It’s time to embrace the AI

Thought experiment for you guys. Which is more likely? 1. I started shilling AI to make more money with T3 Chat 2. I built T3 Chat because I started liking AI more and wanted a better interface Spoiler: there’s a correct answer here

While reading and reviewing code is valuable and certainly part of learning, it's not the same as actually creating it. Writing your own code, solving problems, making design decisions, implementing solutions etc. is where the deeper learning happens. That hands-on process forces you to think critically, confront mistakes, and truly understand what your code is doing. Code review builds recognition and helps you learn patterns and structure, but writing code builds understanding and intuition. It's like learning a sport: you can study game footage, analyze plays, listen to expert commentary, provide feedback etc., but until you get on the field and practice, you won’t develop real skill or muscle memory. Without that, you're not very useful to the team. That’s why it’s so important that junior developers spend meaningful time writing code, not just reviewing it. If we want them to grow into capable engineers, they need space to build, break things, and learn from doing, not just reviewing. The same principle applies to more experienced engineers as well, especially when developing more advanced skills. Over-reliance on AI can lead to skill atrophy and, over time, a decline in overall engineering capability. Striking the right balance between leveraging AI and maintaining hands-on expertise will be a key challenge for the industry moving forward.

56:00 person selling stuff for n years doesn't realise people don't love ads.

My experience for AI in programming; If working with extremely popular and stable libraries/platforms, then AI is incredibly useful. But if one is using lesser used libraries or libraries that are changing in big ways (i.e incompatible upgrades) then the AI is giving nonsense and effectively slowing me down, corrupting working code and so forth. Modifying React and Vue codebases works fairly well and only occasionally hallucinating. The long-term effect would then be that "New Development" of libraries, platforms and languages will come to a halt.

Shoutout to the editor. We see what you did there

Damn, I love your comment about t3 chat's architecture. I need to start thinking more like that

>malformed aijeet spergs for one hour this is poetry

56:50 that zoom in on the beat of the hand motions was superb

I have been an AI hold out for a long time because it was very difficult to differentiate between the hype and the real benefits. Until very recently, nearly every example of AI utility I saw felt like it was, trading one problem for another with a marginal net benefit at best. In the past few weeks, I have seen far more cases where the AI tools are actually achieving a level where they are actively worth playing with them. I don't think that being a hold out on this was a bad decision or has been harmful to me in any way, I just was waiting for the tools to evolve to a point where they represent a big enough benefit that they are worth the effort of learning.

I only started getting into AI seriously after Deepseek R1. After I realised how far it had come, I dove right in.