I built an open-source tool for learning anything using AI: - Type what you want to learn - It creates a full course for you Move from beginner to mastery of any subject, even complex things like Quantum Physics It has three formats: - Explanation: bite-sized lessons that you swipe like TikTok/Instagram - Practice: solve problems in real-world situations - Quiz: test your knowledge in Duolingo-like lessons - Language courses have a different format covering vocabulary, reading, and listening skills It's more hands-on than reading a long textbook or watching a video Plus, it uses everyday language, so it's easier to connect complex concepts to your reality. I built the tool I wish I had when I was in school/uni Of course, I'm biased but I've been using it for the past few weeks and I'm truly loving it. I can't imagine myself going back to learning through traditional ways

I think we need a better OS for AI in general. The current way computers, phones, and operating systems were designed was for humans to interact with a screen. But moving into the future, we are going to start using voice much more often, so we need operating systems that completely rethink how we interact with devices. We need operating systems that realize they are voice-first and just take actions based on what we want. I don't think the future is us going to individual websites or navigating anymore. It is going to be just asking for information, and the device brings it back and gives it to us. Companies building operating systems and hardware should really rethink how their OS works Even though AI feels like living in the future, existing operating systems still feel like living in the past.

I’ve changed my mind about this topic Yes, Opus 5.5 is the best model right now, and it still makes silly engineering mistakes. But does it really matter? I used to think every line of AI-generated code needed reviewing. Now I think it depends on what you’re building If you’re building safety-critical software, absolutely. But most software isn’t that. For a startup that hasn’t found PMF yet, shipping and iterating faster is far more valuable than having a perfectly maintainable codebase Focusing on automated quality checks and making sure your agent knows how to check things e2e is probably more important in most cases

I feel like AI coding tools are too verbose. I'd prefer something that feels more like working with a colleague For example, it'd be useful to have threads organized into sections: - Questions: Things that need my attention. - Decisions: What the AI decided by itself and why. - Thinking: The whole thinking process, if I want to see it. - Overview: Everything that changed. Right now, questions often get lost in the conversation. I'd rather have the default UI focus on what needs my attention and let me explore the details when I want to It would be much easier to review the work as we move towards not reading the actual code anymore

What companies have unlimited tokens for employees? Asking for a friend

A dream: unlimited tokens

Big. Now do .agents/skills too please

I have a scheduled task on Codex to read my X feed and give me a report on relevant posts filtering out topics that I'm not interested in. This has been pretty useful