ai
Practical applications of AI in software engineering, local LLMs, prompt engineering, and integrating intelligence into production systems.
4 articles
My Local AI Setup, Six Months Later: 48 GB, Qwen3.8, and a Different Bottleneck
TL;DR Six months after the 16 GB setup: 48 GB, one default 27B model, Splash inference in LM Studio, and 64K context, because the constraint moved from fitting a model to keeping long agent sessions usable.
A few months ago, I wrote about running a useful local AI development setup on a MacBook Pro with 16 GB of unified memory. At the time, memory shaped almost every decision. I experimented with smaller models, limited context carefully, and looked for a combination that was capable enough for coding while still leaving room for my IDE, containers, browser, and the rest of my development environment.
Read more →My Local AI Coding Toolkit: How I Keep Agents Fast and Token-Efficient
TL;DR How I optimize the environment around AI coding agents to reduce token waste and keep sessions efficient.
AI coding agents have become part of my daily workflow. I use OpenCode and Claude Code for everything from exploring unfamiliar codebases and investigating bugs to implementing changes, reviewing code, and working through architectural questions.
Read more →Making Angular AI Work in Real Projects
TL;DR AI tools are great at generating code, but real Angular projects succeed only when you give them rich project context, architecture boundaries, and conventions.
AI can generate a component, write a test, or explain a compiler error. But in real applications, results depend more on project context than model capability. It doesn't understand your architecture, your conventions, or the boundaries you built over years.
Read more →Local AI on a MacBook Pro with 16 GB RAM: What Actually Works
TL;DR Practical guide to running local AI coding assistants on a 16GB M1 MacBook Pro, what works well, what doesn't, and the real setup trade-offs.
Running AI coding assistants locally sounds appealing: no API costs, your code stays on your machine, and you get a Claude Code-like experience for free. But getting it to actually work well on a 16 GB MacBook Pro M1 takes more trial and error than most guides admit.
Read more →
Ammar Najjar