Daily Edition

illmethinks — September 14, 2026

Agent interfaces are learning to show the work behind the answer: which machine is in view, which child is waiting, and which settings belong to one task. That visibility creates its own obligations. Traces can carry private text, imported tools need explicit consent, and a design decision is not the same as an enforced boundary.

OpenClaw proposes a full-workspace machine view with explicit desktop takeover

OpenClaw’s proposed Systems workspace moves machines into a contextual sidebar and gives the selected desktop the main pane. Access remains view-only until a person takes control; offline targets remain inspectable, and connection records are not presented as execution history.

OpenClaw proposes showing whether a child is working, waiting or still delivering

OpenClaw’s subagent proposal separates execution from result delivery and names what a child is waiting for. Selected-task waits, explicit incoming-message intent and descendant settlement…

ZeroClaw records an identity architecture whose full rollout is still unfinished

ZeroClaw’s proposed ADR separates credential verification, principal resolution and permission enforcement. It records acceptance conditions for sessions, private memory and recovery, but…

Hermes proposes task-level reasoning controls alongside desktop model selection

Hermes Desktop would let users set reasoning effort separately for auxiliary tasks, matching the runtime’s existing configuration keys. The proposal distinguishes leaving an override alone,…