This is how to show AI instead of telling it
Half of what you’d struggle to describe, you can photograph or attach. Skill five: your camera and your files are inputs.
This is how to show AI instead of telling it.
People spend ten minutes typing a description of a rash, a room, a spreadsheet, a clause, when the thing itself was right there the whole time. Describing is lossy: you leave out what you didn’t notice, and what you didn’t notice is usually the answer.
The skill is a reflex: before typing a description of anything, ask whether you can photograph it, screenshot it, or attach it.
What "showing" covers
- The physical world: the dashboard light, the plant, the water stain spreading on the ceiling, the assembly step that makes no sense. This site’s Hidden Powers section is mostly this skill in action.
- Documents: the lease, the bill, the benefits PDF, the contract. Attach the whole thing and ask about it; don’t retype the one clause you THINK matters.
- Screens: the error message, the confusing settings page, the suspicious text message. Screenshots are the fastest input that exists.
- Data: the spreadsheet you inherited at work. "What is going on in this file" is a legitimate and spectacular prompt.
Watch it on a lease
The ask:
(attached: the 14-page lease) I’m signing this Thursday. What would a tenant’s lawyer flag? Anything unusual compared to standard leases, anything that costs me money in a non-obvious way, and the two questions I should ask before signing.
The version where you type "the lease says something about maintenance responsibilities, is that normal?" gets you a paragraph about typical maintenance clauses. The attached version found the clause that made the tenant responsible for the first $150 of EVERY repair, did the math on what that costs in a year of normal apartment entropy, and drafted the question to ask about it. The difference is the difference between asking about what you noticed and being told what you missed.
How to ask about what it’s looking at
- Say what you need, not just "explain this": "what would a lawyer flag" beats "summarize my lease."
- Tell it what you already know, so it skips the obvious.
- Ask "what should I look at more closely that you can’t read well?" It knows where the photo was blurry or the scan was cut off, but only tells you if asked.
The limits, honestly
- It misreads. Numbers especially: confirm any figure it quotes against the document with your own eyes before acting on it.
- Photos carry more than the subject. The screenshot with the embarrassing tab, the document with your SSN in the corner. Crop before you attach; the what-not-to-paste skill applies to pixels too.
- Some things deserve a human eye first. A photo of a mole goes to a dermatologist, not a chatbot. Showing is for preparation and understanding, and the medical/legal/money rule of this site does not bend for images.
Once this reflex installs, the amount of life that becomes askable roughly doubles. That’s not a figure of speech; count your camera roll.
This is how to use AI.