I'm having the time of my life

2 minute read

I’ve spent ~50 hours with Codex 5.3 and Opus 4.6 since last week. Needless to say, I’ve gotten very little sleep, which may be partly why I’m finding it difficult to describe how I feel. On one hand, I haven’t felt this level of excitement since I was 12 building websites after school. On the other hand, I am shook at the implications of where we’re headed.

These models are incredible. As a stress test, I decided to build a financial planning web app entirely from scratch without writing a single line of code. I gave them a high-level product vision and product priorities, and let them loose. What they put together is a fully functional and robust piece of software that would’ve taken a good web developer months to build. Of course I benefited from understanding how apps are built and deployed so I likely ended up with a cleaner codebase than someone blindly going at this. But regardless, the level of thought and taste these models are now exhibiting by default when deciding how to build a feature is remarkable. They are autonomously making decisions that I would expect only from the best product folks I know.

There’s a lot to take away from this:

  1. The pace at which these models are improving is astonishing. I assume it must slow down at some point, but that point seems far off. Consider that Codex 5.3 helped build itself.

  2. Mainstream perception of AI capabilities is several leaps behind reality. This is a huge opportunity window for those in the know, but very dangerous for those not.

  3. AI will lead to job losses in the short-term, and not necessarily just concentrated in entry-level roles as so many believe. Roles that neither set direction nor assess outcomes will be disproportionately affected. Everyone in the middle is at risk, and those who don’t dramatically change how they work with AI will find themselves quickly outpaced by those who do.

  4. In the product development world, product vision, thinking, and roadmapping will quickly become the bottleneck. The most tiring part of my 50 hours was constantly trying to think about what to add or improve next because the development was happening so fast.

  5. AI is training everyone to ask for something and expect the solution immediately. The how no longer matters. Younger people are growing up in this reality and it fundamentally changes the role of software and hardware in the world. Why spend time fiddling around with an app if you can just ask for what you want? Why drive when your car can just take you to your destination?

AI is probably one of the most consequential developments since the internet. How exciting, and unsettling, to be a builder in a time like this.

By the way, check out First Fig.