CodeWalk: Discovering Computer Science in the Physical World with Local Gemma

Dhanush A G — Sun, 11 Oct 2026 18:55:21 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

"Computer science is no more about computers than astronomy is about telescopes."

— Edsger W. Dijkstra

What I Built

The Problem: Trapped Behind Abstract Glass

As software developers, computer science students, and tech enthusiasts, we spend thousands of hours isolated behind high-resolution monitors. We grind abstract LeetCode algorithms, debug distributed architectures, and optimize virtual databases—frequently forgetting that the core axioms of computer science were inspired by observing real-world systems, physics, and human behavior.

When was the last time you stepped away from your IDE to watch a physical queue, trace a real-world bottleneck, or witness how an unsorted library shelf mirrors linear search?

Introducing CodeWalk 🌱

CodeWalk is an open-source, local-first companion application that challenges developers to close their laptops, put their smartphones away, and step into the physical world to observe foundational computer science concepts in action.

Whether it's studying:

Data Structures & Algorithms: FIFO Queues at bakery counters, LIFO Stacks at tray return stations, or Binary Search across street house numbers.
Systems & Architecture: Bottlenecks at single-lane construction merges or Priority Scheduling at pedestrian crosswalks.
AI & Data Principles: Anomaly Detection in fruit displays or Sampling bias in commuter flows.

Designed to Minimize Screen Time

CodeWalk is deliberately architected so that the screen is the shortest part of the entire experience:

Choose a Concept & Time: Pick a CS category and set your outdoor time (5, 10, or 15 minutes).
Receive Your Physical Mission: An open-weight Google Gemma model generates a safe, targeted real-world observation objective.
"PUT YOUR PHONE AWAY": The app locks into an ambient countdown reminding you to pocket your phone, step outside, and look around.
Record Observations: Return to your desk and type what you saw.
Real-World to CS Theory Mapping: Gemma analyzes your observation, evaluates whether you spotted the principle, maps each physical actor to its software/hardware counterpart, and suggests your next outdoor walk.

Demo

🚶 Field Test Story (Taking It Outside)

To test CodeWalk in the wild, I selected Concept: Queue (FIFO) with a 10-minute timer and walked to a nearby coffee shop on the corner of the street.

The app instructed me:

"Walk to a nearby coffee shop counter or bus stop. Stand at a respectful distance, put your phone away, and observe how people arrive and leave for 10 minutes. Notice: Where do new people join? What happens when an order takes extra time?"

I pocketed my phone, grabbed a bench 15 feet from the register, and observed:

Customers joined strictly at the rear (enqueueing at the tail).
The cashier processed one transaction at a time from the front (dequeueing from the head).
When one customer placed an elaborate order with five custom drinks, the line behind them swelled from 2 to 9 people—a vivid physical manifestation of Head-of-Line (HoL) Blocking and single-threaded backpressure!

When I returned and submitted my notes, CodeWalk gave me:

Verdict: STRONG_MATCH (95% Confidence)
Real-World Mapping:
- Customers in line → Buffer / Queue elements
- Cashier register → Single Consumer thread
- Complex 5-drink order → High-latency task causing Head-of-Line blocking
CS Learning Point: In network packet routing and FIFO queues, head-of-line blocking happens when the lead packet suffers processing delay, stalling all subsequent packets regardless of their payload size.

DEV Community: Dhanush A G