One of the core tenets of a good working engineering team is to limit the Work in Progress (WIP). This has been written about so many times, I could spend an entire article just documenting all the people who say this is the key to efficient flow, so lets assume that’s true. I was reading this article and was reminded of that once again. One of the tenets of this article is to create small, reversible changes.

I’ve been working on an MCP server for my service atlas api. My original idea was to do this in Golang, but I gave up on that path. I wanted to talk a bit more about the why and the alternative in this post. Go vs Python? Golang offers a very robust library for MCP server development. A library I really, really wanted to take advantage of. But as I got past HelloWorld, I found a major problem: the library was really clunky to use.

So lately I found a new pattern to increase my productivity. I work on APIs pretty extensively, and have adopted using Bruno to test with and store my test calls in my repo with the api. Since Bruno calls are basic text files, I found I can do the following: use a swagger page to collect curls for all the calls I make for setup and testing feed those calls into copilot, telling it to use them to set up the pre and post script runs then tell copilot what specifically to test for and have it create a bruno file for that call (or update one that has the bare bones for the call I’m making) additionally, I will tell copilot to update the docs section with how to manually run the test in case QA wants to review manually I’ve found that this process greatly speeds up how I test and has completely replaced integration tests in some instances.

I subscribe to the JetBrains all IDEs package. Recently, I learned that JetBrains released a new agentic AI named Junie via this video (which is strangely unlisted now). I thought I would give it a try since I’m technically already paying for it and I have to say I’m quite impressed. I installed it in GoLand to help me on a project I’m learning to use Go on and overall it seems much better than other AIs I’ve used.

I didn’t get a chance to work on another Staff+ engineer post this week, so I thought I would share how I’ve been using generative AIs and my experience with them for this week. Recently I’ve had the chance to compare Amazon Q and Copilot. Both seem pretty good, but I’ve found that Q has different ideas of what is a security concern over Copilot. Both do a pretty decent job with overall for the simpler things I ask.