AI-Powered Game and App Creation: Revolutionizing Development or Falling Short?

I asked AI to recreate my classic 1980s platform game, and it failed miserably, but I'm still impressed by the tech

I have a confession to make - yesterday, work in the TechRadar office reached something of a standstill because Claude, the popular AI chatbot, had added a big upgrade to its Artifacts feature in the sidebar. In Claude's own words, the upgraded Artifacts now allow you to "turn ideas into shareable apps, tools, or content -- build tools, visualizations, and experiences by simply describing what you need." Or in other words, you think of an app or game idea, describe it in prompts, and Claude will code it for you in seconds so you can try it out. In minutes, I made a playable Space Invaders copy, Lance made a very cool Asteroids copy, and Eric made what he describes as a copy of a Super Mario Brothers-style game. Claude doesn't get everything right from your first prompt, but it's easy to refine your game once it has been created. For instance, in my first go at creating a Space Invaders clone, the aliens didn't speed up as you shot more of them, which is an essential part of the game, but it's easy to ask Claude to make that happen using text prompts. It then re-coded the game and let me play again. You can keep fiddling with your game like this until you get it right. Creating games from just a description, also known as vibe coding, is all the rage these days. Back in the 1980s, there was no such thing as AI or vibe coding. We had to code things by hand, and that's what I did on my Acorn Electron home computer using BBC Basic. I spent hours of my life learning how to create games in BBC Basic, designing graphics by hand using graph paper then converting them into hexadecimal numbers and typing them in, then trying to debug the whole thing when I'd put it together, then saving the results as data onto a tape recorder and praying that they loaded back up tomorrow without anything going wrong. It was very time-consuming but also great fun, and I wouldn't have changed it for the world. I've always felt a bit sad that those days of raw creativity are gone, especially now that with AI, you can achieve similar results just by typing a few text prompts. I actually got one of my games published in a computer magazine of the time called Electron User. My game Sprog! followed the adventures of a caveman who had to jump over potholes and objects in a sideways-scrolling landscape. Your controls were left, right, and jump, and as you completed each level, things got progressively harder with more holes and objects to avoid. Back in the 1980s, computer magazines printed entire listings of the code, and people typed it in by hand so they could play your game. It seems utterly crazy today, but that's what we did! If you want, you can actually play my game using an emulator on the BBC Micro Game Archive page. It says it appeared in Electron User magazine, volume 6, issue 6, in March 1989; however, the date inside the game is 1986. (Fans of the era will recognize that my amateur game was heavily influenced by the commercially released Frak! By Nick Pelling) Considering how easily Claude could whip up a classic arcade game in minutes, I wondered if I could use it to recreate my game, and that's when I started to butt up against the limits of vibe coding. There's no way to upload the code of the game into Claude, but you can upload screenshots, so I uploaded the game's screenshots and described how it worked with the prompt: "I want you to recreate this sideways-scrolling game. The caveman should be on top of the ground, never merging through it, and able to jump off it, and move left and right by pressing the Z and X keys, and jump with the space bar. If he falls into a hole, he dies. After 1 minute, you progress to the next level with more holds and objects to jump over." Unfortunately, the game Claude created looked nothing like it and didn't play like it. The graphics that Claude uses are created from CSS blocks, so they can never accurately create something down to the pixel level. So, creating a caveman that looked like the hand-drawn one I created all those years ago proved to be pretty much impossible. I tried a few prompts to refine it, like "Make the caveman character look like the caveman in the screenshot, and make the ground scroll from right to left so he has to jump over holes that are coming towards him." But the results were disappointing. I tried a few more tweaks, but ultimately, I could tell it was a waste of time. What I was asking was beyond what Claude could do, and the game was essentially unplayable. Claude's new upgraded Artifacts are well worth spending an afternoon recreating classic arcade games, but it does have limits. Perhaps some things from my childhood are better left in the 80s, where they belong.

I recreated Asteroids with AI in five minutes, and now I think game development is about to change forever

Six months ago, I was gushing about ChatGPT's ability to build a simple Tic-Tac-Go game (like Tic-Tac-Toe, but with a lot more squares) based on a handful of prompts. That, it turns out, was child's play. Now, I've used five prompts to create a remarkable facsimile of the classic Asteroids game I played as a teen in the 1980s. It took Ed Logg two weeks to prototype Asteroids for Atari in 1979. It took Claude less than five minutes. I was inspired after hearing about Claude's new Artifacts feature, which enabled one TechRadar writer to build a personalized Super Mario-like game in seconds. Super Mario, as a game about an adventurous journalist, is very cool, but I couldn't resist trying Artifacts with one of my all-time favorite arcade games. Rarely has generative AI expectation met reality so perfectly. Here's the first prompt. I've also boldfaced some words to show you the core requests.. Note that while I explained in some detail what I wanted, I said nothing about game physics, like how the asteroids would move and react to the missiles. I didn't define how the booster would make the ship move. I defined the shape of the ship, but not the look of the asteroids. I explained how the asteroids would break up, but not what that animation might look like. It was, at best, a game design outline. And yet this is what I got. The only way that's possible is that Claude already knows what this game looks like from, for instance, the countless YouTube videos showing the original Asteroid Arcade gameplay. I quickly realized, though, that I forgot to include scoring and levels. As I wrote this next prompt, I tried to use my memory of the countless hours I spent playing at a small grocery store arcade where I worked at as a teen: "Amazing. Let's update it with scoring. Every asteroid hit is worth 50. A destroyed asteroid is worth 100. If you destroy all asteroids, you open a new level that increases the number of asteroids by 10%. You also get an extra ship. Finally, if you score over 5000, you can put your three initials in a leader board that appears at the end of every game." This time, Claude took a little longer, maybe an extra 30 seconds, to deliver a new game that I accessed via an updated URL. The scoring worked, and all was well until the game threw an error when it tried to launch the next level. The last time I was trying to use AI to code a game and I hit an error, it was basically game over. This time, though, I had another idea. I copied the error and entered this prompt: "Great, but the generation of a new level produced this error, "Uncaught Error: Uncaught ReferenceError: nextLevel is not defined" Can you debug and fix the code?" I thought Claude might stall here, but after a minute or so, where sometimes I could see its work in what appeared to be HTML (Claude Artifacts will show you all its decisions after the work is done), a new game appeared. I copied the link, opened it in a browser window, and started playing. Now I could beat the level and seamlessly move onto the next. When I cleared all the asteroids, I got, as I had directed, an extra ship. Claude, on its own, made it possible to keep collecting ships as I beat levels. I shared the game with coworkers who suggested I make it more difficult. I agreed that the game felt a bit basic and repetitive. I added this prompt: "Great! Let's have each level increase the asteroid load by 25% and make the asteroids move 10% faster in each new level." Obviously, I was guessing at how these adjustments would express themselves in the game, but Claude did not disappoint. Soon, I had a new multi-level game that did get somewhat more intense with each level. Even so, I was mostly just winning. Then someone suggested I add another element, an attacking alien ship, which would mean you were dodging and shooting asteroids while also dodging and shooting at a wily attacker. Time for another prompt: "After level 2 let's add a small, fast alien ship that will chase and fire at us. It cannot be destroyed by the asteroids, but it can destroy us. We can also shoot it and destroy it for 500 points." That, it turned out, was my last free prompt of the day with Claude, but it was worth it. As I played the new build and hit level 2, a tiny spaceship (that looked like a little, classic 1950s idea of a flying saucer) appeared and started chasing me. Claude decided on its own to break the monochromatic rule and made the alien ship's cannons shoot red missiles. The game was now much harder. It's clear to me that we'll soon have Claude-made and other AI-built games of far greater complexity appearing online and elsewhere. The authors? AI. The credit, though, should go to the countless human developers who built real games over the past five decades, and whose work is now being hoovered up for training these systems and remixed on the fly with nothing but a prompt.

I Tried Out Claude AI's New App Creator, and I Didn't Have to Use Any Coding

David Nield is a technology journalist from Manchester in the U.K. who has been writing about gadgets and apps for more than 20 years. Last year, Anthropic rolled out a new feature called Artifacts in its Claude AI chatbot, letting users see real-time results from their conversations -- think diagrams, reports, websites, or code snippets, for example. Now the feature has been given a substantial upgrade, letting you create and share AI-powered apps just by describing them. The apps you create through this "vibe coding" approach can have Claude AI smarts built right inside, and it's available for all users -- so you can give it a try whether you're on a paid subscription or not. And if you share your creation with other people, when they run the app it'll be counted against their Claude usage allowance, not yours. One of the examples Anthropic gives is a flashcard app. You could previously get Claude to code you an app showing flashcards, but you'd need to specify all the details and card content yourself. Now, you can leverage Claude's AI to create card content specific to individual users, because they'll be able to give the chatbot prompts about the text they want to see. I haven't done a whole lot of coding since I learnt BASIC back in high school: I used it to code a soccer simulator that picked scores and scorers at random, which was a lot of fun at the time. With Claude's new and improved Artifacts tool available, I was interested to see just how good an app I could make from vibes alone. My first thought was to ask Claude what I should create, a worrying reminder that AI is taking away our capacity to think and make decisions ourselves. In the end I used my own brain cells to make a choice: I wanted to make an app to generate quiz questions on any specified topic, with a sliding scale of difficulty and the answers available if needed. I explained what I wanted using natural language and zero code: an app that would prompt the user for a topic, then display a question on that topic, with the option to reveal the answer with a click. I also wanted options for adjusting the question difficulty, and switching topics. It actually only took a single prompt to describe everything. Claude gave me an overview of what it would do before it started building, telling me it loved the vision for my app (I bet it says that to everyone). When I gave the go-ahead, the browser window split to show the app and its raw code on the right (you can toggle between these views with a click). The resulting app was pretty great the first time around, certainly in terms of its functionality -- everything worked as I wanted and was well designed. The AI's choice of questions and assessment of their difficulty was a little wonky, however: All the questions had the right answers attached, but the app tended to fixate on a limited set of subjects within each topic (specific movies, actors, or bands, for example). It was probably unfair for me to ask for a question about Twin Peaks at child-level difficulty, but "where is Twin Peaks set?" is maybe a little too easy -- and Claude generated the same question at the next difficulty level, too. The hardest question was about the model of tape recorder FBI agent Dale Cooper uses, which had me stumped. That's more about the limitations of AI models and their understanding, though. In terms of the actual app building, Claude mostly impressed: It responded to my questions and edits very well, letting me change colors and layouts with simple text prompts. It's cool to see an app being rebuilt in real time, and the AI bot then explains the changes that have been made and why. There was one issue with a new "suggested topics" box that Claude took a long time to get right, despite my vibe-debugging attempts. Eventually, it got the box in the right place, but it took multiple iterations to work properly. At times the chatbot interface is too opaque, though the raw code is always available for the serious programmers out there. For now, this seems best suited to casual, small-scale projects -- as demonstrated by the examples Anthropic has put in the showcase -- but I can see it being useful for people who want to develop custom tools for admin or productivity. I definitely enjoyed the experience, and felt like I learned quite a bit about coding along the way too, even if the end result was a little glitchy (more down to the AI content than the actual code).