Prompt Proposal

#285

I was voice journaling with Claude yesterday, talking through a technical question for a CrossFit workout, and I realized that when I started being more specific with my ask, the responses turned more helpful. I clearly stated what I wanted to accomplish, what I'd already tried, what felt stuck, and this made an increasingly frustrating Q&A session into more of a helpful brainstorming discussion with deeper insights.

Gathering info on how friends and family use GenAI, I learned that some approach it like a vending machine. Slot in a question, get out an answer. I faced this in my early adoption days too.

Me: "How do I get better at pull-ups as a beginner?"

AI: "Practice consistently. Strengthen your back and shoulders. Use resistance bands. Increase volume over time."

Me: "Okay but like, how many times a week should I actually train them?"

AI: "Three to four times per week is standard for most beginners."

Me: "Should I do them on consecutive days or spread them out?"

AI: "Spreading them out allows for better recovery."

Me: "What about my diet? Does that matter?"

AI: "Yes, protein intake supports muscle recovery."

Me: "How much protein though?"

AI: "About 0.7 to 1 gram per pound of body weight daily."

Similarly, when I ask questions around time and life management, I very quickly get frustrated because of not getting exactly what I want. The conversations escalate as:

Me: "OMG can you give actual helpful advice on how to not burn out?"

AI:

• "Prioritize self-care
• Take regular breaks
• Set boundaries with work
• Practice mindfulness"

Notice the pattern? Generic prompts get generic, vague responses. Frustrated prompts get bullet points that don't help. Curt prompts get curt answers.

You get back what you give. 

These rookie exchanges reminded me of an Akbar-Birbal story from Indian folklore, specifically about why Birbal remained the emperor's favorite. 

The story goes:

A jealous minister approached Akbar: "Why do you always choose Birbal for important tasks? Give me a chance to prove myself." Akbar agreed. He sends him to investigate a commotion near the kingdom.

The minister returns: "There's a carnival happening outside."

Akbar: "How long is he gonna be here?"

Minister: (Runs back) "One week."

Akbar: "What's causing all the noise?"

Minister: (Runs back) "A traveling is performer with drums."

Akbar: "Is anyone doing anything about it?"

Minister: (Runs back) "No one's complained yet."

And so on. Multiple trips. Multiple direct answers to each separate question. No anticipation or further thought whatsoever - what does the king actually want to know and why? How to make the exchange more efficient? 

Birbal shows up later. Akbar asks the same of him.

Birbal: "A traveling performer arrived this week. He'll be here seven days before moving to the next town. He's playing drums at the carnival, which is loud. I checked, no one's complained yet. But I also learned he's performing to raise money to support his family (his wife and son), so he might be open to adjusting his schedule if the noise becomes a real issue."

One answer. Everything Akbar asked for, plus context Akbar didn't even know to want. The other minister immediately realized why Birbal was the favorite.

I applied similar logic to prompt better and leverage AI for learning:

CrossFit workout:

Me: "I've been doing CrossFit for three months. I can do five strict pull-ups in a row, but my form falls apart after that. My shoulders get fatigued before my grip does. I've been training pull-ups three times a week. What I'm trying to figure out is whether I should focus on shoulder strength work first, or keep practicing pull-ups and adjust my training volume."

AI: "Your shoulder fatigue before grip fatigue tells me your shoulder stabilizers are the limiting factor, not your pulling strength. Try this: add two dedicated shoulder stability sessions weekly, focus on scapular work and horizontal pulling. Keep your pull-up practice to twice a week for now, but prioritize quality over volume. Once your shoulders stabilize better, you'll see form hold longer."

Nutrition course:

Me: "I'm studying nutrition. The textbook explains that carbs convert to glucose, which powers muscles. But I'm a visual learner and I'm confused about the timing part. Like, when exactly does glucose get used versus stored as glycogen? Does it matter if I eat carbs before or after my workout, and why? Can you explain it simply?"

AI: "Timing matters because of insulin sensitivity. Post-workout, your muscles are primed to absorb glucose directly into muscle cells as glycogen, not store it as fat. Pre-workout carbs give you energy for the actual effort. If you eat carbs at random times with no training nearby, they're more likely to be stored. Think of it like this: post-workout is like an open door to your muscle cells. Pre-workout is fuel in the tank before you drive."

If the nutrition prompt was in the same chat session as the CrossFit one, Claude would even tie those threads: "So when you do pull-ups and eat carbs post-workout, your muscles absorb that glucose directly as glycogen to repair and rebuild the shoulder stabilizers you just worked. You're not just eating rice, you're refueling exactly what you broke down."

A note for people new to AI tools (this works with Claude, ChatGPT, Gemini, or Perplexity): 

The practice matters more than the tool.

When you show up with thought already happening and share specific instructions on what outcomes you're looking for, the AI can actually think with you. When you hand it vagueness, it hands you back a generic, non-useful blob.

Yes it takes longer to create those prompts. But in reality, those save time (and tokens!) You could even try using voice over typing to get the job done faster.

Be like Birbal.