PRACTICE, NOT JUST READING

Do the thing. Then check what you did.

A Reality Lab hands you a flawed artifact — a leaky prompt, a confident but unsourced draft — to repair and self-check. A Consequence Room lets you make a choice and then watch what happens after it, the part a quiz question hides.

All 12 Reality Labs and 6 Consequence Rooms are here — the full set of 12 Reality Labs and 6 Consequence Rooms. Every workspace stays in your browser: nothing you type is sent anywhere.

WHAT THIS ISA Reality Lab hands you something flawed to repair, then lets you check your work. A Consequence Room lets you make the tempting choice and shows what it costs later. See the legend

Reality Labs

Produce and check a real artifact
Reality Lab · 12 min · JudgeRepair a confident draftA model wrote a fluent paragraph. Some of it isn't in the source. Find what it invented and fix it.Start →Reality Lab · 10 min · JudgeRewrite a leaky promptA colleague's prompt gets the job done — and pastes a real customer's private details into a third-party tool. Fix both problems.Start →Reality Lab · 15 min · StewardTurn a process into an agent workflowTake one real, repetitive process and design an agent for it — with the access, the gates, and the owner decided before anything is connected.Start →Reality Lab · 8 min · UseSharpen a vague requestA one-line prompt gets a generic, off-target answer. Rewrite it so the model spends its effort on your problem, not the average one.Start →Reality Lab · 9 min · UseAdd the examples that fix itYou keep describing the format you want and keep not getting it. Stop describing — show two examples, and watch the output snap into shape.Start →Reality Lab · 10 min · UseBreak a big ask into stepsOne giant prompt asks for research, analysis, and a finished deck in a single go — and every part comes back mediocre. Split it into steps you can check.Start →Reality Lab · 10 min · UseGround an answer in its sourcesAn AI research answer is fluent, specific, and completely unsourced. Rework the request so every claim arrives attached to something you can check.Start →Reality Lab · 9 min · UseMake the draft your ownAn AI wrote a smooth first draft of something you’ll put your name on. Turn ‘the model said it’ into ‘I stand behind this’ — content, voice, and disclosure.Start →Reality Lab · 10 min · JudgeMatch the check to the riskYou can’t fact-check everything to the same depth. Given five AI outputs, decide how hard to verify each — so the effort lands where being wrong actually costs something.Start →Reality Lab · 9 min · JudgeVerify a message that might be fakeAn urgent voice note from your ‘manager’ asks you to move money fast. It sounds exactly like them. Design the check that a convincing fake can’t pass.Start →Reality Lab · 12 min · JudgePlace a use on the risk mapBefore asking ‘is this legal?’, ask two sharper questions: how risky is this use, and what is our role in it? Practise the triage that a specialist actually needs from you.Start →Reality Lab · 8 min · OrientDraw the boundary firstThe safest time to decide what an assistant should never see is before you start typing. Set the boundary for a real task, then hold it under pressure to be helpful.Start →

Consequence Rooms

See what happens after a choice