I was technically obsessed

- 2/2025


  • Started working full time in Web Dev in 2018

  • back then React was hot and I decided to go head first in it to find good jobs

  • My first profetional experience was to make a first version of an app in a encubator

  • The stack was:

    • React (classes at the time)
    • Redux
    • Redux-Saga
    • Style Components
    • Jest
    • Firebase for the entire backend
    • home made SSR that was migrated to NextJS mid project
  • As a beginner it was so hard to understand Redux and Redux Saga, it took me weeks to feel confortable with it

  • Everything had to be unit tested, even the browser events, spent a lot of time mocking stuff

  • As soon as we lunched it we inmidiatelly saw the Firebase bills coming it, we had drastically under estimated the costs (it was hard to control the backend calls with how complex Redux had gotten)

  • in the latest months React Hooks were released, and we migrated many components to if (we havent even lucnh and we already have a bunch of tech dept)

  • long story short, the app had no business model, and the founder was burned out even before lunch so after a full year of development the app only lasted a couple of months online and is now dead

  • But it was my first experience of seeing how to make a website from scratch, a full app

  • not long after I started a new project, a SAAS for restaurants to make QR menus (see totomenu for the full story).

  • I wanted something “fast”, over optimizing from day 1, and decided to go with Gatsby for their image optimization that looked greate

  • Gatsby is a React based, compiled time site generator, that connects to APIs via GraphQL

  • I deployed in Netlify bc is what the Gatsby comunity was after

  • For the backend I learned my lessons about Firebase, I want control. But I still had that thing that I didn’t want to code auth, image uploading, security and all that, I wanted something that already did most of the work, with a big comunity, bc “making it myself was a waste of time”

  • I thought on using wordpress as an API and Gatsby for the front-end, but it lacked many things I wanted and ended up landing on Strapi and liked it inmidiatelly for it’s nodeJS focus and features (spetially the upload of images to S3, I didn’t want to do that)

  • But that is not all, Gatsby is just the client front end, I need a “back-office” for the restaurant to deal with the products and updates

  • And because I “didn’t want to do it myself” I want for a batteries included Ant Design Pro, a React base UI library made by the Chines firm Alibaba. And it has it all; auth, icons, roles, layouts, translations, tables that are pluged to the API library to automatically call and refresh. It is massive.

  • So now I have a a client side repository, a backend nodejs in Heroku and a back office repository

  • lunched the first version and inmidiatelly started seeing the Netlify bills go up, bc Gastby is “compiled time”, meaning everytime someone made an update the entire website had to re-build (including optimizing ALL of the images)

  • Sandly I was so deep in the development that I just decided to pay and deal with it later

  • my parents had a IT guy that wanted to code, so it was about to join me working on this (this is still a side project)

  • Embaraced from my code and stack, and considering that he had some coding experience but was not a profetional, see all of the mess I had made with react and Styled components, everything all over the place, I decided to look for a solution, “how can I make this daim thing simpler?”

  • looked for days until I decided to try Svelte, and inmidiatelly understand how much work I was doing in React that didn’t needed to be, plus how much simpler it would be for this new guy to work on the code

  • The good and bad thing at that time was that Strapi was going from v3 to v4, and bc they broke EVERYTHING the entire app had to be re-written anyways

  • moved to sveltekit at the same time as strapi v4 (strapi was more work than svelte itself)

  • we grew the app to more users, and everything fine. I hade many complains about Strapi, bc of a bug on their image optimization that lasted years

  • and their authentications and role mehtods were not really good for me

  • so I went with Strapi for the images and auth, but ended up having to code it myself bc theirs actually didn’t do the job for me

  • Their dashboard was not responsive on phones (very annoying for us) and they added a bunch of features that we didn’t want that were getting in the way

  • in 2024 they release strapi v5, and once again they broke everything, all of the endpoints had to be done once again

  • at this time we started working on Parlamento AI, so we needed to set up the project stack, and I didn’t want Strapi anymore, very heavy with “bateries included”, but non of the bateries were useful to us

  • thought on using Primsa bc many people were using it, but landed on Drizzle and it was a instant ‘yes’

  • it had so little abstraction, it was insane

  • then looked into Lucia Auth bc it coped will with Drizzle

  • I made for the first to my first authentication

  • Everything worked so well with Svelte and AI

  • AI allowed for all of the “boring boilerplate” problem to go away, AI is fenomenal for boilerplate

  • I have been working with this stack for a couple of months and the app is far more complex that the last one (feature wise) and the code is so much simpler, is insane.

  • and we are doing complex ffmpeg and scraping with it, and instead of going with some “dockerize” solution with redis to deal with the multiple processes, we are simply lunching subprocesses with the tools arelady existing in nodejs, and verything works prefect

  • my conclusion is that the industry pushed me into building very complex stuff, that was unmantainable, bc it was supposed to “save me time and worries”, but did the exact oposite