Back to balancing businesses this week.

App development versus YouTube development.

Luckily one is building momentum on it’s own which helps a ton.

And the other is being heavily carried by AI.

Much to discuss.

Topics for this week:

• Final Push

• Google Back on Top

• More Models

• YouTube

• Who’s Asking?

Final Push

I’ve been dancing around an app release for a few months now but a final release looks to be on the horizon.

The goal is to have a published app by summer and summer is rapidly approaching.

In less than 4 weeks, the Wave nightlife app should be a real thing.

It’s nearly feature complete (probably too many features) but still needs a few more tweaks and a lot more testing.

Looking forward to that major milestone though, a full, published app to multiple platforms.

Gotta wrap up this newsletter to get back to wrapping up app development.

Google Back on Top

One way that I’ll be wrapping up app development is by heavily prompting Gemini Pro 2.5.

I’ve been switching to each new major model as my primary developer to see how well it produces code and Gemini seems to be the winner right now.

I mentioned it last week that Claude Sonnet 3.5 and 3.7 have been great, but they kind of mindlessly do what you tell them regardless of it being a good idea.

Gemini Pro genuinely will push back against something if it thinks it’s not a great way to do things.

I wanted to show a list of profiles that Liked a particular post.

I had already created a “NotificationListItem” component that I could use to display Comments, Follows, and other typical social media interactions as well as Likes.

But technically this list of likes isn’t a notification, it’s just showing basic info about someone liking a post.

I figured I could save a little code and re-use the NotificationListItem to show this list of Likes too.

When I asked Gemini about updating that component and the related components, it was rather insistent that I create a separate Like component, and didn’t make the change I asked for.

I was not expecting that and almost didn’t even notice what happened until I went to implement some of the changes and realized there was code missing.

I asked Gemini what happened. Here’s it’s summary:

Pretty cool!

It actually pushed back on a bad quick fix approach.

Claude Sonnet would have gladly made that change and accidentally changed 5 unrelated lines in the process.

This is the kind of thing that will lead to trustworthy AI development and not having to rebrand unpredictable generative results as “vibe coding”.

Beyond the code practice pushback, Gemini has genuinely helped refactor a number a places in the app to be more stable and use less code.

Looking forward to using Gemini to continue fixing and improving my current app development and future projects.

It’s free to use right now on AI Studio so go check it out while you can!

More Models

While Gemini is back on top for the moment, there’s always a new release waiting around the corner to possibly dethrone the current leader.

Just today, Sam Altman mentioned that o3, o4-mini and GPT-5 are all a few weeks or months away.

I could see a full version of o3 being another competitor for the top spot now that they have to contend with Gemini and Grok.

Sneakily, ChatGPT-4o has moved up to #2 on the list which is a big surprise.

Altman posted about it being good a few days ago, but I didn’t realize it was that good.

Should have known given the all caps “GOOD”.

I do wonder if 4o image hype is helping push up the ranks, or if hype in general causes llms to be tested less and fall down the leaderboard in general.

I’ll have to give 4o another go beyond just using it as my go-to example for AI Automations.

And these are just Google, xAI, and OpenAI models.

Plenty of other models in the works too, including rumors of an upgrade to Deepseek releasing soon.

At this rate, I’d be more surprised if there wasn’t some big announcement to talk about next week.

More models to come I’m sure.

Passive Growth

With a looming app deadline, I spent much more time developing this week, and much less time recording automation tutorials.

Despite not uploading for 5 days, my YouTube channel stats continued pretty solidly on their own.

This is the beauty of YouTube.

“The Algorithm” is always out there working 24 hours a day to find viewers the right content, even though I hadn’t touched YouTube for almost a week.

If you produce the right content, YouTube will find the viewers.

There happens to be a massive trend on YouTube right now for an insane amount of searches and not nearly enough results so it’s possible to get a good amount of growth like this just by producing what people want but aren’t getting enough of.

More than 4 million searches for n8n per month right now on YouTube, and not nearly enough search results to match that means this growth trend should continue for a while until the n8n hype slows down.

We’ll see what happens, but I plan to keep creating n8n tutorials 2-4 times per week to try to ride the wave, but also hopefully provide some useful content!

Stats are saying I’m on the right track.

Level of the Question

I had already previewed that Naval would be on Chris Williamson’s podcast this week so I made sure to carve out 3 hours for this one.

Ended up being a 2 session commitment because 3 hours is a long time, but I enjoyed it.

There were a ton of great insights, too many to break down here, even though I probably could, it would just be an unreadably long newsletter instead of the usual way too long email.

One quote I particularly like:

“The reason to play the game is to be free of it.”

That pretty much sums up my entire approach to work and business.

Beat the game so I can move on to something else.

But the biggest moment from this podcast for me was Naval explaining philosophical paradoxes.

Something like “does life have meaning?” or “is there free will?”

Hearing Naval explain his answer made me say “wow” out loud, sitting by myself.

His answer: answer the question from the level that is was asked.

So obvious that it’s easy to miss.

The question gets asked at a human or personal level, but then answered at a universal level.

Obviously life has meaning to a person and obviously humans are an infinitesimally small part of the universe.

Zoom far enough out, nothing matters, but zoom far enough in and everything matters.

The trick is that the levels of the question and answer are different so don’t switch them.

Answer at the same level of the question.

So asking at a human level, “does life have meaning?”

Answering at a human level, “yes.”

That was enough to consider this podcast a win, but I’m sure I will be re-listening to it very soon.

Worth working through eventually if you have the time!

And that’s it for this week! An app timeline, more AI models, YouTube stats, and the deepest philosophical questions.

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!