Your drafts, source files, prompts, transcripts, workflows, and output history should stay understandable and movable.
Own your AI work.
Practical local-first AI, private workflows, and automation tests for people who want the work to stay theirs.
Useful AI keeps the operator in charge.
Privacy is not a feature checkbox. It shapes where models run, what gets uploaded, and how tools are wired together.
A workflow is not done until it produces the file, decision, draft, edit, or handoff it was built for.
Tool claims get tested against friction, failures, costs, constraints, and the parts that actually change the work.
Keep the work close
AI should help you make, decide, compare, write, edit, and ship without turning your files, habits, and private context into someone else's asset.
Read the notesBuild workflows that do the job
A useful workflow has inputs, steps, outputs, and a reason to exist. If it only makes a demo look clever, it does not belong here.
Browse workflowsChoose where the work runs
Local AI is a practical default for privacy, cost, speed, and control. Cloud tools still have a place when they earn it.
Open the labThe marketing page is the promise. The workbench is the proof.
Private workflow kits
Repeatable ways to move from raw notes, files, and context into work you can actually use.
Local AI notes
Device-local model findings, setup friction, useful limits, and hardware-aware recommendations.
Automation tests
Small systems that save repeated effort without hiding what they are doing or where data goes.
Decision notes
Plain calls on what is useful, what is not ready, and what needs more evidence before it gets recommended.
Go there when you want the workbench.
The base domain explains the point of view. The AI subdomain keeps the articles, workflows, local lab, and review material where they can keep changing.
Notes from the bench
Short notes when a workflow, local AI test, or automation finding is useful enough to keep. No drip campaign.
Looking for the library right now? Visit the AI workspace.