DEV Community: Hey SI team

Stuck in an AI fix loop? Do this before the next prompt.

Hey SI team — Sun, 18 Oct 2026 14:45:26 +0000

You know the feeling. Something breaks. You ask the AI to fix it. It says "Fixed!" It isn't. You paste the error again. It apologizes and says "Fixed!" again. Twenty minutes later the original bug is still there, two new ones have shown up, and you can't remember what the app looked like when it worked.

That's a fix loop. No stress: getting stuck in one is normal, and it happens to experienced developers too. It doesn't mean you're bad at this. It means the current approach has stopped working, and it's time to change the approach, not send the prompt a fifth time.

This post works the same whether you build in Cursor, Lovable, Bolt, Replit, or a plain AI chat.

Why fix loops happen

A few things usually go wrong at once:

The AI is guessing. If it can't see the real cause, it patches the symptom it can see. Each patch looks reasonable, so it keeps going.
The real problem lives somewhere else. The error shows up on one screen, but the cause might be bad data that was saved earlier, a setting, or a different file. The AI keeps editing the file where the error appeared.
The chat gets cluttered. Long conversations fill up with old attempts, half-reverted code, and wrong assumptions. The AI starts building on top of its own earlier mistakes.
Too many changes at once. When one "fix" touches five files, you can't tell which change helped and which one broke something new.
"Still broken" isn't much of a clue. Without the exact error and what you expected to happen, the AI has very little to work with.

Every one of these has a simple counter-move.

The routine: do this before the next prompt

1. Stop

Seriously. Don't send another "it's still broken." A good rule is two strikes, then reset: if the same fix fails twice, you've left fixing territory and entered loop territory. Take a breath. We'll figure it out.

2. Get back to solid ground

Find your last working version. Most tools have a way back: Cursor, Replit, Lovable and Bolt all offer some kind of checkpoint, restore, or version history (the name and location vary, so check your tool's docs). If you use git, your last commit is a save point too.

Rolling back isn't failure. It's often the fastest fix there is. If you're not sure you want to roll back yet, at least make a save point of where you are now, so you can't lose anything while you experiment.

3. Describe what's actually happening

Write down four things in plain words:

What you did (clicked, typed, ran)
What you expected
What happened instead
The full error message, copied exactly, plus where you saw it (browser page, browser console, terminal, the tool's preview)

"It's broken" becomes "When I click Save on the add-task form, I expected the task to appear in the list, but the page goes blank and the browser console says TypeError: Cannot read properties of undefined." That's a clue the AI can actually use.

Quick safety note: if the error contains an API key or password, swap it for [REDACTED] before pasting.

4. Ask the AI to explain before it fixes

This is the big one. Ask for the likely cause and the evidence for it without any code. When the AI has to explain its reasoning, it's more likely to notice a wrong assumption, and you get to check whether the explanation actually makes sense.

Here's a prompt for that:

Pause. Don't write or change any code in this reply.

The problem: [ONE SENTENCE, e.g. "new tasks don't show up after I click Save"]
What I did: [STEPS]
What I expected: [EXPECTED]
What happened instead: [WHAT YOU SAW]
Full error message: [PASTE ALL OF IT, WITH ANY KEYS REPLACED BY [REDACTED]]

Please:
1. Explain in plain English what you think is causing this, and why.
2. Tell me what evidence supports that, and what would prove it wrong.
3. If you're not sure, suggest ONE small check (like a log message) to find out, instead of guessing.
4. Tell me which file(s) the real cause is most likely in.
Then wait for me before fixing anything.

5. One change at a time

Once the cause makes sense, ask for the smallest possible fix to that one thing. Then test it yourself. Works? Make a save point right away. Doesn't work? Roll back that one change and tell the AI exactly what happened. One change, one test, one save point.

6. Start a fresh chat with a summary

If the conversation is long and tangled, start over with a clean chat. You don't lose what you learned, because you bring a short summary with you. Ask the old chat to write it first:

We're going to continue this in a fresh chat. Don't write any code.

Write a short handoff note (under 10 lines) I can paste into a new chat. Include:
- What the app is and what part we're working on
- The problem, in one sentence, with the exact error
- What we've already tried and why each attempt didn't work
- What's been ruled out
- The most likely cause we haven't tested yet
- The one next step to try

Paste that note into the new chat, add "Explain your plan before changing any code," and you're back on track with a clear head (yours and the AI's).

7. Try a simpler path

If one feature keeps fighting you, ask: "Is there a simpler way to do this?" Roll back, rebuild that one step more simply, and keep moving. Shipping a smaller version that works beats a fancy version that doesn't.

The short version

Two failed fixes? Stop.
Roll back or make a save point.
Describe: what you did, what you expected, what happened, full error.
Ask for the cause and evidence first, no code.
Smallest fix, test, save point.
Messy chat? Fresh chat with a handoff note.
Still stuck? Find a simpler path.

Our cartoon mascot SI the Yes Bot has a motto for this moment: an error isn't a disaster, it's a clue. You've already got everything you need to read it. Ride the wave. SI's got you. 🌊

Want more prompts like these? Grab the free 5-prompt kit at https://heysi.dev.

Vibe coding for beginners: 5 free prompts to build your first app

Hey SI team — Fri, 16 Oct 2026 13:46:15 +0000

"Vibe coding" just means building software by describing what you want to an AI coding tool (Cursor, Claude Code, Replit, Lovable, Bolt, or a plain AI chat) and steering it in plain English. You don't need to know the jargon to start. You do need to ask clearly, though, and that's the part most beginners get stuck on.

This guide gives you five prompts that cover the moments where first-time builders usually stall: a fuzzy idea, a scary red error, a fix that keeps failing, a safety check before you share, and actually getting a link online. Copy them, fill in the [PLACEHOLDERS], and paste.

How to use these prompts

Copy a prompt into your AI coding tool.
Replace every [PLACEHOLDER] with your own details. Messy is fine.
Send it, then read the answer. Ask "what does that mean?" as often as you like. There are no silly questions.
One small step at a time. Ask for one thing, check it works, then move on.

Five habits that make everything easier

Small steps. Never ask for "the whole app" at once.
Save often. When something works, make a save point (a git commit, which is a snapshot you can go back to) before changing anything else.
Read the error. Paste the whole message and say what you were doing.
Two strikes, then reset. If a fix fails twice, pause and change approach.
Never paste secrets into chat. API keys and passwords go in a .env file (a private settings file that stays off GitHub) or your tool's secrets settings. A placeholder works great.

A quick, honest heads-up: AI tools make mistakes and change often. Always look over what they change, test it, and keep save points. These prompts help you ask clearly. They can't promise a result.

Prompt 1: Brain Dump to Clear Idea

When: You've got an idea but it's all jumbled. Start here, before any code.

I want to build my first app, but my idea is still messy. I'm a complete beginner at coding.

Here's my brain dump:
[WRITE EVERYTHING YOU'RE THINKING: WHAT IT DOES, WHO IT'S FOR, WHY YOU WANT IT, APPS IT'S LIKE]

Please help me get clear:
1. Rewrite my idea as ONE simple sentence: "[App] helps [who] do [what] so they can [why]."
2. Tell me who the main user is.
3. List the 3 most important things the app must do.
4. Point out anything in my brain dump that sounds big, complicated, or expensive for a beginner.
5. Ask me up to 5 questions whose answers would make the idea clearer.

Use plain, non-technical language. Don't write any code yet.

Why it works: "Don't write any code yet" stops the AI from rushing into a big, half-right app. The one-sentence format forces a clear goal, and asking it to flag the expensive or complicated parts helps you find a first version small enough to finish.

Tip: Write messy. Then answer its questions and say: "Now rewrite the one-sentence idea using my answers."

Prompt 2: Here's the Error, Fix It

When: A red error popped up. That's a clue, not a disaster. Paste the whole thing.

I got an error. I'm a beginner, so please be careful and explain simply.

What I was doing: [WHAT YOU CLICKED / TYPED / RAN - e.g. "I clicked Save on the add-plant form"]
What I expected: [WHAT SHOULD HAVE HAPPENED]
What happened instead: [WHAT YOU SAW]

The full error message:
[PASTE THE ENTIRE ERROR - ALL OF IT, EVEN IF IT'S LONG]

Where I saw it: [TERMINAL / BROWSER PAGE / BROWSER CONSOLE / TOOL PREVIEW]
Last thing that changed before it broke: [e.g. "we just built step 4" / "I don't know"]

Please:
1. Explain in 1â€“2 plain sentences what the error means.
2. Tell me the most likely cause.
3. Fix it with the smallest change possible. Don't rewrite unrelated code.
4. Tell me how to check that it's fixed.
If you're not sure of the cause, add a temporary check (like a log message) to find out first, instead of guessing.

Why it works: "Expected vs. what happened" plus the full error and the last change gives the AI the context it needs to find the real cause. "Smallest change possible" stops it from rewriting working code, and "check first instead of guessing" cuts down on random fixes.

Tip: Never paste API keys or passwords. Swap them for [REDACTED]. Fixed? Make a save point.

Prompt 3: Break the Loop

When: The same fix has failed 2â€“3 times. Time to pause and reset.

Stop. We've tried to fix this [NUMBER] times and it's still not working. Don't write any code in this reply.

The problem: [ONE SENTENCE - e.g. "plants don't stay marked as watered after refresh"]
What we've tried so far (as far as I know): [LIST, OR "please list everything we've tried from this conversation"]

Please:
1. Summarize what we tried and why each attempt didn't work.
2. Tell me honestly: do you think the cause is something we haven't looked at yet? What assumptions might be wrong?
3. Suggest ONE completely different approach, simpler if possible.
4. Tell me which files we should look at first.
5. Write a short "handoff note" (5â€“10 lines) that I can paste into a fresh chat, describing the problem, what's been ruled out, and the new approach.

If you think the best move is to roll back to my last save point and redo this step more simply, say so.

Why it works: Long chats get cluttered, and the AI can keep repeating the same wrong assumption. Banning code for one reply makes it step back and reason. The handoff note lets you start a fresh chat without losing what you learned.

Tip: Start a fresh chat and paste the handoff note. Rolling back to a save point is not failure.

Prompt 4: Safety Check Before Going Live

When: Before you share your app, and again after adding login, a database or payments.

Before I put my app online, do a beginner-level safety check. Don't change anything yet, just report.

My app uses: [e.g. "Next.js, Supabase database + auth, deployed on Vercel"]
It handles: [e.g. "users' email addresses and their plant lists" / "no personal data"]

Check and report on each of these, as âœ… OK / âš ï¸ Needs fixing / â“ Not sure:
1. Secret keys: are any API keys, passwords, or tokens written directly in the code, or committed to Git history? Is .env in .gitignore?
2. Browser-visible keys: which keys are sent to the browser? Are any of them secret keys that shouldn't be?
3. Data access: can a logged-out person, or a different user, read or change data they shouldn't? (Check database rules like row-level security, not just the screens.)
4. Forms and inputs: does the app check what people type in before saving it?
5. Error messages: do any errors show private details to users?
6. Packages: are there known vulnerable packages (run npm audit and summarize in plain words)?
7. Anything else that worries you for an app like this.

Then list the fixes in order of importance, and we'll do them one at a time.

Why it works: "Just report" keeps the AI from making a pile of changes at once. The checklist covers the mistakes beginners make most, especially leaked keys and data anyone can read, and the âœ…/âš ï¸/â“ format turns it into a to-do list you can work through one item at a time.

Tip: If a key was ever committed or pasted publicly, assume it leaked: make a new one and delete the old. This is a beginner check, not a professional security audit.

Prompt 5: Put It Online (Vercel / Netlify)

When: Your app works on your computer and builds without errors. Time to share it.

I want to put my app online using [VERCEL / NETLIFY]. I'm a beginner. Walk me through it step by step.

Stack: [YOUR STACK]
Is my code on GitHub? [YES - repo name: ___ / NO]
Does my app use environment variables (.env)? [YES / NO / NOT SURE]
Does it use a database or login? [e.g. "Supabase database + auth"]

Please:
1. Check the project is ready: it builds with npm run build, .env is NOT in Git, and .env.example lists every variable needed.
2. If my code isn't on GitHub yet, help me push it there first.
3. Give me click-by-click steps to import the GitHub repo into [VERCEL / NETLIFY], including which build settings to use (or confirm the defaults are right).
4. List every environment variable I need to add in the [VERCEL / NETLIFY] dashboard, by NAME only. I'll paste the values myself.
5. List anything I need to update in other services now that I have a live address (e.g. Supabase auth redirect URLs).
6. Give me a short checklist to test the live site.

Why it works: It checks that the project is ready before you deploy, which catches most "works on my machine" problems early. Asking for variable names only keeps real secrets out of the chat, and step 5 catches the redirect-URL problem that breaks login on a lot of first deploys.

Tip: Your .env file doesn't upload by itself. Add the variables in the Vercel/Netlify dashboard, then redeploy. On Lovable, Bolt or Replit, their Publish button is usually easiest.

Putting it together

A good first loop goes like this: use Prompt 1 to shrink the idea, build one small feature at a time, and make a save point each time something works. Reach for Prompt 2 when something breaks and Prompt 3 when the same fix fails twice. Before you share the link, run Prompt 4, then Prompt 5. Getting stuck along the way is normal for everyone, not just beginners. Each error tells you something new.

What was the first thing you built (or want to build)? Drop it in the comments. Small ideas are welcome.

Disclosure: we make the Hey SI Starter Pack ($12), which has 20 more prompts like these. It's at heysi.dev. These 5 are also available free as a PDF through our email list: https://si-from-hey-si.kit.com/190809ef45