Gone Full-Stack, Back Full-Baked

Hope that everyone is doing great! This Blog would be completely written by typing manually. I would be sharing my 2-month journey of going full-stack, the GenAI coding experiences, the open source contributions I did during this period of time, and what's next in my tech adventure.

List of Contents

28 Days of Full-Stack: From Zero to... Half-Baked?

I usually never liked doing frontend before , mostly because I was interested in cybersecurity from 2023 till mid-2025 and usually just liked interacting with my terminal. But past 4 months I started to enjoy doing frontend and wanted to properly learn it, Luck by chance i stumbled upon Scrimba's Full Stack Course which is of WHOPPING 108.2 Hours! It sounded impossible to finish within 1 month. What I did next was... I decided to challenge myself to complete it within the next 14 days (But took me 2x time). That means I had to commit to watching almost 9 hours of content PER DAY. so i tried to do so

I didn't stop here for sure , I kept learning tools like GSAP , Three.js , TailwindCSS and planning to learn p5.js but as an hobby

My Thoughts on GenAI Coding

My Advice to anyone learning how to code and is in their first or second year, DONT DO IT! It's not that you shouldn't use them, but rather

• Learn on your Fundamentals
• Practice , Practice , Practice
• Understand the Concepts

Because GenAI coding in my opinion HEAVILY relies on how you understand the fundamentals of the language you are using. Why? You can give better prompts that would target what you really want to , not what the AI blindly guesses because you said "please fix it" , AI models are not mind readers but they are your best possible coding buddy you could ask for , I do not hate GenAI coding, But the claims on "AI will replace programmers" is just pure BS in my opinion. They are just tools to help you speed up your workflow, not replace you. Human Intelligence works completely different from current LLM models. Watch this Presentation by Yann LeCun who was the Chief AI Scientist at Meta an month ago explaining why the current LLM hyped LLM architecture will not lead to AGI. GenAI coding is an potential "tutorial hell" killer for me atleast.

I will give you examples of where GenAI lacks currently: The LLM knowledge base is heavily dependent on when it was made Eg.

• TailwindCSS v4 (March 2025) was a major update that changed the way tailwind works , like direct css implentation and removal of tailwind.config.js
• React v19 added new hook and added "Actions" for better data fetching

Manually Teaching those LLMs the latest updates was necessary if i wanted to use them effectively.

If you ask an LLM like Claude to make a new project from scratch , it would use tailwindcss v3.4 and keep us in an loop of postcss errors in some cases. This is just one of the examples which I could notice. I know its an no brainer it leads to an cascade of errors later on when we try to add new components from newly updated component providers like fancy-components , shadcn-ui etc.

Disclaimer: I could be completely wrong on this aswell ;.; so take it with a grain of salt.

Heres how I code with GenAI currently:

• If its an new project , I would provide the LLM with the entire plan in either markdown or xml format
• Ask it to bisect every major step in project. Usually I ask it to Make Daily Milestones(Day1.md - Day7.md)
• As my primary objective is to learn, I Created an Agents.md file to teach me what it did in an manner i like to learn eg: Teach me like you are an Tutor from youtube.
• I do not use a single chat for more than 30 minutes. What I usually do is switch to a new chat every 30 minutes. But before i do so , i ask it to create an md file on what changes it did
• I am aware this is inefficient method. But this method really reduced the amount of prompting i had to do

Diving Deep into the Open Source Ocean

I had plans to contribute to open source projects since Oct 2023, during Hacktoberfest 2023. But this excitement towards open source was for sure outsourced from one of my seniors. As i had interests on cybersecurity , I went on a voyage on finding the right open source projects (which I never could during that time because finding beginner friendly cybersecurity projects is extremely hard trust me). Fast forward to Sept 2025, I finally decided to contribute to FossFLOW (an isometric diagramming tool)

FossFLOW: My First Major Contribution

I wanted to contribute because I found the same issue every small screen browser user was going through at the Add Item Dialog,#125 also because the number of active contributors for this repo was significantly lesser than usual at the same star range (currently the repo stands at 11.8k stars)

My second contribution was fixing the icon scaling issue when we import newer SVG or isometric icons,#112

My third contribution was adding a new feature to export the diagram by removing the white spaces around the diagram , which i had to implement with two different approaches , the first one was finding the outermost icons on the x and y axis then do padding on that basis. but the maintainer wanted to preserve the original non cropped image download

,so i converted it into an toggle feature that would allow the user to crop the image as they want to.

#138,#142,#146 - These were the PRs I made in the span of 10 days

Hacktoberfest 2025: The Pull Request Marathon

Personally , Not the best way to contribute to open source , but I did it anyway. In my opinion finding an issue that was not assigned to already to someone within the next 2 hours was the toughest part. contributed to like 5 different repositories and made 7 PRs - 6 merged and 1 closed because i wanted to touch some grass.

During this phase I got a chance to attend Digital Ocean x Hacktoberfest mini hackathon, we had to leverage DigitalOcean Gradient in our project which we managed to do so , from creating the AI Agent , adding the knowledge bases to connecting it to our frontend surprisingly within 3 hours

Project Spotlight: Karya.AI

This project was an unfinished hackathon project which me and one of my friends decided to complete. I refactored the frontend using React.js + Typescript + TailwindCSS + Components using ogl.

Project Spotlight: Qoal

This was my personal project because i wanted to refine my Go skills and also wanted to make something with GSAP and Three.js . This project was initially planned to be an Django Backend project then decided to switch to Go backend later on. Mostly this decision was made because of my syntax preferences. instead of talking about the features which is already mentioned in my github repo, I will be focusing on challenges I faced.

• Render has an issue with React Router Redirects as it is client side routing: we have to manually add the trailing slash to the URL.
• Initially I had planned to keep file conversion logic entirely local, but its not possible because i would run out of memory and conversion needs a reliable storage other than localStorage shifted to S3 Buckets with Lifecycle Policies to auto delete files after 1 day
• I couldn't test all the possible edge cases (all possible file conversions)
• Go provides robust testing built in the language, I will probably use it for the above reason
• PostgreSQL would be only active till the next month , would need to migrate the Postgres to another db service and provide the url

Why did i choose Render? Because it was free and I have Go backend, I cant deploy it in Hosting Services like Netlify or Vercel. It is possible to do so but i have to write Go Serverless Functions instead which is not ideal

The End Credits (For Now)

Heres my hitlist of things to learn next:

• Kubernetes
• Ansible or Terraform
• Jenkins
• Grafana & Prometheus

Before I go conclude this blog post I will be hinting on what the next blog post could be about

• An Deployed Next.js Project with GSAP, Three.js and Spline?
• System Design Case Studies??
• Guide to Open Source Contributions? (I am clearly not the best at it)
• Learn DevOps Properly?

For Now, Thank you for reading this blog post till the end, I really appreciate it :) See you in the next one!