Welcome to Day 6 of the challenge.
Most people who give up on AI don’t give up because AI is bad. They give up because of one moment.
The moment Claude didn’t understand what they were looking at. The moment it generated something wrong and they couldn’t articulate why. The moment they got stuck on a screen and didn’t know what to ask. They tried, it didn’t land, they bailed.
Today’s move is the way out of that ditch. It’s the smallest technique in the challenge. It’s also the one you’ll use more often than any other from here on.
A word before we start
Today isn’t a teaching day in the same shape as the others. There’s no big task to run. There’s one habit to install.
When Claude isn’t getting it: screenshot what you’re looking at and paste it in.
That’s the move. Everything below is the detail of when and how.
(The brief on you from Day 2 still applies. The TCCA structure from Day 1 still applies. Today’s habit sits on top of both. When the structured brief isn’t enough, the screenshot finishes the conversation.)
The move: show, don’t tell
Claude can see images. Drop a screenshot into a chat and Claude reads it the same way it reads typed text. Better, in some cases, because the screenshot carries layout, design, mistakes, error messages, all of it. Stuff you’d struggle to describe in words.
This matters whenever:
- You’re stuck on a screen and don’t know what to click
- Claude generated something wrong and you can’t articulate the problem
- A file isn’t loading, or a connector isn’t behaving, or an error message popped up
- You want feedback on something visual (a slide, a poster, a draft email layout, an Instagram post)
- You’re following instructions from somewhere else (a tutorial, a manual, a colleague’s email) and want Claude to read along with you
The default reflex used to be: type out what’s wrong. The new reflex: screenshot and paste.
How to take a screenshot (in case it’s been a while)
On a Mac:
- Whole screen: Cmd + Shift + 3 (saves to Desktop)
- Selected area: Cmd + Shift + 4 (drag to select, then release. Saves to Desktop.)
- Specific window: Cmd + Shift + 4, then press Spacebar, then click the window.
On Windows:
- Whole screen: PrtSc (copies to clipboard, then paste straight into Claude)
- Selected area: Win + Shift + S (opens Snipping Tool, drag to select)
On iPhone or iPad:
- Side button + Volume up (saves to Photos)
You don’t need to save the screenshot to a file. You can paste directly from clipboard into the Claude chat box. The screenshot just goes in.
Try it now
Step 1. Open Claude. Doesn’t matter which mode. Works in browser, desktop, Cowork, anywhere.
Step 2. Find something on your screen that’s confusing you, or a thing you’ve been meaning to ask about. Could be a UI you don’t know what to do with. Could be a document that’s not landing. Could be a poorly-written email someone sent you.
Step 3. Screenshot it.
Step 4. Paste the screenshot into Claude.
Step 5. Type the question. Examples:
- “What am I looking at here?”
- “How do I fix this error?”
- “What’s wrong with this layout?”
- “Walk me through what to click next.”
- “Read this and tell me what they’re actually asking for.”
Hit send. Notice how much shorter that conversation was than the version where you typed the whole thing out.
When we start ‘coding’ or at least vibe coding and using Claude Code or Codex this approach really helps. I have used it a lot!
Three more rescue moves while we’re here
The screenshot move is the big one. Three smaller ones, all worth knowing:
1. Strip the brief back. When Claude is going wrong, the instinct is to add more context. Usually wrong. Try the opposite. Delete everything except the core task. Sometimes Claude was over-fitting to one of your instructions and the fix is fewer words, not more.
2. Start a new chat. When a conversation has wandered and Claude keeps making the same mistake, start fresh. The brief on you from Day 2 is still there. You’re not starting from zero. You’re just clearing the bad memory.
3. Switch to voice. If you can’t get the words right by typing, hit the microphone icon and just talk it through. Voice catches the things typing edits out. Especially when you’re stuck. (If you installed Wispr Flow back on Day 0, this is where it earns its keep. Hold the hotkey, talk, paste straight into Claude.)
What you’ll notice
The “I give up” moment doesn’t go away. It still happens. You just have a way out of it now. Most of the time, the screenshot makes Claude understand instantly. The other rescue moves cover the rest. After a few weeks of using this, you stop describing problems and start showing them. That’s the actual shift.
Why this works
Claude is multimodal. It reads images the same way it reads text. There’s no penalty for using screenshots, no separate skill to learn, no “image mode” to switch into. You just paste.
People underuse it because we were trained on text-only AI for two years. Habit hangover. The screenshot move recalibrates the reflex.
Key takeaway
When Claude isn’t getting it, screenshot what you’re looking at and paste it in. One habit. Solves more friction than any other in this challenge.
Plus the three back-pocket moves: strip the brief back, start a new chat, switch to voice.
Where this is heading
Six days in.
You can brief AI (Day 1).
Claude knows you (Day 2).
Claude asks back (Day 3).
Cowork is installed (Day 4).
You’ve done your first proper task (Day 5).
Now you’ve got the rescue moves.
Tomorrow we close out Week 1 with your first connector.
Tomorrow on Switched On
Day 7: Connect Gmail. Catch the annual subscriptions you missed. Your Day 5 table is incomplete. The inbox knows what the statement missed. Today we connect Claude to Gmail and finish the job.
Nishma x
Catching up? Days 0 to 5 are on the Switched On Substack archive HERE.
Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge
SwitchedOn #SwitchedOnAIChallenge
