@Prehqsy

How I Used Context Engineering to Improve Retrieval

19h ago · 8 min read · In my previous article, I talked about how I cached users' intentions instead of their exact queries to reduce token usage and latency drastically. Context engineering played a HUGE role in doing that

How I cached intention, not queries

Jun 3 · 3 min read · While building a project for a client recently, the system runs an LLM pipeline with multiple LLM calls. I ran into 2 obvious problems — latency and high token usage. I needed a way to kill both birds

Threads vs Processes in Task Queues: The Reliability Tradeoff

May 15 · 3 min read · When I started building Tarsq, the execution model was thread-based. At first, threads felt like the obvious choice: Lightweight Easy shared memory Simple ctx injection Lower overhead Straightfor

The Reflection Pattern in AI: Teaching Models to Think About Their Thinking

Feb 10 · 4 min read · I was building an AI system, and I kept running into the same frustrating problem: the AI would generate outputs that looked great at first glance, but had very subtle issues. wrong calculations, unrealistic assumptions, logic that made no sense. So ...

Retrieval Is Not Resolution: Building a Hallucination-Resistant RAG System with LLMs and SQL Server

Dec 31, 2025 · 6 min read · The problem (it looked simple until I tried it) I started building what I thought was a straightforward system: user types a gadget name → system returns the price That’s it. Nothing fancy. Then reality hit. Users don’t type “apple iphone 12 pro 12...