Arda Gür

Computer Science student at @universitemDEU
Writing about AI, backend & things I build

[Arda Gür @devardagur]

i benchmarked Jev vs OpenAI Decisions API on the same 1,000 customer messages.
nearly tied on accuracy, Jev is ~2.8× cheaper, OpenAI is ~1.4× faster.

full breakdown on my blog: https://t.co/KKv7ohYbu5

[Arda Gür @devardagur]

@MrrrOzi couldn't agree more :)

[Arda Gür @devardagur]

openai released the new "Decisions API" yesterday and after my recent research on jev, this caught my attention immediately. ai is developing really fast in this direction
if you re interested: https://t.co/GBxG2FOG5X

[Arda Gür @devardagur]

@topphamilia definitely. writing is the ultimate test of true understanding. thanks for the feedback, appreciate it!

[Arda Gür @devardagur]

i finally launched my own blog.
https://t.co/nlpk3USbLW

i’ll be writing about AI, backend development, and the small projects I build. I also want to share the problems I run into while learning and how I solve them.

my first post is already live. From now on, I’ll try to put what I learn into practice and write about it along the way.

[Arda Gür @devardagur]

im reading Alex Xu's System Design Interview book.
the diagrams make the concepts really memorable visually but some sections fall a bit short on explaining "why was it designed this way, what would the alternative be" — i fill those gaps by asking claude.

[Arda Gür @devardagur]

is every advanced technique in RAG actually an improvement, or just added complexity?
im trying to measure the real impact using an eval framework like Ragas. i want to move forward with numbers, not assumptions.

focusing on this today!!

[Arda Gür @devardagur]

A lot of people using CharacterTextSplitter in LangChain end up wondering why their chunk_size keeps getting exceeded.

The reason's simple: it only has one separator. If that separator doesn't show up enough in the text, there's nothing left to fall back on.

RecursiveCharacterTextSplitter works differently — it takes a list: ["\n\n", "\n", " ", ""]. If splitting by paragraph isn't enough, it drops down to lines, then words, until it actually fits chunk_size.

End result: less broken-up text, more preserved meaning. That's basically why it's the go-to for most RAG pipelines.

[Arda Gür @devardagur]

the most confusing part of rag pipelines in production is usually the "advanced" optimizations - chunking, retrieval quality, reranking...

before diving into those, i ve been revisiting the fundamentals of rag over the last couple of days to build a solid base, and i'll be moving into advanced topics and optimization problems next

[Arda Gür @devardagur]

If you're experimenting with RAG locally, you don't need to sign up or set up a server for a vector database. Just add ChromaDB to your project and it'll keep your vectors right in your working directory. Just a tip

[Arda Gür @devardagur]

even as i keep working in Ai, taking a moment to refresh the basics and explore what different platforms offer is always a good time

if you re just getting started and want to build a strong foundation, this is a fantastic place to begin
Definitely recommend!!

[Arda Gür @devardagur]

First, i need to have a good relationship with my assistant. I will take the following courses to use claude code more productively:
1. Software Development with Claude Code - Datacamp
2. Claude Code 101 - Anthropic Academy

[Arda Gür @devardagur]

The only thing i've seen lately is:
"AI will take our jobs."

I wanna be someone who uses ai powerfully, not someone who fears it. Ai can be a good assistant for coding and a good teacher for learning. I wanna use it very productively, develop great open-source projects and create content with it. The idea of ​​sharing my process with you guys here excites me a lot.

[Arda Gür @devardagur]

@ScalerSohom thank u for comment 🙏 it was an exciting project for me

[Arda Gür @devardagur]

i developed a RAG technology chatbot for e-commerce sites. i'll leave the link here:
https://t.co/8MMOK1kVYi

[Arda Gür @devardagur]

if you search for 'recursion' on google, you're greeted by a recursion algorithm
google computer scientists are legends 🤯

[Arda Gür @devardagur]

i just learned the text classifier node and added it to my project
it is a node used to classify the input data according to the category you want
enter category name, description and select llm you want. then enjoy.

[Arda Gür @devardagur]

today i am starting a new project with n8n. i will share it when it is finished 🔥

[Arda Gür @devardagur]

there are great lists for n8n tutorials on n8n's yt channel.

link: https://t.co/cagvlJscyD

[Arda Gür @devardagur]

today i will try to learn about embeddings and vector databases