Commandline - 2

2026 (7)

August (5)

commandline: Stop restricting the agent. Start restricting its environment.

Microsoft Command Line Blog

2026-08-21

Azure SRE Agent argues that AI safety is better achieved by constraining the runtime environment rather than the agent itself. This approach aims to preserve autonomy while reducing risk compared with human review or standing credentials.

commandline: How we built ThinkingBox to measure whether agents finish the job

Microsoft Command Line Blog

2026-08-19

ThinkingBox splits the execution framework from the benchmark package so each can be updated independently. The article frames this as a way to measure whether agents actually finish tasks.

commandline: Watch TypeScript creator Anders Hejlsberg on The Peterman Pod

Microsoft Command Line Blog

2026-08-17

After TypeScript 7.0, Anders Hejlsberg discusses porting TypeScript from JavaScript to Go, along with the benefits and drawbacks of agentic AI for coding. The post is a podcast announcement rather than a technical changelog.

commandline: Who spent all the tokens? Real-time, run-scoped cost control for AI agents

Microsoft Command Line Blog

2026-08-13

TokenOps is presented as a run-scoped cost control mechanism for AI agents that enforces a shared budget in real time. The stated goal is to stop runaway runs before costs accumulate and invoices arrive.

commandline: 🌿 Agent Terrarium: Bringing AI agents to life

Microsoft Command Line Blog

2026-08-07

A weekend open-source desktop companion app, Agent Terrarium, explores a more interactive, less purely transactional way to work with AI agents.

July (2)

commandline: Tool search: Finding the right tool at the right time

Microsoft Command Line Blog

2026-07-29

The article says the value in adding more tools to AI agents comes from surfacing the right tool at the right time, instead of exposing every tool upfront. It focuses on discoverability as the key problem.

commandline: From research to reality: An interview with Microsoft VP of Security Research Taesoo Kim

Microsoft Command Line Blog

2026-07-27

Microsoft interviewed Taesoo Kim, VP of Security Research and head of Security FORGE Labs, about codename MDASH and the lab’s work on security research. The interview highlights a multi-model, agentic approach as a potential direction for improving security capabilities.