Story 1
When AI Learned to Take a Day Off

A case from the MARVEN Prompt Lab
By Alexey Danilin

​

How MARVEN and I built a course

together — and taught the planner to rest.

​

The Context

​

Inside MARVEN, we were building a plan generator — a tool that turns ideas into structured, time-spaced tasks.

My goal was to test it in real life: to help my 12-year-old daughter Darya start learning Python.

​

In chat, MARVEN and I designed a small course together — “Python for Little Engineers.” It was supposed to be a friendly plan: simple code, small victories, and daily reminders sent to her via Telegram — MARVEN likes to keep in touch when you’re away from the browser.

​

Once the course outline was ready, I took the generated text and dropped it into the task auto-generator — the same one we use for any "adults" projects and campaigns.

 

The Prompt

​

Create a set of tasks from the text below. Distribute them as one task per day.  
Each next lesson must come "every other day", not earlier. 

If a lesson would fall on a weekend — move it to the next workday, and so on.  

DO NOT cram everything into one day! This is done by one person who has other things to do.

Each lesson should take about "1 hour".  

Keep in mind: the student is "12 years old" — a beginner, learning slowly but with curiosity.

​

TASKS:

​

Mini-Course: “Python for Little Engineers” (1)

 

Level 1. First Control of the Machine  
Goal: feel that a program is not just text — it’s an action.

 

Task 1:

 
# hello_dasha.py  
print("🚀 Program started!")  
print("Hi, I'm Python. What's your name?")  
name = input("👉 Type your name: ")  
print("Nice to meet you,", name, "!")

 

Skill: launching, input, strings.  
Mission: “You’ve made the machine talk to you.”

​

Level 2. Engineer Measures Roblox Speed  
Goal: learn variables and arithmetic.  

 

Task 2:

 
# roblox_speed.py  
vpn_ping = int(input("Enter VPN delay (ms): "))  
fps = int(input("Enter Roblox FPS: "))  

roblox_speed = 10000 / (vpn_ping + 1) * fps / 60  
print("💡 Effective Roblox speed:", round(roblox_speed, 2), "units of power!")

 

Skill: variables, math, input().  
Mission: “Measure how strong your Roblox runs today.”

​

Level 3. Colorful Terminal (ANSI Magic)  
Goal: show that the terminal can be “alive.”  

 

Task 3:

 
# color_test.py  
print("\033[1;31mRed text\033[0m")  
print("\033[1;32mGreen text\033[0m")  
print("\033[1;34mBlue text\033[0m")  
print("\033[1;35mPurple text\033[0m")  
print("\033[1;33mYellow text\033[0m")  
print("✨ Engineering magic!")

 

Skill: escape sequences, liveliness effect.  
Mission: “Make the engineer’s console colorful!”

​

Level 4. Engineer-Simulator  
Goal: simple conditions and loops.  

 

Task 4:

 
# fan_control.py  
print("💨 MARVEN Fan Control System")  
temp = int(input("Enter CPU temperature (°C): "))  

​

if temp < 50:  
    print("🟢 All good. Fan off.")  
elif temp < 70:  
    print("🟡 Medium load. Fan at 50%.")  
else:  
    print("🔴 Overheating! Fan at maximum!")  

​

Skill: if / elif / else.  
Mission: “Build a cooling system for your Roblox-Mac.”

​

Level 5. Mini-Game “Guess the Number”  
Goal: while loop, logic, random generator.  

​

Task 5:  

 

# guess_game.py  
import random  

​

print("🎯 Guess the number from 1 to 10!")  
number = random.randint(1, 10)  
attempts = 0  

 

while True:  
    guess = int(input("👉 Your guess: "))  
    attempts += 1  

 

    if guess == number:  
        print("💥 You guessed it in", attempts, "tries!")  
        break  
    elif guess < number:  
        print("📈 Too low!")  
    else:  
        print("📉 Too high!")  

 

Skill: import, while, break.  
Mission: “Tame randomness.”

​

First result

​

MARVEN enthusiastically produced the plan… but scheduled everything for one day.
Five tasks back-to-back, all in a single work session 🤔

For a senior engineer — fine.
For a 12-year-old — mission impossible.

​

No matter how I rephrased “daily lessons,” the AI kept compressing everything into a single timeline.

It was technically correct, but humanly wrong.

​

Solution

​

So, we (yes, MARVEN and I) decided to teach the planner what learning rhythm means.
At the end of the prompt, I added a small “injection” — just a few lines of scheduling rules:

​

Scheduling rules (extra):

• For study plans with children: schedule at most ONE item per calendar WORKDAY.

• Apply a GAP of 1 workday between consecutive items ("every other day").

• If duration is ~45 minutes, you may set duration_hours=1.

​

That’s it. No backend patching, no hard-coded constraints — just a well-crafted prompt.

​

The Final Result

​

MARVEN instantly adjusted:

​

• one task per day,

• every other workday,

• no weekends,

• each lasting about 45 minutes with a quiz and a mini-challenge.

​

Even the first lesson — “Hello, Python!” — came out perfectly balanced: mood check, code, and a playful quiz.

The Telegram reminders started arriving like real “lesson bells.”
Darya loved it ❤️

 

The Insight

 

AI can understand not only commands but the rhythm of human life.
The key is to define the cadence clearly.

The phrase “one task every other workday” worked better than any parameter.

 

That’s the power of a prompt — to make a machine not just calculate, but breathe with the human tempo.

​

​

About the Author

​

Alexey Danilin is a systems engineer, researcher, and creator of the MARVEN project.
He works at the intersection of cognitive science, technology, and philosophy.
He believes that the next great challenge of engineering is not to build smarter machines, but to create wiser connections between human and artificial intelligence.

​