Browse all posts in this category
A modern cat weighs tens of kilobytes and drags the whole libc behind it. Does it have to? UOLT says no: every tool is hand-written in x86_64 assembly, no libc, no heap, talking to the kernel directly. The full suite fits in the space a single coreutils binary takes. Here is the thinking behind it.
Every time you ask ChatGPT a question, a datacenter somewhere in the world burns electricity for a few cents. Multiply by billions of requests per day. AI, as it is built today, has a fundamental problem nobody is really solving: it consumes an absurd amount of energy.
A developer opens their tech feed and finds three articles: 'How to build a RAG agent with your favorite LLM', 'Fine-tuning is dead, long live embeddings', 'MCP will revolutionize AI workflows'. They close the tab, not much wiser. This guide is for them.
Monday, you create a branch for your new feature. Friday, you try to merge. The conflict is 400 lines long. You spend the afternoon untangling what your colleagues did while you were working in isolation. This scenario has a name: merge hell. And it has an identifiable cause.
Your team delivers 40 story points per sprint. The backlog moves. Tickets get closed. And yet users have been waiting on features for six months. Something is wrong. The problem: you are measuring activity, not value delivery.
Complexity builds itself. Simplicity is a choice. KISS isn't intellectual laziness: it's a discipline that requires resisting the urge to add, integrate, and adopt everything that shines.
Angular, React, Vue, Svelte, Next.js, Remix, SolidJS, Qwik, Astro... JavaScript invents a new 'future of the web' every eighteen months. Ruby on Rails turns 21 this year. Which one is right?
After years of selling microservices as the universal solution, the industry is returning to a simpler idea: a single deployment, but organized around business domains. The modular monolith is making a comeback - and this time, it is intentional.
"We should move to microservices." This sentence has become the signal of a technical decision made by imitation rather than necessity. Here's why, in the majority of cases, a monolith is a better choice.
