Fitness influencers these days be like 👇
DON’T BE AVERAGE 😈
💡10,000g protein before 11 AM
💡Gym 15x/day minimum
💡Sleep in a plank position
💡Never drink water — it has zero protein
80% of "Claude sucks" tweets come from Level 1 users.
L1: accepts first response ✋
L2: asks for revisions 🤷
L3: "Critique your own answer before I see it, then rewrite." 🧠
L4: critique → rewrite → critique the rewrite 🔥
What level are you stuck on?
Tell me you're a junior without telling me.
Here's the Claude code-review prompt you use:
😐 "You are a code reviewer."
🙂 "You are a senior dev, be critical."
😈 "You are a grumpy staff engineer who missed their flight because of a bug like this one. Review it."
Which one are you?
90% of bad Claude outputs come from misunderstood inputs.
✅Fix: append this to every complex prompt:
"Before answering, restate my request in your words and flag any assumptions."
Catches the misunderstanding before Claude wastes 4,000 tokens on the wrong answer.
Math checks out. Try it.
Claude has a feature 99% of devs have never used:
VISION.
Screenshot your 👇
📌Stack traces
📌UI bugs
📌SQL query plans
📌Terminal output
📌Whiteboard diagrams
Paste them in. Watch Claude spot what your eyes missed.
Text is dead. Long live PNG.
Constraints > instructions.
Every time.
"Write a good landing page" → slop
"Write a landing page. No buzzwords. No em-dashes. No sentence over 12 words. No bullet points." → gold
Telling Claude what NOT to do is 10x more powerful than telling it what to do.
Stop writing prompts.
Claude should do it for you.
Paste this:
"I want to accomplish [X]. Write the ideal prompt to send to Claude for the best possible result. Include persona, constraints, examples, and output format."
Copy its output. Paste in a new chat. DONE
The single best Claude prompt I've ever written is 14 words:
"Before you answer, ask me 5 questions that would make your answer 10x better."
That's it. Save this. Thank me later.