Poke holes in my idea
Work & EmailCreative
Give the idea and who it is for, and get three weaknesses and two ways to test them instead of praise.
Fill the blanks and it copies as you filled it
Give me an honest read on the problems with this idea. Skip the praise. - The idea: - Who it's for: - What I've already checked: - Three weaknesses or risks, and two ways to test them
0 of 3 blanks filled · anything left blank copies with its brackets
What this prompt does
Compared with asking the same thing without the length and format lines.estimate
Estimated output tokens
700
Asked loosely
320
This prompt
Energy (Wh)
1.215
Asked loosely
0.423
This prompt
Token counts are estimated from character counts, not measured. Energy uses the per-band figures measured by Jegham et al. 2025 as published.
What that saves
15.8 Wh · 5.54 gCO₂e · 74 mL — assuming 20 uses a day
Tap a card for how the conversion was done.
Reported backfire — Specifying output lengthSpecifying a length longer than needed backfires (a 34% increase in 11% of scenarios)
Reported backfire — Asking for several options at onceAn increase of over 50% in 9% of scenarios. Don't use it on questions with one right answer
Where these figures come from
We read the prompt to see whether it carries that instruction. The percentages below are published figures for the rule itself, not a measurement of what this prompt saves.
Assigning a role
44%mean reduction · 95% CI 29–59%
Podder et al. 2026 BP#4 · 19 use cases / 61 prompts
by modelGPT-4o 34%Command R+ 50%Mistral-7B 49%
Specifying output length
38.7%mean reduction · 95% CI 29.7–47.7%
Podder et al. 2026 BP#5 · 18 use cases / 57 prompts
by modelGPT-4o 35%Command R+ 49%Mistral-7B 32%
Reduced in every scenario. No backfire reported
Asking for several options at once
36%mean reduction · 95% CI 25–47%
Podder et al. 2026 BP#1 · 22 use cases / 71 prompts
by modelGPT-4o 50%Command R+ 44%Mistral-7B 39%