July 2026

July connected both ends of the loop. Themes now produce Work Items and Changes that export to GitHub or Jira in one click, and Status split from Resolution so every closure records whether something got done, turned down, or simply aged out. Backgrounds, card surfaces, and font sizes became yours. Theme of the month: Decide and hand off.

In this edition

From Andy

It started as a small pass to reduce visual noise. Two weeks later I was updating Tailwind and rebuilding elements I had never planned to touch, which is not what "reduce visual noise" is supposed to mean.

I spent years telling teams to keep refactoring separate from feature work. Then I went and did the exact thing I was preaching against, because AI makes it look cheap. It is cheaper. It is not free, and I underestimated the effort even with the AI doing most of the typing.

The day I could finally put the old menu next to the new one was a good day. That part is genuinely leaner and I was happy just looking at it.

But the visible half of July turned out not to be the important half. Underneath the new shell, closing something stopped being one bit of information, and a Theme learned how to become a ticket that leaves the building. Nobody is going to screenshot that. It is the change I would keep if I could only keep one.

Two weeks for a leaner menu and a pile of things nobody notices. I would do it again. Just planned as two weeks next time.

— Andy

A theme can become a ticket

The most manual part of the loop has been the moment after the decision. You work a pile of feedback into a theme, you agree it's worth doing, and then you go somewhere else entirely and file a ticket that describes it in worse words than the theme already did. July closed that gap. Work Items and Changes are created directly from a Theme, and a single click exports the Work Item to GitHub or Jira, so the handoff happens at the point of decision rather than in a separate tool later.

The Delivery Graph got a matching layer. Work Items render above Themes, which means you can see which tasks sit under which strategic initiative and, more pointedly, which themes have nothing underneath them at all. That second reading is the useful one. A theme with no work under it is either not really agreed or quietly forgotten, and both are worth knowing.

Tracking then maintains itself. When a linked pull request is merged or closed, the corresponding GitHub issue closes on GitHub and in the workspace, so the two records stop drifting apart the way they always do when the syncing depends on somebody's memory. For workspaces running the MCP Agent, the whole Work Item surface is available to automation: create, update, delete, search, link, unlink. An agent can maintain the task layer on your behalf, which is the point where a connected agent goes from reading your workspace to actually keeping it tidy.

Closed, and it says why

Closing something in Outcomet used to be one bit of information. It's two now, and the second one is the one that survives contact with next quarter.

Status (Open / Closed) separated from Resolution (Completed / Rejected / Obsolete), making every closure explicit and auditable
New Obsolete (Retired) resolution for items that age out rather than get decided, and "Won't Do" renamed to Rejected
Both labels customizable in Decision Vocabulary Settings
Explicit Resolve and Closure dialogs on feedback, themes, and board cards, listing valid dispositions including Validate and Test, pre-selecting the current choice, and warning about linked work before anything cascades
Resolved themes can be reopened from the Kanban board and stay open, and new themes no longer start life as Obsolete or Rejected
Status-first inbox filtering with a secondary Resolution selector for Closed items
Redesigned close and reject confirmations with a scrollable evidence list and three clear choices: cancel, this item only, or everything

Evidence you can work through

The unit of feedback stopped being the whole message. Evidence chunks became first-class this month, with their own coverage state, their own resolutions, and lists you can actually search.

Incoming emails preserve their full raw text and automatically generate AI-assisted evidence sub-chunks, so a long mail arrives already cut into review-ready highlights
Toggle a chunk as Covered inline from Feedback or Theme views, and resolve individual chunks as Completed, Rejected, or Obsolete
Parents close automatically when all children are solved and reopen when any child reopens, with simplified counts like 2/3 evidence on inbox cards and inconsistent states flagged
Real-time search, filters, and an explicit sorting dropdown (Type, Severity, Resolution) on Evidence lists in Feedback and Theme drawers
Right-click quick actions for solved state, severity, type, and resolution, with hover targets narrowed to indicator areas to stop accidental changes
Learning content folded into Feedback Themes and the standalone Learnings tab retired, so insight and action live in one place
Send a feedback item, theme card, or evidence quote into the active Agent Chat with one click, input auto-focused

One surface, made yours

Two opposite moves, same month. Everything got standardized onto one shell, and then the parts worth varying were handed over to you.

Unified topbar with the active workspace name, left-aligned pipeline navigation, and profile and feedback controls top-right
A single non-stacking Details panel that resizes and remembers its width, plus an inline split-pane for Inbox and Theme Detail with a draggable resizer
Custom canvas backgrounds per area (Inbox, Themes, Planning, Graph) and adjustable Feedback and Theme card width, styling, and positioning
Detail and list blocks placeable left, center, or right, with text kept left-aligned for readability
New Appearance setting with Small, Medium, and Large global font sizes, plus adaptive typography for small screens and high-DPR devices
Four-tier typography scale, one flat theme background, and unified card and popover surfaces across the app
KPI dashlets hidden by default with a unified signal configuration flow, so pages start quiet and show what you opted into

By the numbers

5 release weeks
22 shipping days
9 marquee modules touched: Feedback, Themes, Learnings, Discovery, Delivery, Planning, Signals, Settings, Agent
2 delivery integrations now reachable in one click from a Theme: GitHub and Jira
1 module retired: Learnings, folded into Feedback Themes
1 customization layer across Inbox, Themes, Planning, and Graph

Also shipped

Theme Detail remodel: a Brief tab with synthesized learnings, capability mapping, evidence, and comments, and a Technical tab with the Root Cause Hypothesis, sentiment KPIs, and linked work, with Alt+1 and Alt+2 hotkeys to switch.
Interactive Prose Metadata: status, disposition, and owner rendered as plain sentences with inline click-to-edit controls.
Auto-syncing themes: theme titles and the Root Cause Hypothesis update whenever a Learning Card is regenerated or auto-synthesized, and learning insights now generate without a manual refresh.
Themes Kanban rework: compact text-first layout, collapsible lists on mobile, tightened and reordered statuses, a Closed column that hides when filters exclude it, and highlighting for themes with zero evidence.
Agent Chat redesign: cardless editorial stream, right-aligned user messages, compact tool status lines, execution details collapsed by default, and clickable in-chat links to workspace entities restored.
Graph and planning controls: unified canvas styling, graphs opening centered on the current NOW point, tabbed control popovers (Canvas, Signals, Layout), and a floating ThemeFilters toolbar for in-graph search.
Roadmap layout: vertically centered triangular layout, Now/Next/Later horizons snapping to level boundaries, and interactive edge locking that survives a re-layout.
Inbox detail improvements: persistent inline layout with metadata under the title, borderless evidence and screenshot gallery, compact S0 to S3 severity codes, and date-only headers.
Feedback search restored: plain-text queries work again, with better relevance on both topic-focused and named searches.
Discovery and change details: Capability Suggestions render reliably in the Change detail panel, with expandable sections for Work Items, Capabilities, Themes, Bets, and Hypotheses.
Billing visibility: a read-only banner explaining why workspace editing is disabled, with a Manage billing shortcut for admins.
Companies context: linked contacts surfaced directly in the Company detail panel while browsing the Companies tab.
KPI presentation: clean borderless styling, restored sparklines, and clearer metric warnings.
Cross-view consistency: selection and interaction treatments standardized across Delivery, Capabilities, Discovery, Strategy, Command, Campaigns, Stories, Contacts, and Signals.
Style and drawer durability: page style copy and reset now preserve Signals settings, detail drawers keep their controls and styled backgrounds while scrolling, and legacy browser preferences load safely.
Reliability: Delivery KPI cards render dependably, card exit animations and page centering restored, and a run of layout and rendering crashes fixed across Delivery, Planning, and graph pages.

Looking ahead

June predicted that the GitHub and Jira sync would reach deeper into Delivery and that the MCP tool surface would grow past safe reads and a single write. Both happened, and they turned out to be the same story: a decision made in Outcomet can now leave the building as a tracked ticket, and an agent can maintain that layer without a person in the middle. August should push on what happens after the handoff, closing the loop back from shipped work to the theme that asked for it.

If you want a walkthrough of creating work items from themes, the new resolution model, or the canvas and card customization, reply here and I'll set up a quick demo.