Texti #56: Google I/O in a nutshell and latest AI news. 🤖🚂✨

Hey there, how’s the AI state feeling for you? doesn’t it feel like everything’s moving really fast right now. It feels like you’re always behind the train, but worry not, I will fill in the gaps, so that we’re all on board and talking to the conductor of the train to learn first about the direction we’re heading to. 🚄 🗺️💬

So let’s see what wagons we have missed and need to make sure this train is not too long ahead. 🚋📍📉

The highlights of the week:

General Specter

1. MIT Researchers are trying to make Concrete Green using AI.
2. Same MIT Researchers but different are trying to teach AI to recognize when it’s hallucinating instead of providing wrong answers, (somebody’s looking for the humble knob 🤷)
3. Reddit sues Anthropic for unauthorized data consumption, while reddit is just filling their pockets like Scrooge McDuck (due upcoming US Government contracts)
4. AI Driven ads at Meta. This day is coming and it’s going to be within the next 6 to 12 months. You’ll ask hey Meta tell me what is better coke or water? and it will reply have you tried the new coke zero zero (no sugar, no caffeine) with the taste of water?
5. An internal roadmap from OpenAI want to make chat gpt an one stop for all, including e-commerce so get ready Shopify - OpenAI is preparing the oven to roast you alive.

Software Engineering related:

1. Anthropic is going for the big buck, it will provide AI for the US National Security. If you’re an investor, well you just hit the jackpot.
2. Snowflake tries to make their interface intuitive and user friendly by providing an AI based interface - FINALLY!
3. Samsung and Perplexity will collab, and make the AI search finally good. (I’m a believer that perplexity is probably a highly underrated tool people are missing on) I use it, and I honestly love it as it does really well as an AI search engine.

Google I/O

Google I/O, was huge among every influencer out there, because it scared the shit out of all of them. The Reason Veo 3 the demos put out there are freaking incredible, the synchronization of sound effects to the video action is mind-blowing. It is just on a different level, and compared to what we’ve seen so far is nuts! 🎥🔥🎶

Here are some examples to play out with sound. 🎧🎬🔊

Remember how about 10 years ago we were amazed how well the CGI was able to replace Paul Walker (RIP) in the fast and the furious movie series? Well apparently this is now technology available to the general public. 🕰️🎞️🌐

Last year I was saying that anybody can be a filmmaker with the state of AI models at the time, now we’re closing in on the complexity we had before. We’re moving real fast, and I believe the day when some 5 year old will be creating fully fledged AI-Powered Movie films that could rival large holywood studios. 🧒🏆🎥

A shameless PLUG 📣📱💡

Embracing the power of AI I have created a dedicated Telegram channel providing the latest AI news directly in your phone. It always maintains the same style and keeps information very much short and consistent. 🗞️🔁📲

In case you want to stay always on top of the news, it provides valuable information every morning, the news are localized based on your language. 🗺️🌅💬

English 🇺🇸: https://t.me/textiNews​

Russian 🇷🇺: https://t.me/textiNewsRu​

Other languages are coming soon. Reply to this email with the desired language so I can prioritize it. As you all know I do this on my own and time is quite limited. ⏳🧍‍♂️📬

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:

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. 🌍✍️📣

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

​

Lastly if you want to support me feel free to become a patreon here​

​