We're back, baby! After struggling with the nerfed ChatGPT 4.0 for the past month or so, I took a crack at 4.1. The results speak for themselves. I was able to wire up lock screen notifications for messages and training circle updates almost perfectly on the first try. The only issue ended up being with my system. Tune in for the nerdcore!

I have won some difficult battles throughout this project, but none have been as hard-fought as the one I am in now against Codemagic and iOS.

Just when I was getting comfortable with all the new moving parts within the app build, I had to go and get into Firebase App Distribution. Honestly, it probably wouldn't have gone nearly as bad as it did if ChatGPT were still the brilliant little buddy it was when I first started this project. But my AI friend has led me into the middle of the desert and left me to die a slow agonizing death.

As I approach beta testing, I take a moment to address some of the negative feedback to AI that I've been seeing online. Much of it revolves around the expectation that humans will intellectually devolve as they become more reliant on AI. But I guess that depends on how dumb somebody is to begin with. In my case, there is an argument to be made that the only direction left is up.

The emotional rollercoaster continues as I make massive progress wiring in tons of new features, but it all came crashing down with one seemingly simple action: creating a second test profile on my phone. When you have a product that is data-driven, you have to make sure that the data is user-specific. I learned the hard way.

I am back on track after what felt like weeks of suffering. I now have leaderboards, badges, stats and block summaries. All made possible by my little friend, the widget. But with widgets, come more files to keep track of and the project has ballooned up to 34 files. It is a little concerning but not nearly as concerning as the persistent eye twitch I am developing from staring at screens all day.

I might have broken ChatGPT 4.0 and I am not entirely sure 3.0 mini high will be able to shoulder the load, but we're going to give it the old college try. One thing that is working is the Texas Method, which was not nearly as challenging to build as I had anticipated. The workout footers are still broken, but it seems like things are moving in the right direction again.

I am back in the gym and testing a product! The app is great, but it's definitely missing one key feature - the leaderboards. So after carefully planning my attack, I decided to jump in and start building. Unfortunately, things went sideways as they often do and I was unable to make the leaderboards happen, but I think I figured out why chatbots get dumb after extended runs of being absolutely brilliant.