Blog

Thoughts on Python, Agentic Systems, AI Projects, and Web Development.

• The New Developer Workflow: Shipping with Codex, Issues, and LLM-First Execution
  2026-03-18

  Modern developers are no longer just writing code. They're framing problems, writing better issues, steering tools like Codex, reviewing diffs, and managing an end-to-end workflow with AI in the loop.

• Building Ghostfrog eBay Edge: An eBay Intelligence Tool for Sellers Who Want to Move Faster
  2026-03-18

  Ghostfrog eBay Edge is being built as a focused intelligence product for sellers who want clearer market signals, stronger listings, and faster, evidence-led decisions.

• Building a Magento 2 UCP Extension (Docker Dev): The Plan for a Best-in-Class Checkout Experience
  2026-01-27

  We're building a Magento 2 UCP adapter (Universal Commerce Protocol) inside a local Docker setup. This tutorial explains the goal, the architecture, and the step-by-step plan to deliver a faster, cleaner checkout experience — without adding extra services or Python.

• From eBay V1 Scripts to a Proper Pipeline: Introducing LangGraph
  2025-12-27

  After proving the concept with a first-pass eBay scraping

• LLM Runtime Glossary: The Stuff Nobody Explains
  2025-12-17

  A plain-English glossary of LLM runtime, API, and inference terms you actually encounter when building real systems with local and hosted models.

• Lesson 1: LLMs Are Not Functions (And Why That Changes Everything)
  2025-12-14

  My first real AI engineering lesson: large language models are not functions. They're probabilistic generators — and unless you constrain them, they will break your system.

• How LLMs Actually Work: The Mental Model Every AI Developer Needs
  2025-12-11

  Forget the mystical hype. Here's a blunt, developer-friendly explanation of how transformers think, plan, and hallucinate.

• How to Architect a Real AI Application in 2025
  2025-12-11

  Everyone builds AI demos. Here's how to build production-grade AI systems with layers, tools, memory, observability, and fault tolerance.