IDEAS, DECISIONS & LESSONS
Notes on AI, architecture and building useful products.
85 articles
05 Sep 2026 · 5 min read
ai-agents
CrowdStrike shipped an identity provider built specifically for AI agents at Fal.Con 2026 — continuous, risk-aware authorization instead of static tokens. A technical lead's take on the pattern, and why it's different from the NHI advice everyone's already writing.
Read more05 Sep 2026 · 5 min read
agentic coding
Anthropic shipped a beta that lets you add or remove tools mid-conversation without invalidating the prompt cache. Why that matters for long-running agents, and what I changed in my own agent's tool-loading logic after reading the spec.
Read more03 Sep 2026 · 5 min read
ai-agents
JetStream shipped Clearance this week — a reasoning engine that authorizes every agent tool call against a versioned policy before it runs, not after. Here's how it differs from network-layer MCP detection, and what I'm taking from it for my own agent gateway.
Read more01 Sep 2026 · 5 min read
ai-agents
Claude Sonnet 5's introductory pricing and GitHub Copilot Studio's promotional credits both expired on the same day. Here's the real math on what changed, why it happened simultaneously, and how to budget for the next one.
Read more01 Sep 2026 · 6 min read
ai-agents
VS Code 1.135 shipped the Agent Host Protocol with an experimental /rubber-duck command — a second AI model critiquing your primary agent's plans and code in real time. Here's what it changes about how teams should structure AI-assisted review.
Read more30 Aug 2026 · 4 min read
ai-agents
Anthropic's Aug 27 research preview of MHS gives AI agents a standard interface to operate lab and manufacturing hardware. As a Tech Lead who has wired MCP servers to internal tools, here's why this is MCP's shape applied to physical devices — and what breaks when the tool call controls a centrifuge instead of a database.
Read more26 Aug 2026 · 4 min read
ai-agents
Ollama v0.33.0 shipped a UI-driven gateway that lets Claude Desktop run Qwen, DeepSeek, and Kimi models locally — no MCP server, no Anthropic feature, entirely Ollama's own proxy. Here's how it works, the tradeoffs in tokens/sec and tool-calling reliability, and when it's worth flipping on for a team.
Read more26 Aug 2026 · 4 min read
ai-agents
Docker shipped microVM-isolated sandboxes that plug directly into GitHub Agentic Workflows, giving AI coding agents a full private Docker daemon and sudo without touching your runner's secrets or host. Here's the architecture, a working YAML example, and where the rough edges still are.
Read more25 Aug 2026 · 4 min read
ai-agents
Microsoft's new agent benchmark stopped grading transcripts and started checking database state instead. A 65% first-try success rate that collapses to 25% over 20 tries should change how technical leads think about agent reliability testing.
Read more
