Issue 002 · Published editionGet the next issue
Friday, 21 August 2026DubaiIssue 002

The weekly decision brief for AI product people

What changed.
What deserves a decision.

4

signals, selected for a working PM. About eight minutes, start to finish.

Get the next issue
DISCUSSAgent capability distributionPrimary vendor release notes

Agent plugins went 1.0. Distribution is becoming the durable layer.

One package can carry skills and MCP configuration across VS Code, Copilot CLI, the Copilot SDK, and the Copilot app. Figma moved in the same direction with Community skills and MCP access to Weave tools.

Decision affected

Decide where repeatable agent capabilities are packaged, discovered, versioned, trusted, and governed.

My read

Changing models is becoming easier than changing how capabilities are distributed. Teams should decide whether repeatable workflows stay inside the product or become portable capabilities with a version, trust model, and owner.

Read the source

Useful enough to keep?

Get the next issue by email
02

The rest of today's signals

Filter by the decision sitting on your desk.

02DISCUSS
Delegated agent identityLinear and GitHub13 Aug

Linear and GitHub wired agents into existing permissions.

Linear now verifies supported Anthropic-agent users through Okta and applies their existing Linear permissions. GitHub added controls to restrict plugins and marketplaces and allow or block MCP servers.

Decision affected

Decide whether agents inherit the requesting user's identity and permissions, and how least privilege, auditability, and admin control work.

My read

The enterprise pattern is deliberately familiar. Agents should inherit existing identity and permissions with narrower delegation. Adding AI first and authorization later leaves two control planes to reconcile.

03MONITOR
Persistent agent memoryGitHub11 Aug

Copilot memory is now product state.

GitHub released Copilot memory for JetBrains. It can retain and recall project details or preferences across agent chat sessions, and users can control it in settings.

Decision affected

Define what memory stores, its scope and visibility, how users correct or delete it, and how permissions affect recall.

My read

Memory looks like convenience and behaves like stored product data. A setting is a start, but the lifecycle needs correction, deletion, scope, permissions, and support for wrong or sensitive recall.

04DISCUSS
AI product economicsGitHub11 Aug

AI cost is becoming attributable.

GitHub's AI usage reports added a per-model breakdown for input, output, cache-read, and cache-write tokens alongside AI credits consumed.

Decision affected

Connect model and token costs to workflows, successful outcomes, customer segments, and routing choices.

My read

The useful shift is explainability, not a better billing export. Token accounting is the raw material. Cost per successful workflow is the product metric.

03

The method

Less news. More judgment.

This is a filter for people who have to make product decisions, not a leaderboard for whoever launched most recently.

01

Start with primary sources

Release notes, incident reports, research and regulators come first. Social posts help find stories but do not prove them.

02

Name the decision

A story earns space only if it can change a roadmap, product experience, operating model, price or risk decision.

03

Keep claims attached

Vendor benchmarks remain vendor benchmarks. Early research stays early. Confidence should be visible, not implied by polished copy.

04

Stop at six

If everything is recommended, nothing was curated. The edition ends while there is still something worth remembering.