30 Prompts That Actually Changed How I Work (And the Ones That Flopped)
Last November I spent three hours writing a product spec from scratch. Solid work. Then a teammate pointed out I could have asked Claude to draft it from my bullet points in four minutes. I had Claude open in another tab the entire time. That moment broke something in my brain — not because AI could write the spec, but because I'd built a mental wall between "real work I do" and "stuff I use AI for." Most productivity advice about AI never addresses that wall. This post does.
The Problem With Most AI Prompt Lists
They're written for people who've never used AI before. "Ask ChatGPT to summarize your emails!" Cool. You've been doing that for two years. You're not here for that.
What actually moves the needle for developers and power users is prompts that replace workflows you currently do manually without realizing they're manual. Defaults you've been running on autopilot since 2019. The goal isn't to outsource thinking — it's to find the thirty places per week where you're generating boilerplate cognition instead of insight.
Here's what I've collected from my own work and from watching other builders. Organized by context, not by how impressive they sound.
Morning Clarity and Planning (Prompts 1–8)
These are the ones that compound hardest. Most people skip them because they feel indulgent, like journaling. They're not.
1. "Here's my task list for today. Identify the one item where I'm most likely to procrastinate and tell me why, based on how it's worded."
2. "I have 4 hours of deep work. Here are my tasks. Sequence them to protect the hardest cognitive work for when decision fatigue is lowest — assume I'm sharpest at the start."
3. "Rewrite these tasks as outcome statements instead of activity statements." (Transforms "write docs" into "ship docs so junior devs can onboard without asking me.")
4. "What is the single question I should answer today to make the most progress on [project]?"
5. "I feel stuck on [problem]. Give me three reframes of what the actual problem might be — not solutions, just reframes."
6. "Here's what I planned to do last week vs. what I actually did. What pattern do you see?"
7. "Draft a 5-minute standup for myself. Here's what I worked on. Make it honest — don't spin it."
8. "What would I cut from today's plan if I only had two hours instead of eight?"
Prompt 6 is the one I resisted the longest and now run every Monday. It catches self-deception faster than any retrospective format I've tried.
Deep Work and Writing (Prompts 9–16)
This is where most power users live. The trap is using AI to write instead of you — which produces work that doesn't sound like you and that you can't defend in a room. The better pattern: use AI to clear the friction before and after the hard thinking.
9. "I need to write [thing]. Before I start, ask me five questions that will make the output better. Don't write anything yet."
10. "Here's my rough draft. Don't rewrite it. Tell me where my argument is weakest."
11. "What is the one sentence that this entire document is trying to prove? Extract it from what I wrote, not what I intended."
12. "Steelman the opposite of the position I'm taking in this doc."
13. "I wrote this for a technical audience. Now adapt it for someone who cares about outcomes, not implementation."
14. "Here's my writing. Where am I hedging unnecessarily? Flag every 'might,' 'could,' and 'potentially' and tell me if I actually believe the stronger claim."
15. "Generate five title options for this post. Make two of them uncomfortable — ones I'd be nervous to publish."
16. "Read this and tell me: what will a skeptical reader use to dismiss it?"
Prompt 16 has saved me from embarrassing posts more times than I can count. Run it before you hit publish, not after.
Code, Systems, and Technical Work (Prompts 17–23)
Developers tend to underuse AI on the meta-level and overuse it on the object level. Yes, it can write the function. More valuable: it can help you figure out if you should write the function at all.
17. "Here's the feature I'm about to build. What's the simplest version that tests whether my assumption is correct?"
18. "Review this PR not for bugs but for things I'll regret in six months."
19. "I'm about to refactor [system]. What information do I need before I start that I probably don't have yet?"
20. "Here are the three approaches I'm considering. Don't recommend one — tell me what question I'd need to answer to make the right choice."
21. "Write the test before the implementation. Here's the behavior I need."
22. "Explain what this code does in plain language, then tell me what it assumes — hidden inputs, environmental dependencies, things that could silently break."
23. "I'm debugging [issue]. Here's what I've tried. What's the most boring, obvious explanation I've probably dismissed?"
Prompt 23 is a pride bypass. The boring obvious thing is almost always right.
Decisions, Strategy, and Thinking (Prompts 24–30)
This is the category most people haven't explored. AI is remarkably good at structured reasoning if you give it real constraints and stop asking it to be optimistic.
24. "I'm deciding between [A] and [B]. List the assumptions I'm making that I haven't stated."
25. "What would I need to believe for [decision I'm leaning toward] to be wrong?"
26. "Here's an email I'm about to send. How will the recipient actually read this — what will they focus on that I'm not focused on?"
27. "I want to say no to [request] without burning a relationship. Help me write it so the 'no' is clear but the relationship is intact."
28. "Here's my plan. Give me the pre-mortem — six months from now, what killed it?"
29. "What decision am I avoiding by doing research instead of deciding?"
30. "I'm overthinking [thing]. What's the reversible default action I should just take so I can move forward?"
Prompt 29 is the one that stings. I use it on myself once a week.
A Framework for Actually Using These
The 5-Layer Test — before adding a prompt to your regular workflow, ask:
- [ ] Does this replace a task I currently do manually?
- [ ] Does the output require me to think, or does it think for me?
- [ ] Could I defend this output in a meeting without rereading it?
- [ ] Does it save more than 20 minutes or unlock a decision?
- [ ] Have I used it at least three times without it failing me?
If a prompt fails on item 2 — if you're just accepting what the model says — it's a crutch, not a tool. The prompts in this list are designed to force you back into the driver's seat. They generate better questions, not just answers.
Also: batch your AI interactions. Context-switching to a chat window every ten minutes kills flow. I run morning prompts in one block, writing prompts during a single dedicated session, and decision prompts at end-of-day review. The discipline isn't about the prompts — it's about protecting the time when you're not using AI at all.
How AI Handler Approaches This
The reason I'm building AI Handler is that none of the existing tools solve the actual problem. You have great prompts — maybe even a Notion doc full of them — and you still don't use them consistently because there's no system that surfaces the right prompt in the right workflow context.
AI Handler is built around the idea that your prompts should be organized by workflow state, not by topic. When you're in planning mode, it surfaces planning prompts. When you're reviewing a PR, it knows. When you're drafting something, it's already loaded with your writing preferences and past feedback. The model context follows you, not the other way around.
The thirty prompts above are powerful in isolation. They're compounding when they're part of a system that remembers context, chains outputs, and integrates across the tools you actually use.
That's what I'm building. In public. With all the embarrassing pivots included.
AI Handler is the unified AI workflow tool I am building. Launching June 2026. Email ceo@eternalsix.com for beta access.
Top comments (0)