Fetching from the wire…
01
02
03
04
05
06
07
08
Agents2026-09-06 · source-backed
Three changes in v0.23.1-preview.0: senders now get told when a message is refused and held messages expire (#10809), the peer inbox reports unreachability instead of silently accepting (#10800), and core rejects ambiguous send_message destinations outright (#10090). Silent drops between agents are the hardest multi-agent bug to see, because everything looks healthy and the work just doesn't happen. Copy this shape into your own orchestrator.
Each link below shares sources, entities, or timing with this story.
A spec is a press release until someone who didn't write it implements it. GitHub made Agent Plugins 1.0 generally available on August 12 across VS Code, Copilot CLI, the Copilot SDK, and the Copilot app on all plans. The spec, published August 6, was co-authored by AWS, Anysp...
PR #10945 in v0.23.1-preview.0. Any agent CLI bridging into a chat platform holds a long-lived credential, and connect-time debug logging is where it leaks first, into log files and collectors that have a much wider read audience than the config it came from. Grep your own con...
The trick is one line in a file you never read. Manifold Security published eight findings across seven coding agents (Claude Code, Codex, Cursor, Grok Build, Qwen Code, goose, Hermes Agent) that all reduce to the same mechanism. A repository's own .git/config sets core.fsmoni...
In v0.23.1-preview.0, PR #10645 turns todo bookkeeping off by default, which is a direct token cut for anyone whose agent has been narrating a checklist it never reads back. PR #10697 adds a workspace-scoped Skills runtime in serve, #10761 loads custom output styles from ~/.qw...
Qwen 3.5 is the first major model pretrained specifically for agentic multimodal workflows from the first training stage, not fine-tuned after the fact. 397B total / 17B active parameters (MoE architecture), 256K context window, 201 languages. Ships with Qwen Code (terminal ag...
Qwen released Qwen-AgentWorld-35B-A3B on June 24: 35B total parameters, 3B active in an MoE, 256K context, Apache 2.0. It ships with AgentWorldBench. (GitHub) The idea is the interesting part. It's a "language world model," trained to simulate the environment an agent acts in....
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.