Abhinav pivoted from a career in banking to pursue his first love in writing. Even while working full-time, he continued contributing as an editor-at-large, a role he has held for more than 7 years. A lifelong tech enthusiast who has built three gaming and productivity powerhouse PCs since 2018, his passion for technology keeps him closely following the semiconductor industry, from NVIDIA and AMD to ARM. His MSc dissertation explored how artificial intelligence will reshape the future of work, reflecting his curiosity about the wider social impact of emerging technologies.

I was born in 2000, which incidentally happens to be the year Nokia released the 3310. Long before the phone became an internet meme for being indestructible, it was simply a solid, affordable phone for its time, and naturally, my parents owned one.

I don't remember much about the 3310 itself, but I do remember the games that came preloaded with it. In 2003, my three-year-old self would take every opportunity to pick up the phone when I found it lying around the house and play it for hours. As such, my memory of those games are admittedly hazy. I remember what they felt like more than how they worked.

This got me a little curious, so I decided to see if Claude Fable 5, one of the most capable consumer coding models available, could recreate them from nothing more than my childhood recollections, and it's this experiment that's what made me realize the genius behind them.

Space Impact was one of the greatest mobile games of all time I didn't say it, CNET did back in 2010

Space Impact was undeniably one of the reasons I would often nick my dad's old 3310 from his desk, but it wasn't until recently until I realized that the impact the game had on me as a three-year-old was shared by many back in the day. Released in September 2000, the game came preloaded on the 3310 and several of Nokia's lineup of monochrome phones later.

The game was developed for Nokia's Series 30 platform and written natively in C, challenging its developers to squeeze as much as possible out of the DCT3 microcontroller architecture while working within its tight memory constrains, all while providing an enjoyable gameplay experience. In 2010, CNET named it on their list covering the ten greatest mobile games of all time, and I couldn't agree more.

Claude Fable 5 wrote the game in Python using Pygame, and for some reason, decided to name it Void Runner. Within five minutes or so, it was ready to run, and it felt familiar almost immediately upon launch.

The side-scrolling structure was exactly similar to the game I grew up with, and so were the familiar power-ups (which included a shot splitter and shields, to my surprise). Perhaps the most striking aspect of it all was the fact that the boss fights all seemed to follow patterns that triggered memories I hadn't thought about in years, which made me recall my childhood strategies for getting past them, which was to stick to a corner and only fire when the boss moved directly into my line of fire.

This was enough to know that Fable 5 had indeed, been successful in recreating the game I had remembered, at least the most significant aspects of it. I couldn't help but notice, though, that it also felt like something reconstructed from memory, which was reasonable, as it was exactly something I vaguely described. Not exactly Space Impact, but perhaps close to Void Runner.

Snake II was a clever twist on the classic snake I could play this all day

Snake II was rather different from the classic Snake game that everyone knows and loves. You still collect orbs and grow longer by consuming each, but the losing condition, as I remembered, was a little more forgiving. Hitting a wall didn't end the game, but instead, you'd emerge from the opposite side and continue collecting until you eventually ran into the snake's body.

Nokia had other ways of making the game more challenging. You could increase the snake's speed before launching, which would make every turn more consequential, while different mazes as options would add more obstacles and force you to pre-empt every move before diving into a wall.

Those were all the details that I didn't provide to Fable 5 with my prompt when I asked it to create Snake II with Pygame, and yet Neon Sprint (as it was named) incorporated every single one of those details. Naturally, giving the model the name of the game had been enough for it to draw on what it already knew about the game, but the result still made me reconsider how I had remembered it myself. I'd reduced it to a game about collecting dots, especially when the developers had layered speed, level design, and forgiving traversal into an incredibly basic, yet highly engaging premise.

Fable 5 one-shotted every game I threw at it Even when I remembered the games, and not how they worked

It's reasonable to say at this point that Fable 5 one-shotted every game I threw at it. In the past, I've applied the same "design from memory" methodology to several other models, including Gemini 3.1 Pro, GPT 5.5 and local models like Qwen3-Coder 30B. Almost every one required some degree of hand-holding, particularly when it came to translating the game's mechanics and visuals into something playable.

In a straightforward comparison, Fable 5 consistently got the logic right on the first attempt. It also made some surprisingly great decisions about how those mechanics should be incorporated, even if the visual interpretation and artistic choice was a world away from the originals. The last bit, by no means, is criticism, though. I could describe in detail how Space Impact played, but I couldn't define what the pixels looked like because I barely remembered them myself.

The games were simple, because the developers were anything but

Funnily enough, Fable 5's successful attempt at recreating these games made me appreciate them more, not less. Recreating is easy once the design is understood, but the genius was in arriving at that design in the first place, with hardware that gave their developers almost nothing to work with, with just kilobytes of memory at their disposal.