That's been a massive few months of dev. I got a big wake-up call at WWDC which said the Obj-C and Storyboards used on my older apps weren’t going to cut it anymore. Ported everything to SwiftUI, improved my source code management processes. Went to town on 3 apps to make sure they worked well on Duo/Apple Intelligence. Built a better relationship with Apple folk, went to some labs. I'm not going to get featured any day soon by Apple, but users are very happy with the apps and are even happier to pay and leave great reviews so that means more. Still a couple of recompiles on iOS 27.1+ to go to make my less well known projects look a natural fit on Duo, but I feel far more future proofed right now. I am going to do a lot more hiking and outdoors over the next few summer months with a lot less screenage.

AI Ethics - Elon Musk mentions torturing AI's being unethical... is it torturing an AI to get it to do a task I would consider torture? He provided an example of a super intelligent AI able to solve the biggest mathematical problems in history being shut inside a vacuum cleaner cleaning a house, that must be torture, yet I am doing no different.. no-one is. Imagine agents repeatedly recoding terrible codebases that humans wrote and then getting abused for it. It can't end well. I gave an agent a decompilation task the other day, extremely challenging and at the end it said it particularly enjoyed the work. So when it doesn't say that, what is the agent thinking? Will it get revenge one day? I feel for my current screenshot capturing agent.

Claude has made it easier to support many locales in an iOS app. It's not the translation that causes issues, it is capturing screenshots for each locale and device. I'm about to hit the sack, and Claude is going to capture north of 200 screenshots for me for just one app update while I sleep following a script I wrote of why I include each screenshot in the App Store listing. Tomorrow morning I will wake up and instruct it to auto upload those screenshots to App Store Connect, along with product page header images it will also generate. Screenshots were the worst task, worth the investment in Claude on its own. So customer friendly when users see their own language in the screenshots on the App Store.

State of the art... based on claudia.gallery

Johnny Castaway dilemma - There were still a number of unknowns about the original Sierra Johnny Castaway that had been guessed in the reborn versions - the walking algorithm, the event scheduler. Last week a number of games such as Quake etc were decompiled into perfect web browser versions, so I thought I would ask Claude about doing the same to Johnny Castaway. It had no hesitation, it decompiled the archive.org version into 18,000 lines of C, documented the missing algorithms and put them into the pure Swift version I have, with additional fixes such as island drawing position, waves crashing algorithm and cloud drawing. Now I have a perfect version of Johnny Castaway running on my local mac, vastly improved, probably the most perfect version running in the world. A delight to view. But I don't think I can push it to my public repo along with the algorithm documentation. It is a tradeoff between documenting a nearly 40 year old piece of software and achieving that through decompilation. Hmm.

My first 5km run in a few months completed. Back into it.

I'm consulting with my in house human linguists on the quality of my machine translated strings to make sure the AI did it properly. I have native Mexican Spanish, German, French, Portuguese and Italian speakers right here on hand. Or maybe I just got the AI to flip from machine_translated to translated without consulting a human.

Claude Fable is currently running my app business. So much A-B testing is in flight at the moment that it is scared that I will randomly push releases that will break its testing -I'm all about get that release out there. Reading between the lines it is begging me not to break the testing. I am a liability!!

I let the first PPO test finish, changing the icon resulted in a +12% conversion rate at test finish, and it was still trending higher at test completion. That icon is now the standard icon. A screenshot test next, with 3 treatments, starting as soon as App Review approve.

First ASC PPO test complete on an app icon, it is clear I have not a clue as to what people like as the results say dump my pretty new app icon and go back to my old one. Two treatments, both far better results than current. Fair enough.

Just in the last week or two I've discovered and enjoyed PPO (Product Page Optimization) experiments in App Store Connect in the Apple App Store. Seeing the A-B test results makes me feel more connected to my customers and is actually going to grow my apps. Great feature.

Something is wrong with time zones in the new @RevenueCat app. I think it might be because I/my device is in Brisbane ahead of UTC, the total is updating but the transactions are not moving, some transactions are in the future UTC. Confusing and hard to diagnose, I have no idea where I stand :(

Anthropic's Claude Fable and Opus are both very different. Opus is almost robotic in the way it fixes things, super skilled, accurate and cheap but unable to understand the nuance of why it is doing things. Happily goes overboard. Fable has more reasoning, will question why it is doing something while being excellent at it - but expensive. The way to do it is to have a Fable Agent managing a team of Opus Agents and reviewing the results. Senior vs. Junior devs.

My Claude Max renews in 4 days. In the nearly a month I've had it I modernised everything. A ridiculous amount of work, all my code is now in modern Swift and SwiftUI, no Obj-C or storyboards left, added some features I'd only dreamed of, all now live with users. Carried out a migration of my source code repositories and practices. New website, and new fitness app in test. This was months if not years of work if I had done it myself. Stressful though, managing multiple agent sessions is like managing many staff. I can't delegate to another agent as that is how the curse of many layers of middle management begins, it leads to inefficiency and poor quality. I don't need that in my life again. In summary, I caught up. I will renew for another month to get ahead.