September 2026

The Problem gets a place of its own

Customer evidence, the Problem it supports and the Initiative chosen to address it now have distinct places in Outcomet. September's main change is how those records connect: a decision can carry its reasons into delivery, and an Initiative can return to an earlier phase without losing its history.

In this edition

From Andy

I spent part of September trying to make my assistant write less. I'd tried changing the rules several times, with fairly modest results. Then I looked at the rules themselves. Pages of prose explaining that the answers should be short.

I think visually. Working through walls of text takes effort, even when the content is right. The useful moment was seeing that I had kept adding instructions instead of changing the format. The rule ended up as one line.

That doesn't make every piece of writing a one-liner. This monthly journal would be a strange place to argue that. But it made me pay closer attention to what the format asks the reader to do.

Andy

A Problem you can inspect before deciding

An incoming message can contain several Problems, background context and something that doesn't need action at all. September gives that analysis a review path of its own. Outcomet proposes an existing Problem to reuse or a new one to create, with the evidence alongside it. Approval remains a person's choice.

The original message is still there. Generated titles and summaries use English, while the supporting evidence keeps the customer's words and language. Reviewers can adjust severity, change classification or exclude a fragment without turning the whole message into one verdict.

Dates matter when evidence arrives later than the event it describes. A historical source date puts that evidence into the customer's chronology without rewriting the system's ingestion timestamp. If extraction stalls, Feedback details offer a restart.

This workspace is called Signals at the end of the month. Quantitative indicators have their own home in Metrics, and existing links redirect to the appropriate place.

One Initiative carries the goal and the work

An Initiative can start with a title. Its Decision brief then brings the goal, related Problems, Decisions and Learning into view. Goal directions can be prepared from customer Problems or from strategic context, and a measurable outcome and success criterion are confirmed before moving forward.

It can also move back. Reopening or returning to an earlier phase retains the historical closure receipts. When an Initiative closes, resolving its connected Problems is an explicit choice rather than an assumed consequence.

The former Bet is absorbed into that same Initiative record. Existing Bets are carried over, old URLs redirect, and hypotheses and experiments connect to Initiatives. Discovery gathers the Decision view, Initiative board, relative Planning, Validation Tree and Experiments in one module.

The Validation Tree is the visual counterpart to the brief. Customer and experiment feedback connect through Evidence and Problems to an Initiative. Details open beside the tree, and completed Changes connect to the Capabilities they affect. Planning keeps Now, Next, Later and Closed groups for the relative view of the work.

A suggestion is still waiting for your decision

Delivery puts proposed capabilities into their hierarchy, with a visible distinction between proposals and confirmed links. A suggestion can be linked, created or dismissed without clearing the other candidates still awaiting review.

That applies inside the tree as well. Linking a parent doesn't dismiss its child proposals. Exploration can consider the workspace's full taxonomy and repository context, including supported documentation matches outside the first shortlist. An empty result stays empty.

Recent GitHub Changes loaded in Delivery get missing analysis without a card click. The automatic pickup gives webhook processing ten minutes and is limited to Changes updated within the past 48 hours. Failed analysis has a manual retry action.

The pipeline follows the campaign

Campaign work continues after the initial setup. A linked playbook can be changed while reviewing what happens to the stages and the members already in them. Bluesky outbound joins LinkedIn, X and Email, with stage settings for board visibility, success, rechecks and cooldowns.

Members outside the main board appear in a Parked lane and can return to the active pipeline. Funnel reporting uses recorded stage history rather than only the column somebody occupies today. Adding contacts checks exclusions and cooldowns across campaigns.

Archive and restore keep older campaigns available for reference. Permanent deletion remains a separate action.

A source worth opening again

Stories can capture sources from custom domains and sort them by platform, shelf-life or topic. Cards carry a short preview, with a reading summary in the detail view. Visual material can contribute to enrichment, and a failed image download can fall back to the available text.

Partial results stay visible when another attempt is needed. Ordinary enrichment keeps a customized title, while a forced re-fetch can replace it. Bluesky post links can show captured likes, reposts and replies. Other platforms without metric capture, including X, remain reference links.

Also changed, with their limits intact

Feedback attachments: paste screenshots into either feedback dialog, keep a draft when dismissing it, and wait for uploads to finish before submitting.
Connected contact tools: upsert contacts and companies by ID, domain or email, with name fallback when those identifiers are absent. A contact can link to a company by domain if the company already exists in its workspace. Writes require contacts:write access.
Problem and Initiative MCP context: connected clients can inspect that context on deployments where it is enabled. It remains off by default, and clients using retired Theme tool names must update.
Nova reasoning: Low, Medium and High chat thinking replace model selection. This configuration was verified on staging. Agent changes remain subject to confirmation.
Feedback lifecycle safeguards: verified in DEV, with production release unverified. Customer close, duplicate and delete events are prevented from sending resolved emails. Deleting Feedback that supplies a Problem's last active evidence closes it as obsolete while preserving its history. Problems with another active source remain open, and evidence cleanup preserves a deleted Feedback parent's prior lifecycle state.