Make an AI agent's working directory a trust boundary: scope, disclose, consent
Gated AI Launch Checklist for Trust
Re-Anchor Your AI Model Pick After a Leaderboard Rule Change
UX Patterns
Design an AI agent users can text without losing trust
A texting agent earns trust when it picks the right channel, limits proactive pings, keeps risky approvals in the app, and states what it cannot see.
Now trending
Latest
Make an AI agent's working directory a trust boundary: scope, disclose, consent
Gated AI Launch Checklist for Trust
Re-Anchor Your AI Model Pick After a Leaderboard Rule Change
Design an AI approval surface users won't click through: block dangerous actions by default
Ship AI actions users can revoke: spend, send, delete, book
Score model voice before you swap models, so trust doesn't drift
Design AI answers users can trust: use task-matched components and a density dial
Design AI agents users trust: name modes, pause for plans, keep-or-skip diffs
Evals
-
Re-Anchor Your AI Model Pick After a Leaderboard Rule Change
Treat a leaderboard rule change as a product event: re-baseline privately, define merge-worthy outcomes, and compare cost before swapping.
-
Score model voice before you swap models, so trust doesn't drift
Treat model voice as a product spec: baseline it, score it, and ship only when trust-relevant changes clear the noise.
-
Export a Runnable Eval Suite Before the Read-Only Date
Treat the read-only date as the export deadline: a runnable suite needs a completed run and rebuilt graders.
Launches
-
Gated AI Launch Checklist for Trust
Gate access, leave pricing and GA dates open, and publish a limitation contract so users know what the AI can and cannot do.
-
Ship AI actions users can revoke: spend, send, delete, book
A six-line audit for AI features that spend, send, delete, or book: approval, caps, scope, data, identity, and recovery.
-
Design AI agents users trust: name modes, pause for plans, keep-or-skip diffs
Ship AI agents with named modes, plan approval, and keep-or-skip diffs so users can see what the system may touch and undo what it did.
Trust & Safety
-
Make an AI agent's working directory a trust boundary: scope, disclose, consent
Scope the working directory, classify folders, surface loaded AGENTS.md files, and require consent for config changes before shipping AI coding agents.
-
The Co-Pilot Era of Poker: The Best Player at the Table Is You, Graded in Real Time
Autopilot plays the game for you. Co-pilot plays with you — and tells you, in seconds, whether the line you chose was the right one. Poker just got its co-pilot. Here's what the human-plus-machine team looks like, and why the human still has to be the one driving.
-
Web-reading AI: a three-point contract for blocked pages
When a page refuses your reader, the answer should say what it missed, why, and what source it tried next.
UX Patterns
-
Design an AI agent users can text without losing trust
A texting agent earns trust when it picks the right channel, limits proactive pings, keeps risky approvals in the app, and states what it cannot see.
-
Design an AI approval surface users won't click through: block dangerous actions by default
A block-first approval surface lets low-risk agent work run silently, stops destructive actions, and gives users a clear manual escape.
-
Design AI answers users can trust: use task-matched components and a density dial
A design checklist for AI answers: emit native components, match format to the task, and ship a user-controlled density dial.
Don't miss
Advertisement
The latest