How to use your AI to build sharper critical thinking in real time
Two prompts you drop into any working session for an instant check
Hey Adopter,
By the end of this edition, you will have two prompts that do something your AI has never done for you unprompted: question your reasoning instead of validating it.
You drop them at the end of any working session, into the chat where your actual work already lives. No setup. The model that has been making your deck look sharp gets pointed backwards at the thinking underneath it. What comes back is not a cleaner draft. It is the question you skipped, the assumption you buried, the gap that clean formatting made invisible.
Here is why that matters more than it sounds.
The work is going faster. The slide looks better. The recommendation reads with authority it didn’t have an hour ago.
Somewhere in that process, the question you never answered got buried under formatting.
This is what AI does to professional work right now. Not because the model is bad, because you fed it your conclusion, it made it fluent, and you both moved on. The gap in your logic is still there. It just ships in a cleaner font.
Think about what happens in most AI-assisted sessions. Half-formed argument, the model rounds the edges, gives it structure. You read it back and feel like you’ve thought it through, because it reads like you have. The sense of clarity is a side effect of the model’s output style, not evidence that your reasoning is sound.
That’s the trap. AI raises the confidence of your output without touching the quality of the thinking underneath. A shaky argument used to look shaky. Now it arrives with headers and a summary table, and nobody spots the hole until a senior stakeholder asks the one question the document never answered.
Two professionals can look at the same AI-assisted deck and both read competence where there is a gap. The model gave them what they asked for: something that reads well.
The fix is not to use AI less. The fix is to point it backwards at your reasoning instead of forwards at your output.
There is a specific way to do that. Two prompts you drop at the end of any working session, into the chat where your actual work already lives. The model already has what it needs. You just redirect what it does with it.
The results, and the discomfort, are worth seeing
What these prompts do
They are not standalone exercises. You drop them at the end of an existing chat, the one where your work already lives. The model has everything it needs from that session. These prompts redirect what it does with that material, from building up your output to pulling apart how you got there.
No pasting content twice. No setup. Drop the prompt, hit send, and answer what comes back.