Recent posts
AI-powered development: Chrome extension and n8n automation with Claude Code
A Chrome extension + n8n automation pipeline built with Claude Code in 30 minutes to automate my AI tech watch sharing.
Why we’re moving from Cursor to Claude Code (and why you should too)
Discover why I switched from Cursor to Claude Code. A Lead Frontend Developer’s honest feedback on the limitations of AI IDEs and the power of CLI coding agents in 2026.
Efficiently cleaning a TypeScript project: Knip + AI agent workflow
Learn how to safely clean dead code in TypeScript projects using Knip and an AI Agent. A step-by-step workflow to automate cleanup without breaking production
How Gemini 3 Pro cut our pipeline time in half
We cut our test pipeline time from 22s to 4s per file using Gemini 3 Pro. Learn how AI identified "God Modules" killing our Vitest performance and automated the architectural refactoring that saved us 15 minutes per CI run.
AI Radar
60 entries
Other people's writing on AI agents and LLMs, picked after reading.
- · generativeai.pub
His entire AI skills library loads in 660 tokens.
Discover strategies to minimize token consumption in AI agent skills using Matt Pocock’s techniques like leading words, progressive disclosure, and modular invocation patterns.
- · generativeai.pub
Jev Doesn’t Talk. It Answers. Why a Model With No Text Output Might Matter More Than GPT-6
Explore TypeSafe's Jev model, which outputs calibrated probabilities instead of text. Learn about its API structure, cost efficiency compared to GPT-6, and the implications of RLCD training for automated decision-making systems.
- · claude.dev
Getting the Most Out of Opus 5.5 in Claude and Claude Code
Learn to prompt Opus 5.5 effectively for coding tasks, configure CLAUDE.md for long runs, manage subagents, and handle security flags in Claude Code.
- · coinbase.com
Interviewing Engineers in the AI Era: Lessons from a Year of Rebuilding
Learn how Coinbase redesigned its engineering interview loop to assess AI fluency instead of pattern recall, focusing on directing models and evaluating output.