I thought LLMs were just Copilot.
This week I discovered agents — and it changed how I code.
Instead of autocomplete, I had Claude Sonnet 4 troubleshooting bugs with me in real time. I redirected it when it drifted, or followed when it made sense. At one point, I asked it to output all its decisions into Markdown — suddenly I had a log of learnings I could share with the team.
A coworker spun up an Model Context Protocol (MCP server), so we all could mine knowledge about our project through hundreds of Confluence pages. Suddenly I wasn’t a solo engineer anymore — I was a team of 1 human + 2 agents.
The result? Faster debugging, better knowledge capture, and new side project ideas. I’m thinking, how could I implement a MCP server to mine Google Drive folders? Or, OneDrive? And hook them into apps or blogging platforms like this one WordPress?
So many possibilities! I’m excited! Hoohaa.
