Guide / Evidence
Run a workflow pilot that can answer a real question
A small scorecard can compare baseline effort, output quality, and rework without pretending a handful of jobs is a controlled study.
3 min readTHE WORKING NOTEBOOK / PRODUCTIVITY FIELD NOTES
Make the work
hold up.
AI can produce a draft. The harder part is a reliable handoff. Start with one task, account for the setup and check what leaves your desk.
Pick the next handoff
All seven pathsStart here
All practical guidesGuide / Research & notes
Turn a research question into an evidence table that survives disagreement
A small source-row protocol keeps the question, claim, passage, and unresolved conflict visible before anyone writes the summary.
3 min readGuide / Automation & agents
Prevent duplicate records before an automation goes live
A stable business key, an idempotency record, and a conflict path are safer than hoping repeated events never arrive.
3 min readGuide / Writing & documents
Review the diff for altered facts, not just improved prose
A version comparison catches changed numbers, dates, qualifiers, and links that a smooth reread can miss.
3 min readGuide / Coding & data
Reconcile a spreadsheet before writing the story around it
Totals, keys, row counts, and granularity checks should pass before a chart or narrative turns an error into a conclusion.
3 min readGuide / Setup & friction
Run a vendor exit drill before the exit is urgent
Export is only the first checkpoint: map dependencies, inspect formats, restore representative work, and disconnect the old system last.
3 min readTHE REVIEW IS PART OF THE JOB
Keep the source beside the answer.
The research path moves from a clear question to an evidence table, checked extraction and a library someone else can open.
Follow the research path
At the workbench
Open the toolsKeep the paper trail
100 archive recordsArchive / Evidence
A 1993 paper named the computer productivity paradox
Brynjolfsson's paper defines the productivity paradox and tests four explanations for it, not a resolved verdict.
3 min readArchive / Evidence
METR's benchmark tracks AI task length, not messy real work
METR's paper found the length of software tasks AI agents can finish has doubled roughly every seven months, with real-world caveats attached.
3 min readArchive / Research & notes
Glean's search tool inherits whatever permissions were already wrong
Glean's own security page says it enforces existing access rules rather than fixing them, and pricing is not published.
3 min read