AI Guides › Judgement & Guardrails
The Honest Case For Slowing Down When AI Makes Something Feel Easy
By Nigel Guy · 3 min read
Something that used to take an afternoon now takes two minutes, and the
speed itself starts to feel like a form of evidence — surely, if it came
together this easily, it's basically right. That feeling has nothing to do
with whether the output is actually correct. Ease is a property of the
tool, not a property of the answer, and mixing the two up is how effortless
output slides through checks that a slower, harder-won answer would have
had to survive.
The rule: how easy something was to produce tells you nothing about how
correct it is — slow down in proportion to what's at stake, not in
proportion to how much effort the output cost you to get.
The mechanism: the ease tax
- Notice the specific feeling of "that was surprisingly fast." It's a
useful trigger precisely because it's identifiable — a small flicker of
relief or satisfaction at how little effort something took.
- Treat that feeling as a prompt to pay a deliberate tax on speed, not
as a green light. The tax is simple: before using the output, spend at
least a portion of the time the task would normally have taken, checking
it instead of skipping straight to using it.
- Ask what you would have double-checked if you'd done this the slow
way. Doing something by hand usually builds in natural checkpoints —
drafts, recalculations, re-reads. Reconstruct at least one of those
checkpoints deliberately, since the fast route skipped past it.
- Separate "this took no effort" from "this needed no effort." Some
tasks genuinely are simple and don't need much scrutiny regardless of
how they were produced. The tax applies when the ease is doing work that
the actual complexity of the task doesn't justify.
- Pay the tax before sharing, not after. The point is to catch
anything wrong while it's still cheap to fix, which is the same logic
behind catching errors early on the cost ladder for any AI mistake.
What to skip
Skip applying this tax to genuinely trivial output — a quick rephrase, a
casual first draft, something you were always going to revise anyway.
The tax is proportional to stakes, and taxing everything equally just
trains you to resent and eventually skip it. Skip, too, assuming that
slow output from an AI tool is automatically more trustworthy than fast
output — the tax is about compensating for a psychological bias, not about
the tool's actual processing time being a meaningful signal either way.
Guardrails
- This is a bias-correction habit, not a technical check — it doesn't
substitute for verifying a specific claim, number, or citation. Use it
alongside those checks, as the trigger that reminds you to run them.
- The size of the tax should track the stakes of the task, not the size of
the time saving. A huge time saving on a low-stakes task doesn't need a
large tax; a modest time saving on a high-stakes one might need a bigger
one than the saving itself.
- Recognise that this feeling gets stronger with repeated use, not weaker —
the more often a tool saves you real effort, the more tempting it becomes
to read that ease as reliability. That trend is worth watching in
yourself over time, not just per task.
All 751 AI guides · JulieMango plans from £17/mo