maximal

Changelog

What changed in the private beta.

Changes that reached private-beta workspaces, most recent first. Planned work is on the roadmap.

25 September 2026

Approving or denying in Slack updates the request at once, and a decision that cannot be carried out is posted back to the channel. Lambda deploys recorded in CloudTrail are tied to the alarms they cause. An AWS connector Max can no longer read is flagged on the Connectors page and in Slack. Changing a workspace's autonomy mode takes effect within seconds, including for incidents already waiting on approval. An incident whose audit trail did not fully reach the database no longer feeds the learning pipeline.

24 September 2026

Pending approvals and remediations in progress survive restarts and deploys: one engine instance leads at a time, hands over cleanly, and resumes or escalates interrupted work. Remediations act only on resources named in the incident's own signal, a rollback is pinned to its exact target and refuses an approval that has gone stale, and verification watches for recovery for up to the contract's window rather than about 75 seconds. Escalations and automatic remediations are posted to Slack. The Connectors page lists only integrations that are available today.

18–22 September 2026

EC2 status-check alarms match their contract. Light and dark themes follow your system setting. Incident and audit history survive a restart, audit records are hashed exactly as they are stored, and verification tells a legacy chain fork apart from tampering. Incidents are labelled with their resource type and a readable name.

Request accessView roadmap