Software Engineering related:
In the department of Engineer Experience and Lifestyle, Google announced Gemini 2.5 Pro (daddy model) & Flash (baby model), these 2 are the beasts of the first half of 2025! 💼💻💪 The models beat everything else in the benchmarks.
I wanted to know how good they are so I did a little experiment, where I gave a simple prompt to gemini 2.5 pro and did the same thing with OpenAI's best model out there o3 and the results from gemini are freaking incredible! 🧪🆚🔢
I asked both gemini and o3 to build me a simple clicker game, using the idle clicking game theory. 🎮🖱️📈
Exact Prompt:
I need you to create a simple game, it should be based on the idle game theory.
New checkout the experiments page 🧪 https://newsletter.texti.app/experiments to see both games in action.
Gemini yielded a game which was running but lacked the starting point. So you could play in theory, but you couldn’t make up the first 10$ to get started with your idle game. Yes poorly structured code all in a single file, bad for the lifespan of a game, BUT it was working flawlessly! It also took like ~30seconds to generate the code. I also follow-ed up saying I can’t make my first buck and he fixed it instantly. ⌛🧑💻🛠️
Gemini DEMO: https://newsletter.texti.app/experiments/gemini-idle-capitalist
---
OpenAI O3’s model yielded a game that was way more basic, it virtually had no styling and was lacking a little touch of design. I had to ask it to make it pretty. The code style is as usual per AI models, everything thrown into a single file, and as stupid as it gets. I’ll leave the second version as well, for you to judge, see that bellow. 🎨🧪👀
OpenAI's game is also way more basic, lacks depth and iteration.
To me a clear winner is Gemini here! 🏆
Open AI (O3) Demo: https://newsletter.texti.app/experiments/openai-idle-miner
The code quality isn’t perfect by far, but it gets the job done, and that makes it really cool. I achieved a starting point in very short time, and could easily elaborate on that. I must say that after a more prolonged experiment, the models get very confused and stop understanding what are they doing. The models simply forget what they’re working on, and start hallucinating real bad, so you then are pretty much lost. 😵 🧠💥
I would like you to give it a try as well and ask it to build something small yet interesting. 🎯🔧👨💻
In conclusion I would like to sum-up: 🧾📌✅
- With a suite of AI models you can actually do cinematic stuff including the audio models. Kids born today will have a completely different perception of the world simply due to AI tech.
- Building games, takes 2 prompts on average. Thanks Gemini.
- Everything is a subscription these days. Want AI? Pay money, so make sure to yield some money from that subscription.
- Texti news are still free of charge, now even in a telegram form [EN], [RU]
That’s going to be it for today my friends, thank you for being a part of this community and please be sure to share it with your own friends. 🤝🗣️💬
Especially the Telegram Channels, I honestly find that very useful and I’m subscribed to them and read them on a daily basis. 📲❤️📖
Reply here to know what language should I add next. 🌍✍️📣
