Wanted Claude Code to query production data for spam-filter tuning. First instinct: bolt on an MCP server. Reality: a simple skill driving the mysql CLI (read-only) did the job with less complexity.
You might not need MCP 👇
https://t.co/s1bwASt86v
#MCP#ClaudeCode#AIAgents
Running coding agents in parallel? E2E tests colliding over the same port? New /run-in-sandbox skill runs your tests inside Docker so agents stop bumping into each other.
https://t.co/BAEdPjJ6DJ
#CodingAgents#Docker#DevTools
A year ago I was tinkering with AI. Now it's core to how I build. The hard part isn't the code, it's deciding what to build.
Will my job exist in 5 years? Not sure. But the real moat is human connection.
https://t.co/9oeRCnVNaP
#AI#SoftwareEngineering#FutureOfWork
Coding agents make hidden design assumptions. Here's a one-line prompt to surface them so you can challenge them.
https://t.co/qsVKgRDLhf
#AI#ClaudeCode#PromptEngineering
18 months of burpees taught me what makes systems resilient: modest effort, done consistently, with enough slack to survive sick kids and bad weather.
https://t.co/LuZ9cNzj3b
#SystemsThinking#Habits#Fitness
Metaprogramming for LLMs: ask the model to write a prompt, show it to you, then run it straight away. I'm calling it "inline-executed metaprompting" - a chain-of-thought trick in disguise.
https://t.co/8P7iGiNlMP
#PromptEngineering#AI#LLM
AI coding agents are tactical by nature: shortest path, first thing that works. That's a problem - AI builds worse on slop. Our job as engineers now is the strategic work: steering agents toward sound architecture.
https://t.co/IMHgR3VREZ
#SoftwareDesign#AI#ClaudeCode
Want better answers from AI? Make it show its workings. Chain-of-thought prompting forces the model to reason step by step before answering - replacing assumptions with explicit intent. Example prompt inside 👇
https://t.co/T99MIafLUm
#PromptEngineering#AI#ClaudeCode