Community Days 2026 was a different kind of conference experience for me. No booth to staff, no talk to deliver, no social media queue to manage. Just me, a badge, and a schedule full of sessions I actually wanted to attend. Turns out that is a pretty good way to spend a couple of days.
I did some networking, did some job hunting, ate some ice cream, and came away with a few things worth writing down.
How to Upskill Within Your Skillet
A panel with Angela Jones, Mark Williams, and Lindsay Shelton covering career development in a world where linear career paths are mostly a myth now. The framing I found most useful was the distinction between expansion and pivot. Expansion builds on what you already know and adds to it. A pivot is more of a reset, starting over in a new direction. Neither is wrong, but they require completely different strategies and timelines, and conflating them is where a lot of mid-career frustration comes from.
The conversation about what to learn next was equally practical. The panel drew a line between hope and signal. Hope is learning something because it feels exciting or everyone is talking about it. Signal is what actually shows up in job postings. Both matter, but leaning too hard on hope without checking the signal is how you end up chasing trends that evaporate before you can capitalize on them.
The question they closed with landed well: what would you tell your five-years-ago self? The answers were pretty consistent across the panel. Stay adaptable. Ask better questions. Build trust. Develop translation skills, meaning the ability to move between technical and non-technical conversations. And get comfortable being a beginner again, because that is going to keep happening.
Future of Code Reviews
A short but pointed session on where AI is taking the code review process. The framing that stuck: we are no longer the code reviewer. We are the Editor-in-Chief. The question is not whether humans are still needed but what that human role actually looks like when the first pass is automated.
Saying to Thinking to Doing
The most technically dense session of the conference, covering the evolution from LLMs to LRMs to LAMs and what that progression actually means in practice.
LLMs (Large Language Models) are the current baseline. One token, one pass, fast and cheap, with a cost that scales roughly linearly with tokens. LRMs (Large Reasoning Models) take a question and run a reasoning trace before answering, which means they are slower and more expensive but produce better answers to complex problems. LAMs (Large Action Models) are the next step: high-level tasks come in, real workflows get automated end-to-end. The tradeoff is that errors can compound, and the cost jumps significantly.
The throughline was simple: there will always be another tool. The teams that do well are the ones who understand the tradeoffs well enough to reach for the right one.
It was a good conference to just attend for once. More of this, please.