I might be starting to understand leverage and the real idea of the 4 hour work week.

Concentrating on the right things can produce way more results than spending time on the wrong things.

And sometimes you pick the wrong thing, despite research (maybe not enough research).

In short, this was a busy week:

• I made it to Tier 2!

• First Booking

• YouTube Opportunities

• o3 Cancelled

• Not Firebase

I made it to Tier 2!

There were a couple stressful stretches this week, but the first one was debugging why the OpenAI API suddenly started returning an error that I had “exceed my limit.”

I completely forgot that they changed their payment model to a “Pay as you go” model where you buy credits ahead of time when they used to just send you a bill for your total usage that month.

Checking my account, I saw that I had -$1.77 so I quickly purchased $25 worth of processing and set it to auto-charge if I hit $5 or less.

A little annoying that I’m also paying $200 per month for the Pro plan, but I must have already burned through the free credits you get with that.

With that latest credit charge, though, I officially reached Usage Tier 2.

Basically means I’ve spent enough money with them that they’ll let me spam them more.

Like I can now send up to 5000 Requests Per Minute for gpt-4o instead of 500 RPM in Tier 1, which I was getting dangerously close to sometimes (couldn’t I batch calls with more data and use less requests? yes but I wouldn’t have reached tier 2!).

The limit increase is great, but I’m now more motivated to reach Tier 3 because that tends to be the lowest tier that OpenAI releases new features for testing.

The real AI pros.

Tier 3 isn’t all that hard to reach either, it requires $100 of paid API usage total, which I seem to be on track with already, but I won’t be going out of my way to spend it.

Instead, I’ll soon be pleasantly surprised to receive another email congratulating me for moving to Usage Tier 3.

First Booking

I started a cold emailing process last week, and in less than a week it shockingly led to a positive lead.

Just over 500 emails, I got 1 positive response, which is actually not bad return.

I’ve only been sending 100 emails per day while I get used to the process, but I’ll take that rate.

If the math maintains, 600 emails per day could be 1 booking per day which is not bad at all for something that mostly happens automatically.

Combine that with a few other marketing efforts, like say YouTube and maybe even an ad campaign, and I may soon have a good problem of being overwhelmed with business.

We’ll see though, the meeting is on the calendar, but doesn’t a guarantee a client.

Still plenty of work to do.

But it’s a step in the right direction!

YouTube Opportunities

When I started divs.design, I started researching what the Webflow space looked like on YouTube.

I had an entire spreadsheet of topics to cover with high opportunities to get views because the lots of people were searching but there weren’t that many search results.

Ultimately I ended up hesitating on the design agency idea, but only recently remembered I could apply the same plan to automation.

I noticed n8n popping up in thumbnails and video titles, and talking to a friend the other day, he mentioned using n8n as well, so I checked the search stats on vidIq.

I’ve been using vidIq for a while to track different search trends, but I’ve never seen an Overall score of 88, usually high 70s at best.

Clearly there is an opportunity to show up in search results, 1.5 million searches and not very many results for those searches.

What Alex Hormozi refers to as a “starving market.”

Looks like I’ll be pivoting to use n8n for a little while.

I see all these platforms, Make.com, Zapier, and n8n as interchangeable so I’m not too concerned about switching.

n8n seems like a great place to start, plus it’s a little more developer friendly than Make or Zapier.

I recorded a quick intro video to n8n this week, but with good new-YouTube-channel-etiquette, I’ll wait to publish the channel until I have at least 5 videos ready to go.

Cold email and n8n tutorials will be the main focus next week.

Hopefully I’ll have a link to a YouTube channel ready this time next Friday 🤞

o3 Cancelled

Going to skip the obvious OpenAI drama from this week, but I did think it was interesting that Altman said they are reworking their AI model structures.

There is a huge list of variations now.

He even said he doesn’t like the model picker and wants to unify to simple options.

From a ChatGPT perspective, that makes sense to me.

I have so many chat threads open that it becomes tough to keep track of which conversation was which.

At a high level, each model has it’s own particular use case, and I would switch based on what kind of question I was asking, like if I didn’t need reasoning I would use o3-mini, or if I only needed to read and output JSON, I’d stick with 4o, but recently I’ve found myself just using “o3-mini-high” for everything because it’s easiest and there’s no limit for its use.

As a chat user, I think just the text box, and buttons for search, reasoning, deep research are enough.

Let the model decide what to do with the question.

But on the backend, I would hope all the different model variations remain as API options.

In that case I think it makes sense to optimize certain models for specific use cases.

Honestly I can’t say that I’m bummed that o3 was cancelled, I use them all a lot but not sure I would really be able to tell the difference if they shipped “4.5 with reasoning” or o3.

I’ll look forward to GPT-4.5 though, usually means better performance and cheaper prices.

So long o3!

Not Firebase

Besides a short API limit inconvenience, another unexpected headache this week was trying to setup the backend for the app I’ve been developing in React Native.

Over the course of 3 evenings, I never once was able to get a stable Sign Up screen to work when trying to connect to Firebase.

Admittedly, I let Claude Sonnet take the reigns for a while, which might have been part of the problem.

I’ve grown to trust it because, overall, it’s done really well with the front end.

But when things weren’t working on the backend, I passed my code over to o3-mini-high and it identified I was using 2 different Firebase libraries, suggesting I stick with the FirebaseJS library.

I started over, implementing firebase@9 and proceeded to go round and round with conflicting modules that prevented me from even pre-building the app.

Reading over Firebase’s documentation, they recommended the unofficial react-native-firebase library instead so I ended up trying that.

Of course it doesn’t play very nicely with Expo.

Luckily the react-native-firebase library is very well documented and did outline Expo setups specifically in their Getting Started guide, and o3-mini-high (o3mh for short?both roll off the tongue worse than “gpt-4o”) did mention that Expo could be an issue with Firebase.

Now I could give up on Expo, but so far it has helped streamline the development process so much over the last couple months, I hated to have to rework my setup, so I decided to stick with getting Expo and Firebase to play nicely.

But soon I was back to conflicting modules and imports, unable to even get the app to pre-build successfully, let alone start up.

Venting my frustrations to o3-mini-high and Claude separately, they both recommended Supabase as an alternative.

Reluctantly I signed up for a free tier.

I then had Claude spit out the code for a basic Supabase connection to the Sign Up button, plugged in my new account credentials, successfully prebuilt the app, booted it up, clicked on the Sign Up button and…

I got an error.

But it was a good error!

I didn’t enter an email or password when I clicked Sign Up so I would hope I’d get an error.

It worked!

The screen loaded, no pre-build or start time issues, and I got a legit error back as a result of a connection to Supabase.

Looks like I’ll be going forward with Supabase which means using a SQL database rather than the lawless wild west of Firebase’s document database.

SQL takes more planning but is probably for the best, forcing me to avoid the inevitable data disaster that would have happened as a result of accidentally dumping inconsistently structured objects into NoSQL.

Strong typing can useful like that.

But in a total of maybe 15 minutes last night, my opinion of Supabase skyrocketed.

If the rest of the implementation goes that smoothly, I’ll be soon be a big fan of Supabase, not Firebase.

When it comes to React Native anyway.

Firebase can wait in the toolbox for a better use case.

And that’s it for this week! So long o3 and Firebase, here’s to many more opportunities.

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!