What is the most impactful and challenging work a software engineer can do in a fast-paced organization?

Not a surprising answer, but it’s not a new API endpoint, a new UI widget, or a prototype for calling an LLM API.

It’s the code that makes all of the above blend and work together seamlessly in production, it’s the opinionated APIs that guide you to be productive, the good set of defaults for retries and error handling that you don’t have to constantly think about, it’s the complex OpenTelemetry collector that lets you write one line to export your metrics, it’s the piece of infrastructure code that automatically deploys your last commit to production.

We see a lot of posts these days saying that AI will make software development obsolete, and yes, it helps us be much more productive when writing code within fixed guardrails, but I’ve yet to see AI produce comprehensive glue work and processes.

At large companies, glue is often created and developed by internal platform teams that integrate with each other to enable feature teams to deliver to their end customers at incredible speed and quality. At startups, it’s often a handful of seasoned software engineers who do the glue work. I was chatting with a brilliant friend over the weekend who often takes on this role, and while it can feel ungrateful at times, it can also be incredibly rewarding to see things run smoothly at scale thanks to what you put in place.

The next time you’re implementing a productivity improvement or integrating something new into production end-to-end, remember that you’re gluing, and don’t forget to celebrate.