🌈 ESP32-S3 Rainbow: ZX Spectrum Emulator Board! Get it on Crowd Supply →
View All Posts
read
Want to keep up to date with the latest posts and videos? Subscribe to the newsletter
HELP SUPPORT MY WORK: If you're feeling flush then please stop by Patreon Or you can make a one off donation via ko-fi

I recently decided to learn how to solve the Rubik’s cube. As a self-confessed geek, it felt like something I should be able to do.

I found a pretty good website that breaks it down into eight steps, from “the Daisy” all the way through to “Finish Him!”. I memorised the moves. Pretty soon I could solve it without looking at the website.

It was incredibly disappointing.

Rubiks be gone!

I was just mechanically following “algorithms” I’d memorised. No skill, no insight. Pattern match, apply algorithm, repeat.

So I thought, this can’t be how speed cubers do it. Steps 1-3 are painfully tedious. There must be a faster way.

This led me to F2L, solving the first two layers simultaneously. The first resource I found suggested memorising 42 algorithms - my response - yeah, I’m not doign that…

Then I found a few suggestions on solving F2L “intuitively”, which sounded more my style. Maybe this approach would actually involve some thinking.

It’s definitely faster and requires a bit more brainpower, but ultimately? Still pattern matching. I wanted to understand why the algorithms work, not just memorise better ones.

Frustrated, I did what everyone does now. I asked ChatGPT.

After the obligatory “That’s a brilliant insight!” it gave me this analogy:

  • Beginner: following GPS directions
  • Intermediate: memorised routes
  • Advanced: understand the map
  • Expert: improvise new routes instantly

I keep coming back to this because it maps so well onto the junior-to-senior engineer journey.

You start out following instructions. At some point you start pattern matching. This problem looks like that problem I solved last year. Eventually the memorised playbook isn’t enough and you start synthesising new solutions from first principles. You develop taste. Judgment. The ability to look at a system you’ve never seen and know where the problems are going to be.

That last part is what I find interesting about coding agents right now. They’re very good at levels one and two. Following instructions, applying known patterns at scale, faster and more consistently (well, most of the time…) than any person could.

And sometimes watching an agent churn out code gives me that same Rubik’s cube feeling. It works. It passes the tests. But I’m just sitting here approving it. What am I even doing here?

But then you hit a problem where the agent can’t see what you see. It solves the ticket but puts the logic in the wrong place. It writes code that works today but will be a nightmare in six months. And you remember what you’re doing here. That’s the difference between mechanically producing a solution and deeply understanding the solution.

For now, experienced engineers are the ones providing the taste and judgment that makes the difference between code that works and code that works well.

I’m never going to be a speed cuber. But maybe, with the help of agents, I might end up being a speed coder…

Related Posts

AI Tools in Coding: Progress or Just Optimization of the Status Quo? - Pondered the future of coding with a colleague of mine recently, as one does. Couldn't help feeling a tad underwhelmed by the programming tools like GitHub's Copilot X and ChatGPT. Sure, these AI-powered coding buddies are smart, helpful and do save time, but they more or less tinker around the edges - optimize the existing processes but don't fundamentally alter them. Software development still remains a tedious enterprise - a loop of problem-definition, code-generation, code-validation, and feedback. Not so different from what it was 30 years ago really. We're surrounded by subpar software everywhere - the bugs, the slowness, the vulnerability. My, what a mess! We need a massive, transformative solution. Imagine an AI system that could take high-level concepts and materialize fully functional apps from them! We've seen glimpses of such capabilities in ChatGPT, generating apps from sketches, but that's just it - glimpses. Still a long way off. My hope for the future? An era where writing code is antiquated. We instruct AI to generate, tweak, even test code - on a constant iterative feedback loop. Maybe, what we know as 'code' could be on the cusp of extinction, replaced by something machines create and understand - and we don't. A brave new world, indeed!
Vibing a PCB - surprisingly good - In my latest adventure, I challenged AI to design a working ESP32-S3 development board from scratch using Atopile and Claude. The idea was as simple as vibe-coding actual hardware without diving into the code. It was a chaotic yet fascinating journey, with some misses like unwired components and a forgotten capacitor. After a few prompts, the AI delivered a surprisingly functional board featuring USB-C, an AMS1117 regulator, and status LEDs. While not yet perfect, vibe-coding feels like a glimpse into the future of hardware design.
Failing to solve Manic Miner with RL - I was never any good at Manic Miner, so I asked Claude Code to build something that could finish it for me on my ZX Spectrum emulator. We started with reinforcement learning, and failed. What actually worked was search - first on the emulator, and then on a model of the game Claude built from the disassembly.
Watching LLMs Think - I learned early in my almost-PhD: don’t watch the program run. Staring at loss values—or an agent’s “Thinking…”—feels like work, but it isn’t. Kick off the job, step away, and use your brain for the human stuff.
Augmented Reality In The Browser - This blog showcases the progress of my idea to create an augmented reality Sudoku solver using technology that enables us to solve puzzles in our browser rather than with dedicated apps. I have developed an AR Sudoku solver with a simple image processing pipeline. It identifies and extracts Sudoku puzzles from pictures, recognizes each cell's numbers, solves the puzzle, and renders the solution on top of the original image. This process is accomplished by converting the image to greyscale, conducting thresholding, OCR-processing, and puzzle-solving. I've also done parallel image location and extraction. It's a technical journey that's achieved some high-quality, low-drift results, visible in the video included. The full codebase is accessible on Github. Enjoy the joy of leveraging technology to efficiently solve Sudoku puzzles!

Related Videos

Surprisingly not CR*P - Vibing hardware - Well folks, what an adventure! After vibe coding our vibing button, I took up the challenge to vibe code the actual hardware for an ESP32 S3 dev board. Used a tool called at pile and gave it a whirl with some AI help from Claude. Interestingly, it handled everything from sourcing components with LCSC part numbers to wiring them up. Had a wee hiccup with missing capacitor in the RC circuit, but overall it was a thrilling ride watching the AI process! A bit of tweaking here and there, and we're good to go! Can't wait to see it in action.
ESP32 SD Card Speedup With a Couple of Lines of Code - In this video, we explore the disappointingly slow data writing speed of the ESP32 when reading and writing to an SD card in our TinyTV project. With 500 kilobytes/sec reading and a dismal 270 kilobytes/sec writing, we embark on an adventure to find a solution. After ditching the Arduino code in favor of IDF functions, we discover incredible improvements. Seeing potential risks, I propose a truly bonkers plan: using a IC to interface SD cards with USB with a USB multiplexer switch and another switch to alternate between ESP32 and the GL823. This could be a total disaster, but I'm game for the challenge. Stay tuned to see if it works out!
ESP-IDF 6 Setup in VS Code: Blink & Debug an ESP32-S3 - Time to get with the program and jump to ESP-IDF v6.0.1 in VS Code. I show you how to install the prerequisites (I’m on a Mac with Homebrew), run the ESP-IDF Installation Manager, spin up a Blink example for an ESP32-S3, tweak menuconfig (GPIO21, 500 ms), then build and flash—switching from stubborn USB JTAG to trusty UART when things fail. We check the monitor, bump CPU to 240 MHz, fix flash to 16 MB, peek at PSRAM, and give PCBWay a quick shout-out.
I Feel the Need – The Need for Hardware SPI… - An insightful iteration on my Arduino Nano esp32 video. Despite criticism regarding the slow display update speed, a solution was found thanks to the helpful fellow, Nick. Turns out, the software SPI was the cause of the issue. A quick tweak in the code and voilà, we've got ourselves an SPI clock whizzing at 80 megahertz. Quite the speed boost for just a few lines of code alteration!
Arduino & C++: Blink Sketch Reinvented with Inheritance & Composition - Learn how to create a single Arduino sketch for multiple dev boards using inheritance and composition in C++, improving code readability and maintainability. This tutorial demonstrates the principles by implementing a multi-board LED blink sketch and comparing code sizes to ensure efficiency.
HELP SUPPORT MY WORK: If you're feeling flush then please stop by Patreon Or you can make a one off donation via ko-fi
Want to keep up to date with the latest posts and videos? Subscribe to the newsletter
Blog Logo

Chris Greening


Published

> Image

atomic14

A collection of slightly mad projects, instructive/educational videos, and generally interesting stuff. Building projects around the Arduino and ESP32 platforms - we'll be exploring AI, Computer Vision, Audio, 3D Printing - it may get a bit eclectic...

View All Posts