Author: Sean Hollister, the Verge; Translator: Peggy, BlockBeats
Google AI Studio is bringing AI programming into a more intuitive phase: users are no longer just having models "write code," but can directly generate an Android app using natural language and install it on a real phone within minutes. From typing prompts in a browser to Gemini automatically generating code, designing interfaces, fixing bugs, and finally the app appearing on the device, the barrier to entry for software development is being further lowered.
The Verge author Sean Hollister recently experienced Google AI Studio's "prompt to phone" capability. He created three apps in one afternoon, including a text adventure game, a calorie calculator, and a Mario-like game, requiring almost no coding himself, and some bugs could be quickly fixed by continuing the conversation. This experience shows that AI programming tools are moving from development environments to consumer scenarios that are closer to ordinary users.
This is also where the most imaginative aspect of the "personal software revolution" lies. In the past, ordinary users could only wait for developers to create general-purpose products; now, they might be able to generate a fitness tracker, a calorie calculator, or even a simple game based on their specific needs. For Google, this is not just a demonstration of AI programming capabilities, but also a potential new entry point for Gemini into the mobile platform, developer ecosystem, and subscription revenue. However, this experience also illustrates that AI-generated applications are still significantly from true maturity. It can quickly create "running" programs, but it may not be able to create reliable, accurate, or user-friendly products: game narratives are crude, mechanics are weak, calorie data can be seriously misjudged, and even Mario-like games may repeatedly crash. More complex issues include copyright boundaries, data sources, product judgment, and long-term maintenance capabilities. What truly deserves attention is not whether AI can already replace developers, but that the starting point for software production is changing. Google has proven that ordinary people can create mobile apps using prompts; however, the transition from "generating an app" to "creating a good app" still requires human expertise, aesthetic judgment, and continuous iteration. AI can significantly accelerate development speed, but the final mile of software quality remains unresolved. The following is the original text: Yesterday, I created my first Android app. Then, I created two more—three in one afternoon. For one of these apps, I simply typed 148 English words into a web browser and left. Ten minutes later, a completely new app appeared on my actual Android phone. Of course, before that, I did need to prepare the phone: enable USB debugging mode and connect it to the computer. But apart from that, as Google advertises, AI Studio almost did all the work for me. I typed in the text, clicked install, and—voilà—a fully functional program appeared. At that moment, I was almost ready to agree with David, Allison, and Jen's assessment: the personal software revolution has arrived, and it's entering your phone. In the future, even ordinary people without programming skills may be able to make complex smart home systems truly run. Then, I started actually using the three apps: a calorie counter and two games. They didn't perform very well. And just as I was beginning to enjoy iterating and trying to improve them, AI Studio reminded me that I had reached my daily usage limit. Next, I either had to pay or wait for my limit to be restored. So, the friction is still there. But undeniably, what individuals can accomplish today is quite astonishing. That same morning, my colleague Stevie Bonifield also created a personal fitness tracking app, and he believes it's good enough to be truly usable. When Gemini popped up a paid upgrade prompt, my first reaction was, "Should I pay for a few months of trial use?" This wasn't the reaction I would normally expect from a Google product. On Tuesday, when Google demonstrated how AI could create a Doom-like game, we joked that I should make a game called MOOD. It would be a Doom-like text adventure game, with MOOD standing for "Modern Online Oratory Dungeon." This information alone was enough for Google to get to work. When I typed into AI Studio, "Make me a Doom-style text adventure game called MOOD, where MOOD stands for Modern Online Oratory Dungeon," Gemini began automatically adding more ideas, trying to continue my creative process. It first entered a sentence: "This game should have procedurally generated levels and challenging turn-based combat." Gemini is trying to autocomplete my app idea. Image source: Google. I don't want randomly generated levels where every level is completely different—I want a classic text adventure game where players explore a designed environment with a realistic map structure. Turn-based combat is acceptable, though. Perhaps the game could even have AI automatically generate the map for me? Gemini then suggested more features like "secrets hidden in rooms" and a "satisfying progression system." Most of the time, I just nodded in agreement. Before I let it officially start writing code, the final prompt looked like this: [Image source: Google](https://img.jinse.com.cn/7470885_watermarknone.png) ... A minute later, it had generated five design prototypes for me: Twenty minutes later, I pressed the "Install" button to transfer the game to a Pixel 9 phone. Unsurprisingly, the copywriting was terrible. There was absolutely no trace of demons in the game. The entire dungeon only had 11 rooms, and players could simply "clear" it by frantically pressing the attack button. If you play seriously, you can finish it in under a minute. At least for now—before that, Gemini fixed two serious bugs that would have made the game unplayable. Here's the actual effect of MOOD: I wasn't too surprised to find that Gemini's promise of "engaging narrative, branching dialogue options, and multiple endings" ultimately condensed into a simple branch at the end of the game: I can defeat "Core Orator"—an AI that somehow manages to convert internet anger into corporate profits—by attacking, merging with it, or entering a backdoor password. Furthermore, the game actively reveals all the promised "secrets" to the player: it presents these as glowing buttons, eliminating the need for players to type anything. When you encounter a glowing treasure chest, the game goes to great lengths to remind you that it's actually a Mimic—the classic monster disguised as a treasure chest from Dungeons & Dragons. It not only explicitly warns you of the "risks of checking treasure chests," but also directly marks it as an enemy and prevents you from leaving, with the system message: "A hostile 'Clickbait Mimic' is blocking your way!" Speaking of which, MOOD will even tell you the backdoor password to unlock the hidden ending when you need it. However, the bug fixing process is surprisingly smooth, provided that the bug is one that Gemini can correctly identify. When I told it that the game would freeze during a conversation with "The Whistleblower" because the end-of-conversation button was missing, it immediately generated a new version of the app. I pressed "Install," the app on my phone automatically restarted, and when I re-entered the game, I found myself back where I had left off—only this time, the button I needed had reappeared. My other apps probably still need more polishing. The calorie counter's best way of determining the calories of a certain food was to call the paid Gemini API, which I didn't have. When I asked it to search for relevant information in other databases instead, I found that its calorie estimates for many foods were seriously underestimated. However, when I told Gemini that a 16-ounce cup of bubble tea couldn't possibly only have 190 calories, it seemed to have indeed found that basic error in its code. Previously, it assumed "milk" was sufficient to match "boba milk tea," and worse, it used low-calorie 1% low-fat milk as its estimation basis. Gemini claims it will now make more reliable matches. Even so, my 3-ounce serving of Taiwanese fried chicken was just counted as 140 calories, and I'm fairly certain the actual calories are at least double that. So, this app clearly needs more work. Finally, and least importantly, I felt it necessary to test whether Google still allows users to create those terrible Nintendo knock-off games, like my colleague Jay Peters did earlier this year with Project Genie; or whether Google has learned its lesson. With deep shame, I present to you—Super Peach Rescue: Three screenshots showcasing a truly awful game. Image source: Google. This is an absolutely terrible program. Princess Peach is depicted as some kind of terrifying, one-eyed, floating alien, and the game crashes instantly whenever she touches any of the power-up blocks—every single time. Gemini hasn't been able to figure out why. Furthermore, the second pipe is completely impassable because Princess Peach can't jump that high.

Thank you, Google… is this what you call “help”? Image source: Google
Nevertheless, Gemini didn’t hesitate to generate such a game. My requirement was: “Make a playable Super Mario game where I play as Princess Peach rescuing Mario, and it has all the elements of a traditional Mario side-scrolling game.” In a sense, it did.
It even proactively suggested that I might as well “add a series of classic Mario items to Princess Peach, such as the Super Mushroom, Fire Flower, and Invincible Star.” It even automatically labeled the control scheme as “NES System.” I think I will delete this game. At least one of the two games I made with Vibe Coding was playable from the start, requiring almost no effort on my part—unless, of course, I take into account the psychological trauma I feel when I think about how many game developers are currently unemployed. To be clear: I'm actually quite glad that the games I made with Vibe Coding were of poor quality. For a completely free, custom-made calorie counter, I might be able to justify myself: after all, no one would make such a tool specifically for me. But when it comes to games, I'd rather spend my time supporting genuine human creators.