Week of September 28, 2026

Plan with Initiatives and compare competitors

Initiatives bring the former Bet workflow into the same record as the goal and linked Problems. Competitor Intelligence adds a separate place to assess market alternatives, while Product Context gives native agents the workspace's own product brief.

In this edition

Keep the goal and evidence with the Initiative

Existing Bets become Initiatives, with redirects from their old URLs. Discovery gathers Decisions, Initiatives, Planning, the Validation Tree and Experiments. The relative Planning view keeps Now, Next, Later and Closed groups.

The Validation Tree connects customer or experiment Feedback through Evidence and Problems to an Initiative. Open details in place, and follow completed Changes to their impacted Capabilities.
Goal directions can start from the title, capabilities and workspace strategy before Problems are linked. Suggestions need confirmation. Intent and custom goal drafts stay editable, and phase changes persist in the board and workspace.
Archive an Initiative without changing its lifecycle phase or losing Problems, evidence and Decision history. Restore it later, or use Include archived within the chosen Open, Closed or All view. Agent searches include archived records by default.
Forward actions and relationship pickers remain accessible across Brief and Technical tabs.

Review a Problem before closing it

Duplicate review puts the candidate and existing Problem side by side, with readable evidence and a comparison verdict. The reviewer chooses whether to link the existing Problem or create a separate one. Comparison results can be reused until the candidate changes.

Linked Changes can also be checked against a Problem's active evidence, with an explanation for each chunk. Adopt the proof note into the closure dialog, or verify open Problems in an Initiative's closure checklist. Fully verified Problems are selected and proof notes enter the conclusion. The user still confirms the closure.

The Feedback lifecycle safeguards were verified in DEV. Deleting Feedback that supplies a Problem's last active evidence closes that Problem as obsolete while retaining its history. Another active source keeps it open. Deleted Feedback retains its prior lifecycle state, and customer close, duplicate and delete events are prevented from sending resolved emails.

Start agent conversations from the product brief

Product Context turns the workspace's website into a report of its value proposition, audience, capabilities and pricing. Nova, Polaris and Vega receive that overview at the start of a conversation and can request deeper detail when needed. Competitor website content stays separate from this grounding.

Analysis shows phases, elapsed time and discovered pages. It can resume after a refresh or be cancelled and restarted. The report handles structured proof points and differentiators, and its database setup was applied to staging.

Inspect competitor positioning and movement

The default Map sits alongside List and Risks. Profiles combine assessments, positioning and linked companies or contacts. Rubric versions define the scoring, with new dimensions marking older assessments for rescoring. Workspace admins manage the rubric, and recording a conversation agent's assessment requires approval.

Website scans preserve dated snapshots. Their pitch separates the vendor's description from Our read and Against us. The brief includes source citations, product-model descriptions and waitlist recognition, with a full-page reading view.
Literal value proposition quotes are checked against website text. Unverified claims remain visible. Dead domains, temporary failures and bot protection have distinct scan outcomes, while stalled scans can recover instead of remaining stuck.
The scan card shows live milestones and an expandable step history. This progress update was applied to staging.
Actual assessment dates drive the Map and its comparison paths. Contours and 3D relief use the full selected snapshot, so search does not change the terrain or measured positions.
Camera state survives theme, filter and elevation changes. Marker and plate customization is retained, and dense maps can omit labels while leaving every marker inspectable. Clearing a comparison date stays cleared across view changes and reloads.

Give connected clients explicit write access

Connected MCP clients gain more ways to work with existing records. These writes have different boundaries from the native agents' approval flow.

Connected clients can enroll Campaign members, log touchpoints and update stages or next actions when contact write access is approved. Configured stages are discoverable, history can be paged, and duplicate enrollment is reported. Stage changes derive status when omitted, while next-action edits preserve exclusions.
Connected clients can remove automation-created touchpoints only. Human-created touchpoints remain protected. Authorized members can remove a mistaken timeline touchpoint in the app after confirmation. The Campaign contracts were verified on staging and require coordinated app/MCP deployment.
Competitor connections can search assessments and risks, inspect profiles and update records. Reads and writes are disabled by default. Writes require competitor access, while rubric changes need an admin and a separate switch. The database access and relationship corrections were verified on staging and production.
Initiative connections can create and update records for an admin or member with approved Initiative access. Context and writes are disabled by default, and the companion contracts were applied to DEV. Creation is not idempotent, updates do not close Initiatives, and relationship writes remain bounded. Separate verification proposals require another default-off permission that the app consent flow does not grant.

Other changes in daily work

Delivery picks up missing processing for recently updated GitHub Changes after giving normal webhook processing time to finish. Failed analysis has a manual retry. Selecting a card preserves existing capability suggestions.
Saving a manual Change refreshes an expired authentication session and reports the specific reason if saving fails.
Feedback dialogs accept pasted screenshots, retain dismissed drafts and wait for uploads before submission. Historical source dates remain editable in details. Global Feedback History reflects rejected or deleted items and opens screenshots at full size.
Signals Inbox evidence and comparison notes render Markdown. Removing the selected item closes its detail pane.
Customer Feedback is named Signals, while quantitative indicators move to Metrics. Existing links redirect, and workspace labels can be configured for eleven entity types.
Nova chat offers Low, Medium and High thinking levels. This configuration was verified on staging, and native agent changes still require confirmation. Agent Preferences use one header save action that tracks changes.

Complete change list

Initiatives replaced the former Bet workflow, gained guided phase changes, archiving, closure history and clearer goal, evidence and delivery context.
Problem review gained side-by-side duplicate comparison, reusable verdicts, evidence checks and user-confirmed closure support.
Product Context began grounding native agent conversations in the accepted workspace brief, with resumable and cancellable website analysis.
Competitor Intelligence gained profiles, assessments, dated website snapshots, comparison views and readable scan progress.
Connected clients gained guarded Campaign, Competitor and Initiative writes with their default-off, role, approval and coordinated-deployment limits preserved.
Daily work gained safer Feedback handling, clearer Signals and Metrics naming, better Markdown rendering, recoverable Delivery analysis and configurable agent thinking levels.