Change of plans.

The big thing I was planning do this week was deprioritized for another big thing.

Way more developing and way less YouTube, but not such a bag thing.

Topics for this week:

• Cold Booking

• Distribution

• Too Complicated for AI

• Supabase Impressions

• The New Number 1

Cold Booking

I planned to drop a YouTube channel this week with 5+ n8n tutorials on how to automate your life but…

It’s a different crunch-time.

Time to get my client’s app out the door, so no automations this week.

But I did have my first call with a lead from cold emailing which was exciting.

Granted I did have to pay if he accepted the call, but I still consider it a win.

Absolutely worth the experience of fumbling my through trying to explain to someone I’ve never met how my brand new business works.

Ultimately the price was too high, but thankfully he was very friendly and as helpful as he could be, offering me other marketing ideas and connecting me with another potential source of leads, and who knows, may be an opportunity in the future when he does have the budget for automation.

Over the last 2 weeks, I’ve sent 872 cold emails and had 2 booked calls (the other call isn’t until March so hopefully it goes).

I can definitely see why cold email is still a viable strategy these days.

Horribly inefficient in terms of conversion, 2 out of 872 is 0.22% but it only takes a couple hours of work to set that system up to run automatically everyday for a couple weeks.

600 per day could net 1 interested response, 3000 per week could 5 interested responses per week.

2 hours of setup per month for potentially 100 leads isn’t so bad.

I’ve let the email campaign run out for now though while I focus on the app, but I’m thinking I’ll crank it up to pro numbers once I dive back into automation full-time.

That with YouTube, and potentially a community element like skool, I’ll have to be an automation expert to keep up with everything.

Distribution

It’s always a distribution problem.

Talking to my first cold email lead ever this week, he identified that problem and recommended I become the distribution guy, which I’ve heavily considered.

It’s been the biggest blocker for me by far.

It’s great to have a good product or message, but no one knows you have it and odds are they aren’t going to come find it unless you happen to hit that magical sweet-spot of creating something that everyone needs, or that lots of people need but they also need their friends to get it.

I’ve been circling around this concept for a long time, which is what led back to automation and offering tools to increase distribution.

Most likely this will end up being my automation niche.

But we’ll see once I pick up the cold emails and get YouTube videos out there to see what people respond want.

Too Complicated for AI

I’ve been pro AI-development for a while now.

Claude and ChatGPT have been able to churn out considerable chunks of code for me in the various projects I’ve worked on, enabling me to make way more progress than I would have on my own.

But I seem to have found a limitation.

React State Management.

It made me remember why I don’t like React.

And AI seems to struggle with it as well.

I ran into a different issue last week with Firebase not cooperating with React Native and Expo, and luckily Supabase has been much simpler.

But once connected to Supabase, Claude, Deepseek, and o3-mini-high all completely tangled up the authentication flow and state management.

I wasted an entire day trying to get the app to successfully:

• Sign Up a new user

• Log In an existing user

• skip Log In if you’d already signed in

There must be too many layers and states in React for AI to be able to reason through that process when giving the existing code it had already recommended.

Usually could 2 of 3 cases working, but fixing the third would break the other 2.

I’m sure I could have provided better prompts or could have had better results by producing pieces step by step, but I had a horrible time getting any of them to generate anything that could handle all 3 conditions.

I know I used AI to generate the authentication process with Go and HTMX when I was working on hypebid.tv and I know I ran into some issues there too but not nearly the tangled mess that I ended up with for this React Native app.

I suppose it’s lesson learned at this point, became too reliant on AI to do everything, but up until this point I have had great results.

Circular conversations with Claude here and there but nothing this blocking.

I’m not giving up on AI coding entirely, but I will be relying a little less on it for really complicated flows, although it still has been pretty good at debugging some weird state issues.

Just might not quite be to the point of “Hey, create me an entire fully featured and scalable app.”

Grok 3 just came out though so maybe it can…

Supabase Impressions

I only briefly used Supabase last week to successfully fail to login it, but I was more than satisfied with the result then.

I’ve spent quite a bit more time with it this week and so far I’ve been pleased.

Supabase has a nice interface for looking at managing tables and it’s been easy enough to generate query syntax using their client.

Their own built in AI on their website has been fairly helpful too in reviewing query logic and offering better solutions, though a couple times it’s offered the wrong solution to it’s own product.

But Supabase is surprisingly restrictive in some ways too.

They don’t give you API access to all the normal Join mechanics of a SQL database which is kind of pain, and they don’t let you run manually written queries either, which that I get, that would be way more of a headache for them to have validate on the fly.

I might also not be using Supabase for what it’s intended.

I’m used to having full access to a database but their focus is scaling efficient use of a database.

It’s a perfect service to get started and scale an app until you run into good problems like having too many users or needing to re-optimize queries for performance.

Overall I’m pleased with it, I would happily recommend it as an easy way to spin up a backend for a new app.

The New Number 1

xAI released Grok 3 into beta this week and so far it seems like it’s performing very well, currently at the top of the Chatbot Arena Leaderboard, which is pretty rare to see OpenAI dethroned from the number 1 spot.

I do have access to it but I haven’t played around much yet, although I may be throwing more code into it if I continue to run into issue with React Native.

During the live stream reveal, it was able to simulate the timing for a rocket to use efficiently use the gravity and orbits of Earth and Mars to more efficiently travel between the two, and it also generated a game that was a cross between Bejeweled and Tetris.

Both seemed to work on their first prompt which is always a good thing, no follow up prompting required.

When I got access to it, I was able to have it research my current automation strategy and thankfully it reported a very positive outlook on the prospect:

I haven’t upgraded to API access yet, but I’m sure I’ll end up dumping some code into the chat interface this weekend to see if “Think” mode can handle it.

If Grok can solve one of the hardest problems in coding, React State Management, then I would happily switch to pay for API access.

At least until the next big AI model comes out next week…

And that’s it for this week! Way more app development than anything else this week, but still plenty of time for AI prompting.

Those are the links and ideas 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!