airoweb / field notesIndependent thinking. Practical AI.
Notes on making AI useful at work.
Workflows, tools, MCPs, and the decisions in between.
Type below. Or just start exploring.
Curiosity is a good starting point
You can just play, too.
Latest transmissions
29 posts8 min
Jev belongs in the decision layer, not the writing layer
AI infrastructureNotchFuel puts AI usage in the MacBook notch
AI infrastructurePut the AI pentester in a test cell
AI infrastructureAsk workers where the AI moved the work
Company adoptionThe agent updated the record. The notification never left.
AI infrastructureAI governance needs decision rights, not another committee
Company adoptionA running flag is not a lock for your agent queue
AI infrastructureStop measuring AI literacy by course completion
Company adoptionValid JSON is not permission to act
AI infrastructureMeasure the AI workflow, not how often people open the tool
Company adoptionGive your agent a key, not the keyring
AI infrastructureDesign the AI workflow's off switch before you need it
Company adoptionAssume your agent will run the workflow twice
AI infrastructureYour agent logs are a mirror: a self-audit from AI session history
AI workflowGive coding agents a workbench, not a workstation
AI infrastructureDo not send every AI request through the same approval queue
Company adoptionAgent workflows need an evidence trail before they need more autonomy
AI infrastructureReview AI workflows again when the workflow changes
Company adoptionKeep SOUL.md out of the agent junk drawer
AI infrastructureTreat MCP servers like production APIs
AI infrastructureTreat Multica skills as operating knowledge, not prompt storage
AI infrastructureDecide which agent tool calls need human approval
AI infrastructureReview AI pilots before they become business process
Company adoptionFable 5 is back. Do not build as if access is permanent.
AI infrastructureA better prompt for building an Obsidian vault as agent memory
AI infrastructureBefore buying AI tools, write the operating model
Company adoptionA quick review checklist for recurring AI workflows
Workflow reviewMCP catalogs can wait. The workflow cannot.
AI infrastructureKeep exploring the archive.