved planning settings and graph coordinates that won't corrupt your view when switching tabs or filters
Lower-latency default preset for the feedback capability graph, and faster, more precise change lookup
Usage widget with compact input and output breakdowns, cache reads and writes, Normal and Verbose views, approximate tool payload size alongside billable usage, and more accurate per-section cost calculations
Visual consistency across Discovery and Strategy: unified icons, standardized cards, consistent colors for Feedback and Learnings, and a single background click to close a slide-in pane
Safari private-mode fix so state persistence no longer errors, plus avatar handling, keyboard accessibility, and quieter empty states

What this means for you this week

Map discovery as a real graph: walk Opportunities, Bets, Hypotheses, Experiments, and Evidence instead of a flat list.
Run experiments on a Kanban and start from 31 proven templates instead of a blank form.
Plan as a team with a shared view and a clear Now boundary that survives reloads.
Track issue-like Work Items that link straight to Changes and Themes, with imported markdown rendered cleanly.
Import from GitHub and Jira with clear permission and error reporting instead of silent failures.

If you want a walkthrough of the Validation Tree, the Methods Library, or the new Work Items, reply here and I'll set up a quick demo.