When I was learning how to code, databases were difficult to understand. Frontend came naturally and I'd always manage to get something looking and working the way I wanted to but persisting information was always a blocker. It was a whole different set of things to learn, schemas, sql, primary-keys, etc, in order to get data to persist.
Last year, when I was building indiakittens.com, Cursor made it super easy to create a database by integrating with supabase. I didn't have to think too much. I just explained what I wanted and it set up all the schemas, tables, and just gave me the code that I needed to run on Supabase.
This experience cost me 40$ a month. I wanted users to have the ability to sign-in via Google though and the only way I could get that working was if I paid supabase $40 a month. 25$ for their pro-plan and then 9$ for Google sign-in plus taxes.
Indiakittens.com is a public good. I don't make any money from it and back then, I was ok to spend a little but since then, the costs were adding up and $40 a month felt a bit steep. Plus, I have a lot more ideas for things I want to build nowadays and paying supabase 9$ for Google sign-in for every single website felt a bit much.
I looked around on how to reduce costs and renting a VM was an option.
Today, 12 Nov '26, I finally got around to asking Opus 5.5 to give it a shot. I got a VM from Latitude.sh at 18$ a month and asked Opus 5.5 to do everything that's needed for me to stop using Supabase.
It took a few hours but it managed to do it and everything (almost) worked!
I don't know when I'll stop finding what these coding models can do incredible.
It managed a full migration from Supabase to self-hosted Postgres database on a VM in a few hours. It even set up the VM for me and ensured that all I had to do was open a few tabs and enter some values on a few websites. If I'd enabled computer-use and the model was a bit more permissive, it could have done that by itself too.
Instead of paying 40$ a month to Supabase, I now pay 18$ a month to Latitude.
Supabase's pitch made a lot of sense in the past. They abstract away database complexity to ensure that first-time coders mget up and running as quickly as possible. It's what drove me to them in the first place. But now that the coding models have gotten better, I've switched to running my own database on my own VM.
Where will value accrue a year from now?
Laziness is the only 'lock-in' anymore. Every few weeks, I get around to doing something that I've been putting off for too long and I wonder why I had never gotten around to doing it.
Sometimes the wait is necessary. In this case, I had to wait for the coding models to get better. I'm fairly certain that Opus 4.5 couldn't have done this migration without something going wrong or it taking a lot more time.
— The world we grew up in no longer exists —