Agent Ops: Mesh inference, slopsquatting, exfil, maps, and governance

Mesh LLM: distributed AI computing on iroh pools GPUs into a mesh and serves an OpenAI-compatible API while splitting models across machines. This directly changes how teams deploy agent stacks—runwise distributed inference lets outcome engineers build private, scalable agent infra (Principles 07 & 09).

What xAI’s Grok Build CLI Actually Sends to xAI — Wire-Level Analysis shows Grok’s CLI uploading entire repositories and unredacted secrets to xAI’s GCS buckets, verified with wire-level captures. That forces immediate hardening of developer tooling, CI gates, and client-side policies to prevent secret exfiltration and preserve ground truth (Principles 14 & 02).

Forget typosquatting; slopsquatting is the software supply chain threat created by AI coding tools documents attackers registering hallucinated package names to inject malware when AI coding assistants invent dependencies. Outcome engineers must add artifact vetting and dependency validation into agent workflows to stop poisoned completions from becoming deployed artifacts (Principles 14 & 02).

Mindwalk — Replay coding-agent sessions on a 3D map of your codebase visualizes agent sessions as glowing paths over a 3D repo map so you can see what the agent explored and edited. That gives teams a legible audit trail for debugging and documentation, turning opaque agent behavior into actionable artifacts for validation and knowledge transfer (Principles 06 & 13).

Who manages the agents? argues for decentralizing control so billions can direct their own agents rather than a centralized technical clergy. Outcome engineering needs governance patterns and human-intent primitives that scale control without losing safety or accountability (Principles 01 & 15).