← Latest Update

Agents at Work: Routines, Infra & Platform Risks

Claude Code turned every engineer into three. Now companies need more product thinkers. Anthropic’s Claude Code and Code Routines multiply individual engineering throughput, shifting the bottleneck from implementation to product thinking and orchestration. Outcome engineers must set clear success criteria, acceptance tests, and agent contracts so multiplied output produces reliable outcomes (Principle 04).

AMD Strix Halo RDMA Cluster Setup Guide. The guide shows a two-node Strix Halo cluster with RoCE v2 delivering ~5µs RDMA latency for tensor-parallel vLLM inference, enabling truly interactive multi-APU models. Outcome engineers can use these patterns to build low-latency, production-grade islands for agentic systems and plan deployment topologies accordingly (Principle 07).

Adrafinil — keep a lid-closed Mac awake only while agents work. This macOS utility keeps machines awake only while agent sessions run, releasing sleep immediately when the last session ends and enforcing thermal safety and privilege constraints. For teams prototyping local agents, it provides predictable agent lifecycle and resource governance without manual intervention (Principle 09).

Google told Meta it couldn’t supply all Gemini capacity, delaying Meta’s AI projects. Limited access to Gemini compute forces Meta to postpone internal AI work and exposes compute as a strategic bottleneck. Outcome engineers must design for compute scarcity: optimize models, implement graceful degradation and multi-vendor orchestration to keep pipelines resilient (Principle 12).

Sources: Salesforce staff worry Anthropic’s Claude Tag could cannibalize Slackbot and give the AI firm more leverage over the enterprise software industry. Anthropic’s Claude Tag integration for Slack raises internal fears that it will displace Slackbot and increase vendor leverage inside enterprise workflows. Outcome engineers embedding agents into collaboration platforms need explicit gate controls, exportability, and anti-lock-in strategies to preserve composability and control (Principle 15).