Returning to my roots this week.

Big AI updates and a podcast that’s not even out yet.

New models, new capabilities, and so many Studio Ghibli memes.

Topics for this week:

• YouTube Channel Growth

• ChatGPT Images

• The Latest #1

• Reve

• Monday

YouTube Channel Growth

Progress!

13 videos into building a YouTube channel and there’s still some potential here.

Not exactly 1 million views or viral success, but that’s also not the goal.

As long as the content is finding other automators, then I’m good to go.

I’ll try to stick with the “Social Media” theme since that seems to do well, but I’d like to throw in some other topics too.

I’ve found the hardest part still is coming up with the video topic, mostly because I tend to do the entire process in one sitting - build and test the workflow first, then record myself rebuilding it, so that can take a while if the build doesn’t go as planned.

I’ll need to get a little better a staggering bigger builds, having them ready to go for a recording later in the week.

Instead, I tend to get trapped in a cycle of over-optimizing for the best idea that’s also straightforward enough that I can build it in an hour or two.

Earlier this week, someone left a comment on a video asking for a specific scenario, which was great, because it was something that could be done in one sitting.

The next day I was able to get a tutorial made on exactly what they wanted.

And I may have another suggestion from the comments to work on next but it will involve a little more work.

But that’s more what I had in mind.

See what the market wants, and eventually I’ll have more video topics than I could manage.

Not sure I’ll stick with 7 days a week of creating videos just so I can take a little more time to build them, but at least 2-4 per week should hopefully get the job done, and I’ll do more than that if it ends up working out.

Promising start to YouTube growth in the last 3 weeks, anxious to keep going.

ChatGPT Images

Just last week I was thinking it had been a while since a major AI development, and sure enough, OpenAI decided to fix that.

Dall-E was ChatGPT’s built-in image generator and it wasn’t great.

It’s usually not hard to spot a Dall-E generated image unless you do some proper prompting, so I had stayed with Midjourney for convenience instead, even though there are quite a few other powerful image generators in the market now that I haven’t taken the time to setup.

But this week, OpenAI announced “4o image generation.”

Not only does it look much more real, but it can handle text surprisingly accurately, which has been a challenge for many image models.

A quick test of my own was really impressive:

This was the first thing I prompted and it pretty much nailed it, only 1 merged together set of fingers!

But the floating title in the background, slightly behind the guy’s head would never have happened correctly in Dall-E or Midjourney or most other models.

It’s clearly able to generate realistic-looking images, but it can also restyle images too.

And the internet went crazy with the “Ghibli Style” conversion which you have probably been seeing.

A lot of potential with this image creation.

I was recently looking into how to automate creating “carousel images” for LinkedIn and this tool may be the easiest way to do that, once it’s available through API.

Right now 4o images are only in ChatGPT and not available for everyone because it’s apparently melting the servers at OpenAI so it may be a little while until this is fully publicly available.

I’ll look forward to that rollout and finally cancelling my Midjourney subscription.

The Latest #1

Also this week, in the on-going battle for top AI model, Google dropped a new contender and crushed the #1 spot.

Gemini Pro 2.5 seems to be the winner now in coding and intelligence.

I haven’t had a chance to try it out, but I’m looking forward to giving it a chance as my newest coding buddy.

1 million token context is still mind-blowing.

You can basically feed it entire novels to read in one prompt.

And it’s funny how much I have started to appreciate the “doesn’t just agree with you 24/7” aspect.

A while ago on a podcast, Sam Altman talked about how eventually the model should work more like a senior engineer and point you in the right direction if you’re possibly going down the wrong path.

That’s probably my biggest gripe with Claude right now is that pretty much just mindlessly obeys whatever I ask it to do when clearly that idea isn’t the most scalable solution.

ChatGPT 4.5 does have a little more push back where it will subtly suggested a better way of doing something, which has been surprisingly helpful.

Looking forward to see how Gemini Pro 2.5 handles those situations.

I haven’t been using Google AI nearly as much as any of the other models, but lately I’ve been doing more Google Console tutorials so it may be time to get Google’s models into the regular rotation.

Reve

If it weren’t for GPT’s 4o image generation, this would have been a bigger deal.

Scrolling on X, I ran across the Reve image model which was also creates impressively realistic-looking images with accurate text.

Unfortunately for them, I think OpenAI might have killed their hype, at least for now.

Things always change, though.

If GPT fails to rollout their new image creation properly in the near future, or has too many limitations or restrictions, there’s plenty of opportunity for another answer to step in, like Reve.

Pretty amazing stuff, it will certainly be on my list of options for image generation in the future, and hopefully they find their role in the market.

Monday

It’s been a minute since I referenced a podcast, so today I’m going to talk about a podcast that hasn’t even come out yet.

On Monday, Chris Williamson releases his podcast with Naval Ravikant, which is kind of a big deal.

According to internet rumors - or some random comments I saw online one time - Naval is a very selective podcast guest, but Chris has established himself enough in the podcast game to warrant an appearance.

Naval is responsible for one of my favorite podcasts that I’ve referenced a few times in this newsletter:

Naturally, I’m looking forward to this one.

I’ve backed off on podcast consumption (and creation) lately so it’ll be fun to jump back in for this.

A quick preview:

And that’s it for this week! YouTube channel growth, AI images, new AI models, and a podcast to look forward to.

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!