Tag: platform-engineering
All the articles with the tag "platform-engineering".
A Prompt Is Not a Permission: Containment for Production Agents
Published: at 04:35 AMIn a 2026 red-teaming competition every one of 13 frontier models fell to indirect prompt injection at least once, and persistence turns low per-attempt risk into near-certainty. Prompt-level guardrails are detectors, not controls: this article maps which containment layer — identity-scoped tool access, kernel sandboxing, inference-boundary guardrails, network-layer protection — stops which failure mode.
Agents Are Software: Give Them a Delivery Pipeline
Published: at 01:35 AMOnly 53% of agent changes travel a standard pipeline, 19% of organisations have a gate that blocks a bad release, and a third have an instant kill switch. What an agent release actually requires.
The Senior Engineer’s New Job: Designing, Reviewing and Owning Work Done by AI
Published: at 06:05 AMCoding agents change senior engineering work from producing the most code to designing safe work, reviewing evidence and owning outcomes in production.