LLM Temperature, Explained Simply
A concise explanation of LLM temperature, how it changes next-token probabilities, and when to use low, medium, or high values.
All posts / AI
Why LiteLLM Matters for AI Engineering Teams
Why LiteLLM is useful for Codex and Claude Code, how to get it running quickly, and why this kind of tooling will matter more as AI agents spread.
The Most Productive Looking Procrastination I've Ever Done
Building with AI agents is so easy it's become my new form of procrastination, one that feels like productivity even when it isn't.
Stitch might be a UI tool worth watching
A short note on why Google's Stitch stands out as an AI-native UI prototyping tool with real potential.
Gemma 4 makes local models exciting again
A short reflection on why trying Gemma 4 made local AI feel practical and exciting again.
How to Reduce Token Consumption with Claude Code and Codex
Practical habits and patterns that keep your token usage lean when working with AI coding agents, whether you use Claude Code or GitHub Copilot with Codex.
Claude's New Telegram Plugin: Talking to Your AI Agent on the Go
Anthropic just released an official Telegram plugin for Claude Code. Here's how it works and why it changes how I interact with my AI agent.
Performance Tracing with Chrome DevTools MCP and The AI Agent
How I used the Chrome DevTools MCP to trace, diagnose, and fix real performance issues on my photo gallery page — without touching Lighthouse manually.
Spec-Driven Development in Practice: A Walkthrough with Spec Kit
A practical walkthrough of using Spec Kit to apply Spec-Driven Development inside an existing Next.js project
The Shift from Vibe Coding to Spec-Driven Development
A look at how the industry is moving from rapid, conversational AI coding to structured, specification-first development and the tools driving that shift.
Improving Lighthouse Performance with Image Optimization
How I improved Lighthouse performance on image-heavy blog posts using staged optimizations: lazy loading, WebP conversion, and LCP/CLS tuning.
Upgrading a 4-Year-Old Next.js Blog with AIAI
What it took to go from Next.js 9.5, 136 vulnerabilities, and an abandoned MDX plugin to a fully modern stack with zero vulnerabilities — using AI-assisted development.
My 2026 Tech Radar: Tools and Skills I'm Exploring This Year
A curated exploration of the technologies I'm learning in 2026—from AI-powered development and data engineering to infrastructure, mobile development, and ethical hacking.
Human Creativity Is the Real AI Breakthrough
The biggest leaps in AI aren't coming from better models — they're coming from creative engineers who know how to build smarter software around the ones we already have.
Processing Your Handwritten To-Do List Has Never Been Easier
How I use ChatGPT to process handwritten to-do lists into structured, time-blocked markdown for Obsidian—saving time every morning.
Are Layoffs the New Norm in the Age of AI?
Tech layoffs are accelerating—and AI is often cited as the reason. Is this the new normal? A personal reflection on staying adaptable.