Cal Newport AI takes are WILD...

Software is literally the foundation of all automation. If you can automate software, you are fundamentally automating automation

The invention of the wheel was only impressive in the narrow field of wheels.

Great video man. I watched at 40x speed.

2-4 was where the progress was obvious in the chatbot paradigm. After that it wasn’t as obvious it was improving under that paradigm. But now we are in the agent paradigm.

It's a matter of perspective. You only notice the change when it can deliver something you want from it that it couldn't before. Casual user -- wow it can proof read and polish my emails... Next version it can proof read and polish my emails a bit better Technical user (eg researcher/programmer etc)-- gets so many things wrong, I have to redo too much, it's not worth it--- next version wow this new version gets it right 90% of the time, still costs me a lot of time to check though --next version, I can delegate whole sections of work, this rate of progress is amazing soon I won't need to use the agency I'll just get AI to do it and save 20k

I think some people judge and value things by how it helps consumers do what consumers have been doing. That isn't exactly what AI does

i can understand it would seem that way if you only use copilot

I'm a developer with 40 years as a professional. I work for large enterprise operations.... my software has helped make everything from plastic bottles to cars, and some of my work is used, today, by hundreds of thousands of people. Today, I submitted a PR that was comprised of over 100 source code files. While it took several passes of queries, as well as layered progression of requests (For example, "Implement these classes" followed by "write the unit tests" - though the actual prompts were more focused and defined) and a few manual adjustments - it only took a few hours to work up code that would have taken me a week - at least - to do "the old way". This is serious code, unit tested, integration tested, QA'd at several levels, then used by thousands of people daily. AI doesn't just write the code, we also use it as a first pass on our PRs, use it to find edge case bugs... it **IS** helping us write better code. No, we aren't "vibe coding" this either. The code is already architected before we were given the specs. GPT 5.3 Codes or Claude 4.6 is coding to the spec, and leverages our extensive codebase to use the same patterns, along with well defined instruction.md files. Anybody who says AI isn't doing real coding work and providing real gains is lying or not using the tools correctly.

The idea that software is a narrow use case is crazy. I've built so many tools for myself in the last month—things I would never have even tried to build before because it would not have been worth it. But now that I can just say what I want and it appears, why not build it?

This is your best video ever Wes! Ever! Well done sir