In partnership with
👋 Welcome to AI Central,
When AI gets something wrong, is your first instinct to rewrite the prompt and try again?
I do this too, but sometimes you end up asking the same thing three different ways and still get nowhere.
The trick is to figure out why the answer went wrong before trying again. Was it wrong from the start, or did the conversation go off track somewhere along the way? The fix depends on which one it is.
In today's guide, we're looking at 9 common ways AI gets things wrong, how to spot them and which problems you can't fix just by rewriting your prompt.
Big hug and love you ❤️
Kris & The AI Central Team
P.S. We're building this with you - so if you're loving it (or have any suggestions) leave us a review here
🔗 TOGETHER WITH THE CODE
The Agentic Engineering Playbook top engineers are using
Most engineers are stuck babysitting AI one prompt at a time. A handful have figured out how to build agents that plan, execute, and self-correct on their own.
Sign up for The Code newsletter and get access to The Complete Playbook for Agentic Engineering. Claude Code, subagents, MCP, all of it.
You're seeing this ad because you're on the free plan - upgrade to remove ads & unlock 1200 AI Tutorials. Ads keep AI Central free for everyone
📊 Where are you on the AI map?
Ten questions, about 40 seconds. You'll get your AI Readiness Type and the one thing worth working on next.
At the end there's a Member Pass card with your stage and profile, ready for LinkedIn if you want to share it.
🔗 BROUGHT TO YOU BY THE CODE
100+ Claude Code hacks to ship code 10X faster
Top engineers at Anthropic say AI now writes 100% of their code.
Are you using AI to write yours?
These 100+ Claude Code hacks show you exactly how. Sign up for The Code and get:
100+ Claude Code hacks — free
The Code newsletter — learn the latest AI tools and skills to code faster in 5 mins a day
You're seeing this ad because you're on the free plan - upgrade to remove ads & unlock 1200 AI Tutorials. Ads keep AI Central free for everyone
📓 How to fix a bad answer
Inside this guide you will learn:
Section | What you can do by the end |
|---|---|
23.1 A bad answer is evidence, and it points at its own repair | Sort any failure into ask, thread or ceiling with a single question |
23.2 5 shapes that mean your request was the problem | Repair five common failures where you're standing, usually in one sentence |
23.3 4 shapes that mean the conversation is the problem | Spot a dead thread, and leave it the way that keeps 95% of its value |
23.4 Some failures are ceilings, and no rewording moves them | Recognise a ceiling before you've rephrased the same thing four times |
You'll need: your quality bar, and one recent answer that failed it, still sitting open in your history. Bring the whole conversation it came from, not just the last reply.
🚨 Something big is coming tomorrow
Ever been curious about newsletters? Or thinking about starting a creator business? What about growing via a channel that you own (and not battling with an algorithm)?
We’ve got just the playbook for you.
We’re sharing the steps to launching your newsletter, growing it, making it pay and running it with AI.
The first drop will hit your inboxes tomorrow.
In partnership with beehiiv
📝 Use these prompts
Try this when the conversation has gone off track and rewording your request isn’t helping.
What most people type:
"No, that's not what I asked for. Try again, and this time keep it under 200 words."
After the third correction, the chat is still carrying the wrong turn. Each new instruction lands on top of it.
What to type instead:
In the old chat: "Before we go on, summarise in one paragraph everything we've agreed so far: what I need, the decisions we made, and anything I've ruled out. Don't add anything new."
In a new chat: "[Paste the paragraph, corrected by you]. Using only the above, [your original request]. Give me [format] and stop there."
One paragraph: the AI writes down what it thinks you’ve agreed, so you can check it
Corrected by you: this is usually where you spot the mistake that sent things wrong
A new chat: you keep everything you agreed and leave the mistake behind
