Frank Andrade

o3-mini-high is the best coding model right now. Its coding average on LiveBench is 82.74. Other models don't come even close: o1 (69.69) claude 3.5 sonnet (67.13), deepseek-r1 (66.74). Here's a cool space shooting game I created with only ONE PROMPT: "create a space shooting game using Python. make the game images look like the real game (not triangles/rectangles)"

Do you want to try DeepSeek R1 but are worried about sharing your data with China? If you're concerned about privacy, just install deepseek-r1 locally in your machine! Unlike the deepseek app (which is hosted in China), running an LLM locally ensures that your data stays on your machine. In my quick 2-minute video below, I show you how to do it 👇

What was the AI product of 2024 for you? I loved Gemini’s ability to see our screen and provide real-time assistance, but I’d say NotebookLM was my favorite because I think it’ll redefine learning in 2025 and beyond. NotebookLM got even better a few days ago after Google rolled out the ability to join the conversation yourself. See how AI hosts discuss "The Invention of Air" by Steven Johnson, and then the author joins the conversation at 0:26 👇

o3 scored 87.5% on ARC-AGI Public Benchmark in high compute mode Humans performance is 85% It's just mindblowing! But it's not AGI There's still a fair number of easy ARC-AGI-1 tasks that o3 can't solve (some of the type "easy for humans and hard for AI"). Here's one task 👇

Google Colab will soon have a data science agent to accelerate your workflow (using Gemini 2.0). You'll only have to describe your data analysis goals, and then your notebook will be generated automatically. It's just mindblowing! 🤯 Here's what Gemini will do for you: - Data exploration/preparation - Feature engineering - Model training/evaluation/optimization - Data cleaning/analysis/visualization Source: @googleaidevs

Google Gemini 2.0 realtime AI is amazing. Here’s how you can use it to learn Python, data science and other technical stuff. 5 examples 1. Turn Gemini into your code tutor by sharing your screen and starting a conversation with it

Live video and screensharing with ChatGPT Voice are finally coming out! Now ChatGPT can see you and what's on your device in real time. Here are 5 amazing use cases to make the most out of these new features 🧵

Today, Apple launched its ChatGPT integration with Siri. Now you can tell Siri to use ChatGPT for tasks like generating text and images. Apple also enabled image and document understanding capabilities. You can combine this feature with ChatGPT by telling Siri to use the content on your screen and then sending the screenshot to ChatGPT, as I did in the video below. Here are other AI features Apple is releasing in iOS 18.2: - Visual Intelligence: Now, you can use your phone's camera to learn about objects and places instantly. Visual intelligence can also summarize and copy text, translate text between languages, and more. - Image Playground: You can create images, with concepts like themes, costumes, accessories, and places - Genmoji: By simply typing a description into the emoji keyboard, a Genmoji will appear - Image Wand: You can use Image Wand in the Notes app to create images in your note using the written or visual context already captured within the note.

ChatGPT Canvas can now execute code! It can run Python code and "become" an automated data scientist or swe, create little software tools, games, and more. You can also let ChatGPT fix any bugs based on console errors. The execution of Python code in ChatGPT Canvas uses Pyodide. Besides code execution, Canvas is now available to all ChatGPT users and it has support for GPTs

I asked ChatGPT Advanced Voice Mode to commentate on a football match. It responded with a wide range of emotions. Excitement when calling a goal, sadness when its team lost, and excitement again when it found out the goal was offside.

Web scraping with ChatGPT Code Interpreter is mind-blowing! You only have to: -Go to a website and save the page as HTML -Upload the HTML file and type "From the HTML file, extract [data]. Export data to a CSV file" -Inspect the data. Copy its HTML element. Add it to the prompt