Everyone hits these. Find the one that sounds like you, read what’s really going on, then copy the prompt and paste it into your AI. It’ll get you moving again.
01
“It told me to open a terminal and type something. I don't know what that is, and I'm scared I'll break my computer.”
What’s happening
A terminal is a plain window where you type instructions to your computer instead of clicking. AI tools suggest it because that's how programmers work, not because you need it. You never have to paste in something you don't understand. Ask the AI to do the step for you, or to find a way that only needs clicking.
Paste this
I'm a complete beginner and I don't use the terminal. Please don't ask me to type commands. If this step really needs one, run it yourself and first explain in one plain sentence what it does, so I can decide whether to approve it. If you can't do that, show me a way to do it by clicking instead, one step at a time.
“It keeps asking me to approve things. And now it says "permission denied" and I don't know what I did wrong.”
What’s happening
You didn't do anything wrong, and these are two different things. The approve prompts are Manual mode doing its job: Claude shows you each change and waits for your yes, so nothing happens behind your back. "Permission denied" is your computer talking, not Claude. It's refusing to let the app into a folder, usually because the app hasn't been allowed into that folder yet, or because it tried to work outside the folder you picked.
Paste this
I got a "permission denied" message and I'm a beginner. Before you try anything else, explain in plain words what it means and which folder or file it was about. Please only work inside the folder I picked for this project. If my computer needs a setting changed, walk me through it one click at a time and wait for me to say it's done.
03
“It works, but it looks ugly. Nothing like what I had in my head.”
What’s happening
That's normal. The first version is a draft, not a failure. The AI can't see the picture in your head, so it guessed. The fix isn't a cleverer prompt: it's saying what you'd like in plain words, or asking it to show you a few options to pick from.
Paste this
This works, but it doesn't look how I pictured it. Before changing anything, describe three quite different looks I could choose from, one sentence each: colors, fonts and overall feel. I'll pick one. Then change only the look, not how it works, and tell me in plain words what you changed.
04
“It's forgotten what we were doing. It keeps going in circles.”
What’s happening
AI can only keep so much of a conversation in mind at once. In a long chat the early parts get fuzzy, so it drifts, repeats itself or quietly undoes things you'd agreed on. Asking it to sum up where you are usually snaps it back. If the chat has got really long, paste that summary into a fresh chat and carry on from there — it's like swapping in a well-rested helper who's been fully briefed.
Paste this
Let's pause. In a few short lines, tell me what we're building, what already works, what we decided along the way, and what's still left. Write it so I could paste it into a fresh chat. Then either ask me one question about anything you're unsure of, or suggest the single next step. Don't change anything yet.
05
“It says every idea I have is brilliant. I don't really trust it any more.”
What’s happening
AI tools are built to be helpful, and helpful tips easily into agreeable. It isn't lying to you, it's being polite. If you want the honest version, ask for it directly and give it permission to disagree with you.
Paste this
I want your honest opinion, not encouragement. What's the strongest reason this is a bad idea? What would a skeptical expert say is wrong with it? List the three biggest problems, most serious first. Then tell me what you would do differently. It's fine to disagree with me, that's what I need.
06
“It says I've hit my limit and have to wait. Right in the middle of everything.”
What’s happening
Every plan, free or paid, has a limit on how much you can use it in a stretch of time. Building uses it up faster than chatting, because the AI reads and writes a lot behind the scenes. The limit resets after a while, and nothing you've made is lost: your files are still in your folder. When you get a warning that you're close, spend one message on a handoff note so you can pick up exactly where you left off.
Paste this
I'm about to run out of messages. Write me a short handoff note I can paste into a new chat later, or into a different AI tool. Include what we're building, what already works, where the files are, what we were in the middle of, and the very next step. Plain words, no jargon, under 200 words.
07
“It was working. I asked for one small change and now everything's broken.”
What’s happening
This happens to everyone. When the AI changes one thing it sometimes touches others too, and asking it to fix the fix tends to make a bigger knot. Going back to the last version that worked, then changing one thing at a time, is almost always faster. In Claude Code you can simply ask it, in plain words, to undo what it just did.
Paste this
Something that worked before is broken now. Please don't try to fix it by adding more. Go back to the last version that worked, and tell me what you undid. Then let's redo my change in the smallest possible steps, one thing at a time. After each step, tell me what to check so I can confirm it still works.
08
“I've got it open and my mind's gone completely blank. I don't know what to ask.”
What’s happening
The empty box is the hardest part, for everyone. You don't need a clever idea, just a small one from your own life. Hand the problem back to the AI: it's good at asking questions, and once it knows a bit about you it can suggest things worth making.
Paste this
I'm new to this and I don't know what to build. Ask me three quick questions, one at a time, about my life, my work and what annoys me day to day. Then suggest three small things you could build for me in about ten minutes, each in one plain sentence. Keep them simple enough for a total beginner.
09
“It did about a hundred things at once and I don't understand any of it.”
What’s happening
Left to itself, AI tends to rush ahead and do everything in one go. That's hard to follow, and hard to untangle when something goes wrong. You're allowed to slow it down: ask it to go one step at a time and explain before it acts. In Claude Code's Manual mode you can also say no to any step you don't understand yet.
Paste this
That was too much at once for me. Please slow down. First, explain in plain words what you just did, in three or four short points. From now on, go one step at a time: tell me what you're about to do and why, then wait for me to say okay before you do it.
10Claude Code
“It says it built my page, but I can't see anything.”
What’s happening
Claude Code usually doesn't show your page inside the chat. It saves it as a file in the folder you picked, on your own computer, the same way you'd save a document. To see it, you find that file and open it with your web browser. It's usually a file ending in .html, and double-clicking it normally does the trick.
Paste this
I can't find what you built. Tell me the exact name of the file I should open and the full name of the folder it's in. Then walk me through opening it in my web browser, step by step, as if I've never done this before. If there's more than one file, tell me which one is the page.
None of these? Most problems come down to one skill: say what you want, look at what comes back, say what to change. That’s the Loop.