It’s January. Time to start this again, as always with the intent that this will go on all year. But hey, at least this year I have a ✨ content ✨ plan!
This is gonna be a bit of a moment-in-time/2025 retro type post, sorry if that’s not your thing.
Claude and Claude Code are my daily drivers now. That’s a weird way to start a post, but truly I am using them every single day, literally more than my car. I have personal and work accounts for Claude and ChatGPT, which I mostly use as a foil to Claude.
I mostly don’t write code for work these days, but it’s my job to understand how all this AI stuff works anyway. That’s meant lots of toy projects with Claude Code — learning about agents and workflows that work, and remembering that git worktrees exist. I solve a lot more problems with code than I ever would have before, but especially since early 2025. I also get to be curious about a lot more things when I have a pile of data at my fingertips.
Of course, toy projects alone won’t do. One of the best parts of my job is that I get to have lots of conversations about AI with devs and others at Swarmia, as well as folks outside of Swarmia. The range is wide: everything from “our devs don’t write code anymore” to “yeah, we use Copilot” to “I don’t use it much” to “out of my cold dead hands.”
AI was a very abstract thing for me until late 2022, right around the time I left one job without another in mind.
Midjourney and DALL-E came out that summer; ChatGPT followed in November. I played with the image generation stuff. Those tools felt gimmicky and their output felt full of risk. I remember then getting a picture of a pit full of cherries when I asked for a cherry pit. ChatGPT was a fun toy during a few weeks of unemployment. I wouldn’t dream of using it for anything real.
When I started at Swarmia, I mostly used it for writing stuff, sparingly, rough drafts only, and had no dreams of leaving the browser. Ninety-five percent of the content I was writing — maybe more — was coming straight from my fingers.
When I was writing Build in the summer of 2023, I used ChatGPT to help me challenge and organize my ideas. It wrote some really bad prose that nonetheless helped me get past the blank-page problem.
Today, generative “AI” — we all do still know it’s just a predictive text engine, right? — it’s become genuinely useful for me in ways I didn’t expect. Claude has proven far and away the best at writing, for me (though I probably need to give GPT5 a fairer shake), and I finally feel like I’m gaining the skills to orchestrate these things. I even started, just recently, learning how to build my own models.
AI is helping people produce code faster. This is real, not hype. In the hands of experienced software engineers good at systems thinking and the concept of orchestration — these things are game-altering. Add real enablement and guardrails in the form of platforms and tools, and geez I can only imagine what you can do.
I hope it’s trite to say at this point, but the job of a software engineer is changing forever, even if we don’t know what it looks like at the end. I am ever more convinced that some portion of people who were employable in circa 2019 as software engineers are going to struggle to find their place in the new world.
We’re not in the Before, we’re not in the After either. It’s also not clear yet whether we might actually be in or on our way to The Bad Place. But here we are.
I still maintain that some of the biggest advantages of generative AI are going to accrue to people who don’t wear a software engineering hat today. Suddenly they, too, get to be a lot more curious about a lot more things, without bothering a dev. Insert shadow IT, security, governance, privacy, etc. concerns here. Meanwhile some dev teams race toward founder mode goals using AI subscriptions they bought themselves.
But don’t you dare take it away from anyone who’s hooked to it today. Claude Code broke the other day and I had to … use Claude Web to figure out what was going on. It’s rough out there.
Still: Something big is happening, and I’m solidly in the cautious optimist territory, though there will be winners and losers and we will start to see that shake out in early 2026. For companies that are trying to get massive software engineering benefits out of these tools, there will be a clear divide between haves and have nots, based on increasingly predictable criteria. TL;DR you probably shoulda been, like, writing tests.
At the same time, the AI industry itself is not like Enron — no really, definitely not like Enron at all.
I’ll talk more about some of that (but mostly not the Enron part) in January 2026, Part 2: A Rising Tide that Lifts Some Boats, which will be a more coherent and deeper dive of my thoughts on the state of … things.
This year I got to hug Nathen Harvey in person and receive my own personal printed copy of the 2025 DORA AI Report, control my inner fangirl when The Charity Majors came on the podcast, and hear Lada Kesseler calmly connect all the bits of gen AI wisdom that had been knocking around my head all year (podcast episode coming soon).
Off the record, I got to talk to senior leaders at tiny startups and well-known fintechs, companies in the post-hypergrowth stage trying to stay afloat, companies full of AI hopes cutting headcount before they started floating at all.
Their stories are all over the map, and I deeply believe some of them could make a good TV show. I still can’t believe who will agree to talk to me, it’s wild.
I wrote a whole lot of words, and got to work with Swarmia’s new content marketing person, which was a treat. We developed a lot of content together through interviews, candid feedback, attempted organization in Notion, drafts and drafts, and all that. Some favorites:
Measuring AI impact like it’s 1995 If you’ve known me from my jQuery days (it turns 20 this month, btw), you still might not be old enough to remember this.
What the 2025 DORA report tells us about AI readiness Yeah it’s an SEO title, sorry, but it’s full of goodness.
Code faster, ship … the same? My written reckoning of a reckoning I think is coming.
So here we are. In the famous words of Jed Bartlet: What’s next?
I don’t generally like announcing commitments — show don’t tell, right? — but: over the next year, I’ll be writing from a vantage point that feels underrepresented. Close enough to real organizations to see how decisions actually get made, but across enough of them to notice what keeps repeating — and as a subject matter expert on developer productivity as a bonus.
I’m interested in the parts of this transition that don’t announce themselves loudly. Where bottlenecks move after code gets faster. How responsibility shifts as tools get more powerful. Why AI tends to surface leadership constraints rather than fix them.
No frameworks, no prescriptions — just what I’m seeing, for anyone else trying to figure out what’s actually changing as 2026 gets underway.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.