Node 20 left GitHub Actions: check the actions, not just your app
GitHub's Node 20 runner retirement calls for two different checks: the JavaScript actions your workflows use and the Node version your project tests.
The library
Knowledge that
moves work forward.
Browse by the kind of help you need, or search the full-text index.
12 field notes available
A fresh desk is coming
No field notes are published yet.
There is nothing to browse right now, but the library is ready for the next useful note.
How the desk works
Audit a Mac app for Intel-only binaries before Rosetta goes away
A small Bash check that finds Intel-only code hidden inside a macOS app bundle before the general-purpose Rosetta deadline.
Connect an AI coding tool to Atlas without sharing credentials
A practical guide to MongoDB Atlas App Connections, delegated MCP access, and the boundary between interactive AI clients and automated agents.
Test a live voice agent before you trust the demo
Use ADK's live evaluation loop to test turn-taking, tool calls, handoffs and spoken conversations before shipping a voice agent.
When privacy and AI safety stop being opposites
What OpenAI's Private Safety Processing preview teaches us about building AI workflows that protect data and remain observable.
Choose the right LLM for the job
A practical guide to choosing fast, general, reasoning, long-context, multimodal and specialist models by workflow phase.
Herdr: one terminal for the whole herd
A practical first look at Herdr, a terminal-native multiplexer for keeping coding agents, shells and long-running work together.
Spend tokens on decisions, not noise
A practical look at the four tools I use to keep AI coding sessions focused: Caveman, Ponytail, RTK and Graphify.
Agent brief template: give the work a finish line
A copyable brief for giving an agent a clear goal, useful context, and a definition of done.
Watch: Agentic AI — workflows vs. agents
A short visual explainer for choosing a fixed workflow, an agent, or a combination of both.