Solved 500 LeetCode problems today.
I honestly thought I'd feel more confident by now, but I still don't. Maybe it's because I haven't cracked an interview yet, or maybe I'm just overthinking.
Half the year gone, GitHub graph looking better than my actual life rn.
Built a collab code editor with CRDTs + WebRTC just to understand how Google Docs sync works.Learned Go and Rust for no real reason except I wanted to.
Let's see how the next half of the year goes.
Most people imagine LLMs "reading" text like humans.
They don't.
Text is split into tokens (subwords, not words).
Tokens become ID.
ID is looked up in massive embedding tables to produce vector.
Those vectors flow through layers of matrix multiplications and other operations
A user requests a file.
You see one request.
The system sees:
CDN, Cache, Rate Limiting, Storage, Queues, and Monitoring.
That's the difference between building an app and building a system that scales.
Thanks @tiago_taquelim_
Don't call your Auth service on every request.
Validate the JWT at the API Gateway.
Reject invalid tokens there.
Forward only trusted requests to your services.
Bad traffic dies at the edge.
That's how scalable authentication works in microservices.
One backend concept every developer should know: Pagination.
Don't return 10,000 records in a single request.
const skip = (page - 1) * limit;
Fetch only what users need, and return metadata like:
• totalPages
• totalRecords
• hasNext
• hasPrev
Simple idea. Massive impact.
Just started my AI/ML journey with @kirat_tw Bootcamp 1.0.
Today's topic: Neural Networks.
2 inputs → 4 hidden neurons → 1 output.
Excited to learn how AI works under the hood.
Just completed my 50-day LeetCode streak. 🎉
Just trying to stay consistent and get a little better every day. Not a huge milestone, but every step counts.
Just watched @bijanbowen's demo of Claude Fable 5, and some of the things it can do are honestly crazy.
The long-horizon reasoning, coding, and tool usage feel like a real step forward.
Based on this video, Claude Fable 5 looks incredibly smart. 🤯
Instead of scrolling for the next few hours, I'm going down a different rabbit hole.
Starting "Building an AI Interviewer From Scratch in 3 Hours" by @kirat_tw.
This is so peak 🙌
Another Productive day 🔥
Submitted my autonomous agent assignment for an internship.
Solved 3 LeetCode problems (2 SQL + 1 DSA).
Spent 2 hours learning Rust.
Spent another 2 hours diving deeper into LangGraph.
Spent the evening building my first chatbot with LangGraph.
It's actually pretty cool how you can think of the conversation as a graph instead of a simple request → response flow. Started with a basic Gemini-powered bot, next step is adding memory and tool calling.
For almost 1.5 years, my routine was simple:
Watch a tutorial → Build the same project.
Today, I'm building my own ideas, contributing to open source, and figuring things out without someone holding my hand.
The journey is slow, but looking back, it's worth it.
Currently building two completely different projects simultaneously for internship opportunities.
One involves AI agents and Figma integration, while the other is in a completely different domain.
Lots of debugging, lots of learning. Let's see how it goes.
Trying to learn LangGraph by building random little projects.
Today's one was a simple state machine that solves a quadratic equation and branches depending on the discriminant.
Small project, but every time I build something like this, the whole idea clicks a bit more. 😄
Happy to share that two of my PRs got merged today 🙂
• Route-level code splitting with React.lazy()
• Manual chunk splitting for vendor libraries
One improved performance.
One taught me that production always finds bugs you didn't 😅
Still learning and still fixing.