#ai
8 articles tagged ai.
Designing the AI Request Pipeline: 8 Layers Between User Input and Your LLM
A production architecture blueprint for every layer between a user's message and a safe, reliable LLM response — with Spring Boot code you can actually use.
We Let Claude Code Refactor Our 200K-Line Java Monolith. Here's the Honest Result.
A real post-mortem on 6 weeks of AI-assisted refactoring — what it got right, what it silently broke, what we'd do differently, and the hybrid model we actually shipped.
Building AI Skills: GitHub Copilot Extensions, Claude Tools, and Reusable Agent Capabilities
One skill, three platforms. A complete guide to building a custom capability for GitHub Copilot, Claude, and your own AI agents — with real code, architecture diagrams, and a decision framework for when to use each.
AI Code Review at Scale: How We Use Claude to Review Every PR Before Humans See It
Six months, 4,200 pull requests, and a GitHub Actions workflow that catches real bugs before your team spends a single minute on review. The architecture, the prompt, the cost, and what we learned.
Feature Flags for AI: The Deployment Pattern That Saved Us From 3 Production Disasters
Why LLM features need flags more than anything else you've ever shipped — and the four flag patterns that let us swap models, A/B test prompts, and kill broken AI features in under 60 seconds.
AI Coding Tools in Enterprise: What 90 Days of Data Actually Shows
A structured 90-day study of AI coding tools across a 70-engineer enterprise team, measuring PR velocity, defect rates, adoption patterns, and ROI. The results were not all positive — and the most important finding had nothing to do with speed.
Agent Memory Patterns: How to Give Your LLM Application a Brain That Persists
Your LLM forgets everything the moment the request ends. Here's how production systems build short-term, long-term, and semantic memory — with Spring AI, Redis, and pgvector examples.
Why Most Enterprise AI Projects Die in Production (And the 5 Patterns That Actually Survive)
I've deployed AI systems to 10,000+ enterprise users across healthcare, logistics, and financial services. The projects that failed didn't fail because of the model. They failed because of decisions made before the first line of model code was written.