AI for coding

Good morning, and welcome to the fourth video lesson of our Python course!

You’ve probably noticed that LLMs can make mistakes when you ask them to write complicated code or analyze tricky logic.

Yet they’re rapidly improving, especially the “reasoning models” that produce detailed thinking steps internally.

These AIs break a problem into smaller tasks, systematically test solutions in their own chain of thought, then arrive at a better answer.

That’s the hallmark of a reasoning model, and it’s a huge deal for coding. Instead of just spitting out code, it’s trained to reflect on edge cases and compare different approaches.

Right now, that strategic thinking might be hidden or partially shown, depending on which interface you use, but you can see the benefits. Benchmarks testing how well AI solves coding challenges show these models are doing great. They’re solving problems more consistently, reducing silly errors, and getting better each month.

But since they’re more powerful, they tend to be more expensive—so if you rely on free versions, you might never have tried reasoning models like OpenAI’s o1 or other advanced options.

Yet the leap from GPT-4o-mini to GPT-4o, which is not even a reasoning model, is huge, saving enough time and effort to justify the subscription, and similarly, a leap from GPT-4o to o1 can be huge as well if used properly.

If you haven’t tried these higher-level models enough, you might not realize how productive they can make you.

We know that you can copy some code, paste it into your AI’s chat box (like ChatGPT), and say, “Can you improve this?” or “Explain this part to me?” or “Implement a different function that does the same job.” Then, we simply copy-paste the improved code back into your file. That’s already pretty cool, but it can be tedious to juggle prompts, code snippets, and multiple browser tabs.

People working on specialized IDEs, specialized editors, or specialized collaboration tools are trying to fix that. Instead of you being the glue that copies and pastes, they want to embed the LLM directly into your environment so the code is always visible to the AI, and the AI can produce changes instantly.

Let’s talk a bit about two examples of that approach: Cursor and GitHub’s Copilot Workspaces. Each takes a different spin on making your life easier, and both are trying to figure out the best ways to mix code editing and AI suggestions...

By the way, if Python programming (or just working with LLMs) is something interesting to you and you finally want to make the move to learn programming with Python as a tool to code with LLMs, now is your chance! With such powerful help and proper guidance, you can learn Python programming faster than we've ever could.

And we've built a course just for that! Check out our Python Primer for Generative AI here.

And that's it for this iteration! I'm incredibly grateful that the What's AI newsletter is now read by over 20,000 incredible human beings. Click here to share this iteration with a friend if you learned something new!

Looking for more cool AI stuff? 👇

• Looking for AI news, code, learning resources, papers, memes, and more? Follow our weekly newsletter at Towards AI!

• Looking to connect with other AI enthusiasts? Join the Discord community: Learn AI Together!

Want to share a product, event or course with my AI community? Reply directly to this email, or visit my Passionfroot profile to see my offers.

Thank you for reading, and I wish you a fantastic week! Be sure to have enough sleep and physical activities next week!

Louis-François Bouchard