#8351: Twenty twenty-five
Here we are: a new year, a new opportunity to look forwards and inwards.
Two things, I believe, characterised the previous twelve months for me: health, and purpose. One thing that has frequently come to mind over the previous years is, basically, what am I doing? I’d churned through multiple startups in barely a couple of years as part of an opportunistic move to get into the field of AI (more on that later) and each time found myself flaming out barely 8 months into the job. One time may be mistake, a second time suggests an error in judgment.
It’s no coincidence that, after moving on from my second foray into AI startup tech (where, upon reflection, I felt out of my depth), I switched gears and turned towards the concept of Tech for Good. It strikes me that so much of the work I can do as a software engineer feels unrewarding: here’s some AI that, fundamentally, spams cold outreach; here’s a new SaaS productivity suite which is a glorified CMS and CRM; here’s something marketplace style thing that actually does seem cool, but its crushed by the weight of its infrastructure and work is slow to finish. I dunno, maybe I’m just hard to please and I’m making things sound less interesting than they really are.
I count myself lucky that I found a company that offers meaningful positive impact that is also tangible, in the world of accessibility and public transport. The problem space here is one you’re unlikely to even think about unless you are disabled yourself or close to someone who is, and therefore require additional accessibility needs when taking a train somewhere. The challenge is two fold: ensuring such a project remains technically maintainable, and doing that while still serving the people who depend on the system. There is a sense of purpose here that I feel I can connect to without much difficulty, and so I’ve happened to feel more content in my career as a result.
And, perhaps, there might actually be a career to think of rather than several years of increasingly self-conscous job hopping.
That said, the early start was rough. I started 2025 with a diagnosis of hypertension (high blood pressure basically), and by the summer was laid low by a kidney stone of all things. Anyone who has had kidney stones know they are intensely uncomfortable and cause quite a lot of pain until they pass. It sort of works out better when you have two kidneys, because one can compensate for the other, but a CT scan and a diligent urologist discovered that I was missing one. Probably from birth, but who knows.
I’m not one to complain too much about medical staff doing the best they can while overstretched in an intense environment, but I did wonder how one could miss the absence of a kidney in a CT scan. I’m fortunate that it was double checked after my blood test results became increasingly concerning and there were indications of an infection.
On the second day of hospital I went in for surgery (quite suddenly by all accounts), and found myself laid up in hospital for a good week. Honestly, a hospital is not the most pleasant of places at any time, and getting woken up every 2 hours for observations gets old fast, but I seriously appreciated being able to read the books my friends brought me with absolutely no distraction. It was life on easy mode, for a short while.
The surgery that week may have been the first but was not the last, as I was scheduled for a follow up to remove the stent that was inserted. This too was infected and rather than removing the stent, the urologist opted to replace it, scheduling a third follow up to remove that one. Low mobility, lots of blood, lots of pain; I’m grateful of all the support I was given as I was struggling.
All in all a huge ordeal that consumed my summer from mid-July to mid-October and even now things feel different, although that may be psychological.
Just before all of that happened I’d fallen into the hype of the weight-loss jab and I’d given Mounjaro a shot. Of course, had I known I only had one kidney at the time I might have reconsidered that. Worried that it might have been connected, I cut that experiment short - gonna have to get back into shape the hard way: back to the gym (or at least being more active), back to healthy eating…you know the drill.
In the midst of all of this, a decision at work was made to adopt AI in our workflow as much as possible. I spent a decent amount of time in the latter half of the year embedding myself further in the world of AI, perhaps revealing how much I might have dropped the ball in my previous jobs in terms of enthusiastic early adoption.
My personal position on AI, or LLMs really, remains nuanced. I am dogmatic neither for or against it, and at best it’s an effective tool for me as a developer to handle busywork while my mind stays focussed on things requiring deeper thought.
An LLM like Claude or GPT is not in itself intelligent; it acquires that from the person writing the prompt. The more specifics, the more knowledge imparted, the better. This changes an LLM interaction (or an agentic one, say, in the context of Claude Code and the like) from a dice roll to something akin to an automated delegation of work. These things aren’t mind readers and Garbage In, Garbage Out remains constant even if the garbage you get is convincing.
I’ve stated on LinkedIn more than once that I’ve used AI as an opportunity to improve the way I explain concepts, because I can then apply that in a mentoring context. This sounded sensible at the time but still, it’s easier said than done. Practicing this kind of communication against a tool that will never contradict you or present you a challenge means you only experience a happy path of sorts, and experience demands you understand more than the most simple and direct path to a positive conclusion.
These days my enthusiasm is tempered and I use AI to help me do things that I’ve either always struggled with or simply don’t enjoy. Diagrams are not my strong suit, but I can describe a system in prose and AI can turn that into a decent Mermaid diagram for me. An AI can craft a complex, deeply normalised SQL query in half the time it would take me to write it all by hand, because I understand the schema and can reason about it in prose. It can flesh out a bunch of jira tickets and help me review them.
I don’t really know what the future looks like right now, for tech or for me personally. Perhaps one of the more unexpected conversations floated the possibility of management.
I’ve always maintained that I find technology easier to handle than people, and I suck with people, and so I’ve always favoured the IC (Independent Contributor) track, but you know what? That’s limited thinking and I fucking hate limited thinking. Maybe it’s time I get back out of my comfort zone and learn how to do something else.
I suppose that sets the scene for the year to come and, honestly, I’m here for it.