Thoughtworks Technology Podcast Themes from Technology Radar Vol.33
110 snips
Oct 30, 2025 Bryan Oliver, a Principal Engineer at ThoughtWorks and expert in Kubernetes and platform engineering, dives deep into the AI-centric themes from the latest Radar. He discusses the critical need for infrastructure orchestration for AI workloads and how new Kubernetes tools can enhance scheduling for ML jobs. The conversation also covers the role of the Model Control Protocol in elevating AI agents, the transformative impact of AI across development workflows, and emerging anti-patterns that teams need to navigate in the evolving landscape of AI.
AI Snips
Chapters
Transcript
Episode notes
Optimize GPU Scheduling To Save Cost
- Even moderate GPU usage benefits from topology-aware scheduling to improve performance.
- Use platform tooling to label nodes and reduce cost by placing jobs on optimal hardware groups.
Agents And MCP Fuel Each Other
- Agents and MCP (Model Contract Protocol) have risen together, enabling autonomous tooling ecosystems.
- Context engineering and agent role descriptions (agents.md) are emerging patterns for reliable agent behavior.
Harden Agent Protocols Early
- Treat new agent protocols as additional attack surfaces and audit their auth model carefully.
- Assume bad actors will weaponize new MCP servers quickly and prepare defenses early.