Back to AI coding this week, continuing the final push for Wave.

The foundation is there, but still a small list to button up.

In the meantime the AI world has not slowed down.

Topics for this week:

• Release Requirements

• One Thing at a Time

• Claude or Opus 4?

• DeepSeek R1 Update

• The Way of Code

Release Requirements

Wave moves ever closer to release as I slowly checkboxes to prepare for go-live.

After some back and forth with Apple Support, we were able to get an official Developer Account created for Wave to publish under the Wave business instead of Divs, so I have a path to publishing.

Now the game is trying to check all the Apple box before submitting the app to Apple and undoubtedly getting denied for some dumb reasons they don’t explain.

Researching online, I have a list of items I’ve been trying to complete in anticipation of App Store rejections, and the ones I’ve been working on this week were account deletion and Apple Sign In.

I had to set up another service to handle account deletion since it’s not exactly secure to allow the app to delete accounts beyond your own account.

Admin functionality like that needs to live somewhere a little more locked-down, so I created a new Lambda function to handle users wanting to delete their own accounts, and that is happily living in AWS.

And the other feature I haven’t gotten to yet is Apple Sign In.

Since the app allows users to sign in using Google accounts, Apple will also require users to be able to sign in with Apple accounts, which is fair enough, it’s their platform.

I don’t think that will take too much work, just need to authorize another app with Apple to allow for that sign-in process but we’ll see.

Luckily Apple Support has been surprisingly helpful so if I do run into more issues, they seem to be able to resolve cases unexpectedly smoothly, but it would be more ideal to get it up and running on my own without having to coordinate with support.

But thankfully, those seem to be the biggest expected blockers at this point but we’ll see how the submission process ends up.

Hoping to wrap up Apple Sign In and a few other minor things this weekend.

I’m not going to put a time estimate on it this time, though.

Expect the Wave app to go live sometime in the next decade! 🤞 

One Thing at a Time

The YouTube channel is still slowly growing despite now complete neglect.

And I haven’t posted anything because I have tricked myself into a cycle of thinking I only need “one more day” to wrap up everything with the app, then I can focus more on YouTube again.

When in reality, the app isn’t going to go anywhere any time soon.

I need to get back to the time-blocking strategy.

Dedicated time each day for each thing, then both can make progress, just maybe not as quickly as fully focusing on one thing.

It’s tough because a pretty common theme in business advice is to “focus” on one thing.

Trying to juggle multiple things makes it that much harder because really just one thing could easily take all my time and attention if I were to push it to it’s full potential.

But sometimes that’s not how things go.

The app is really close to a major milestone and it’s been my primary focus, but I think the continuing expectation that it’s ready to publish tomorrow has led to some burnout.

Rather than pushing long days to finish everything, it’s better to maintain a constant pace, and make time for other projects too.

One thing at a time is ideal, but maybe not realistic right now.

Focus will be the goal for the future, time-blocking is the strategy for now.

Claude or Opus 4?

Back to AI coding this week while wrapping up Wave features, so I figured it was the perfect opportunity to give the updated Claude a shot, or at least I think I tried it.

I’ve mentioned this before, but I’m not a huge fan of how Cursor abstracts their AI agent process.

The option is simply “Auto” so I trust that they decide which models to use and how to use them, could be Claude 4 or could be Opus 4 or could be something else, but that leaves me guessing what they’re doing under the hood.

I typically prompt Gemini Pro with what I need done, then sometimes will take an entire block of Gemini’s answer and dump it into Cursor chat to apply all the changes.

GPT-4.1 is very good at handling multi-step instructions like that now and usually steps through a big block of code and suggestions from Gemini just fine, so I figured I’d give Cursor Auto a shot, but I was not very impressed.

It seemed to spin in cycles a lot, searching files, then searching more files, then searching more tools, then finally delivered a result that was ok.

For one of the sessions I just stopped it before it finished because I could see it going in the wrong direction.

More direct instructions worked quickly and efficiently though, so that was good to see.

To be fair I didn’t test it very thoroughly, I wasn’t really in the mood so I switched back to explicitly GPT-4.1 instead to keep going, but I’d like to give it another shot soon.

Because of that, my workflow remained the same this week, big questions to GPT o3 and Gemini, then GPT-4.1 to implement or make small changes.

I’ll look into Opus and Claude more next week though and see if I’m missing something.

Happy to use the best tools available and hopefully Opus or Claude are one of them!

DeepSeek R1 Update

DeepSeek couldn’t let the other frontier AI model companies have all the fun in the last few weeks.

They released an update to their open source model DeepSeek R1 0528 this week to compete with Gemini Pro 2.5 and GPT o3, which sounds like R1 is pretty close in performance.

I haven’t tried it myself but it’s been interesting to see what people have been posting online.

The most interesting to me is that DeepSeek may have switched siphoning data sources from GPT to Gemini instead:

It’s language looks more similar to Gemini Pro 2.5.

This would make sense for a couple reasons.

Gemini Pro 2.5 has been topping the leaderboards so why not use it as a primary tool for updated training, but also could be that OpenAI cracked down on their access since OpenAI noticed massive data downloads of GPT coming from China not long before DeepSeek R1 was first revealed.

Speculation aside, I do think it’s good to have a powerful open source model available for everyone, specially when Meta dropped the ball with their latest release.

It’s only been a few days since the R1 update so I’ll be curious to see how things shake out, but in cause you are lost in the AI release cycle, @piet_dev provided a great reference in response to DeepSeek:

The Way of Code

I’m recommending this without having read it myself, but I get the feeling it’ll be a good one.

I just found out this week about Rick Rubin’s The Way of Code “book”, subtitled The Timeless Art of Vibe Coding.

It is a book, but also a website: https://www.thewayofcode.com/

It combines the ancient principles of the Tao Te Ching with the current trend of “vibe coding” which I’ve been doing regularly.

It’s interesting that this came from Rick Rubin, who is a music producer, not a developer but it seems he’s fully embraced the vibe coding trend.

Apparently it all originated from a single meme:

Rick vibing in the zone.

Also “designing an entire b2b saas without talking to a single user” speaks way too deeply to me.

The “book” or website is a really cool idea, providing interactive visuals to compliment the verses on the page from what I’ve seen so far.

It’s an interesting time to be alive, combining philosophy with programming, but somehow feels surprisingly inevitable.

Marc and Ben released a podcast with Rick to discuss it’s origin if you want more context on how this came to be and have the time to dive deeper:

And that’s it for this week! Release requirements, focus, and the way of code. Time to vibe… code!

Those are the links that stuck with me throughout the week and a glimpse into what I personally worked on.

If you want to start a newsletter like this on beehiiv and support me in the process, here’s my referral link: https://www.beehiiv.com/?via=jay-peters. Otherwise, let me know what you think at @jaypetersdotdev or email [email protected], I’d love to hear your feedback. Thanks for reading!