The research desk / Expanded edition
More signal.
Better decisions.
New primary-source-informed field guides, the complete original report, and a much wider tool landscape. Read the method behind the shortlist.
Research added / 16 September 2026
The questions behind the purchase.
Product documentation, independent evidence and editorial judgment—with the boundaries visible.
8 guides
Choose notes you can leave with
Local files, structured workspaces and the export test to run before moving your archive.
Research practiceBuild research with a source trail
Separate discovery, extraction and citation management so a fluent summary never becomes your evidence.
Team operationsMeeting notes without invisible trade-offs
Evaluate capture, retention, sharing and summary accuracy as separate decisions.
Automation designBuild automation that can say no
Approval gates, retries and failure paths belong in the first version—not the incident review.
Evidence updateWhat AI coding evidence can—and cannot—tell you
Why METR’s 2026 follow-up changes the measurement conversation without erasing its earlier result.
Buying decisionsThe real cost of an automation
Translate tasks, credits, replays and review into a budget that resembles your actual workload.
Builder decisionsFrom a convincing prototype to a dependable app
What to test after an AI or no-code builder makes the happy path look effortless.
Stack designBuild a small stack that stays small
Choose fewer systems of record, define handoffs and give every extra subscription a job to prove.
No guides match that search. Try a broader topic.
The foundation / 14 September 2026
The original twelve-chapter report.
The map before the shortlist
Assistants, copilots, agents—and what makes each useful.
Find the category that fits
A map of the jobs AI and productivity tools actually do.
The product landscape
The report’s comparisons, plan caveats and trust snapshots.
Does AI actually save time?
What studies measure, where they disagree, and how to run a pilot.
How much should an agent do?
Context, permissions and human review across levels of autonomy.
Start with your kind of work
Different roles need different software, controls and compromises.
From one task to a working system
Thirteen complete workflows, with a person at the right checkpoint.
Before you connect your data
A practical checklist for permissions, retention and an exit path.
Try it. Measure it. Then decide.
A repeatable way to test, adopt, restrict or reject a tool.
The price beyond the subscription
Seats, credits, review time and the cost of switching.
Who is building the category?
The platforms, specialists and research institutions shaping the market.
What to watch next
A dated view of emerging capabilities and their practical limits.
The original report and its 36 product notes are preserved as a dated research edition. New field guides and catalog entries have separate source checks. A vendor-page check does not establish performance, current pricing or independent security assurance.
Read the method and limitations ↗