Chapter üyelerimizin yazdığı teknik makaleleri, etkinlik özetlerini ve deneyim paylaşımlarını Medium üzerinden yayınlıyoruz.
I actually met Flutter a bit before my internship, but that exposure was quite limited. The internship was my first chance to use Flutter in a real project — I worked on the frontend, building different screens and shared components while adapting to the existing project structure.
How I built a natural-language database assistant that refuses dangerous queries, checks its own answers, and tells you how much to trust them — I wanted to show more than "I can call an API," but that I think about what happens after the model gives an answer.
In the first post of this series, ConvNeXt kept coming back to one comparison point: Swin Transformer. This post asks a more basic question: what made Swin necessary in the first place, when Vision Transformer already existed a year earlier?
Most people pick a backbone the same way: whichever one shows up first in a tutorial, or whichever one "everyone" seems to be using. I ran a controlled experiment instead — same dataset, same hyperparameters, three backbones trained side by side so it would actually mean something.
Collapsing the chaos of dozens of services into a single door — and bringing it to life with .NET.
In ResNet, each layer talks directly only to the one before it. But what if every layer talked directly to all the layers before it? That is exactly the idea DenseNet is built on, and it won the Best Paper Award at CVPR 2017.
When Swin Transformer arrived, there was a feeling in the air: the future of computer vision belonged to attention. A group of researchers at Facebook AI Research asked a simple but uncomfortable question: do these gains really come from attention — or from everything else around the Transformer?
Almost everything we talk about today — Vision Transformers, ResNet, YOLO, segmentation models, even the giant networks that generate images — traces back to a single moment in 2012, when AlexNet won the ImageNet competition by such a wide margin that computer vision changed direction overnight.
I've been writing Java and Go for a while now. Both languages taught me to think of a backend service as a living process that stays alive waiting for the next request. Then I started learning PHP, and I had a genuinely confusing moment.
Yazılım dünyasında kartlar yeniden dağıtılıyor. Artık sadece bir yapay zeka asistanınız yok; projeniz için 7/24 çalışan, uzmanlaşmış ve otonom bir dijital geliştirme ekibiniz var.
NGINX stands as a cornerstone of modern web infrastructure, renowned for handling high concurrency with minimal resources. This guide walks through web server fundamentals, NGINX's event-driven architecture, and its roles as a reverse proxy and load balancer.
The first rule of modern backend design is simple: never connect your server directly to the internet. This is where Proxy Servers come in — sitting between the client and the server, filtering traffic and distributing the load.
On November 18, 2025, the internet suddenly had a big problem — popular apps like ChatGPT, Spotify, and X stopped working. Everyone thought it was a cyber attack, but the real reason was a very specific technical detail: the unwrap() function in the Rust programming language.
When you're close to deploying your backend, one of the most important questions suddenly hits you: "Wait… how secure is my API really?" I was using plain HTTP, which meant that all the sensitive data in my request headers, including my JWT token, was being sent in clear text over the network.