What actually worked, and what didn't.
I write from the constrained end of AI engineering - one laptop, 6 GB of VRAM, no cluster to hide behind. Every post below is something I hit, fixed, and would want to read if I hit it again. Published on Medium.
-
How I Stopped AI “Context Pollution” and Drastically Reduced Token Usage
By prompt fifty the assistant forgets your instructions, hallucinates variables and quietly triples your bill - because it re-reads the whole conversation every turn. What actually fixes it.
Read on Medium -
I Built a Job-Search Copilot That Refuses to Lie
Deterministic where possible, agentic where necessary - engineering an honest job-apply pipeline that runs entirely on your laptop.
Read on Medium -
Running Local LLMs on a 6GB GPU Laptop - What Actually Works in 2026 (And What Doesn’t)
Every guide assumes 12 GB, 16 GB, or a $4,000 workstation. This one is for people serious about local AI on a budget laptop with 6 GB of VRAM.
Read on Medium -
Google Just Dropped TurboQuant - And It Changes Everything for Local LLM Inference
Longer context meant the KV cache ate the rest of my VRAM and inference stalled. I sat with that problem for months - then Google Research published something.
Read on Medium -
How I Finally Stopped Worrying About What My AI Agent Was Sending Outside My Machine
A practical deep-dive into running a local agent against a private stock-analysis project - and why it changed how I think about AI agent security.
Read on Medium -
Building RAGNITE: How I Created an Enterprise-Grade Agentic AI Platform in Two Weekends
A technical deep-dive into architecting a production-ready multi-agent system with advanced RAG, vector databases and intelligent conversation management.
Read on Medium
New posts land on Medium first.
Local inference, agent architecture, quantization and the occasional trading detour.