Happy Birthday to Me

Yesterday was my birthday! I'm old now. Actually, I think I crossed that bridge officially 2 years ago, but I'm haven't really felt it... today however just feels a little different.

#Edinburgh Fringe

Alanna and I just returned from a long weekend to Edinburgh for the annual "Fringe" festival. It was perhaps a little too jam packed with people but I'm glad that I went. I walked across the city center a few times over our short three days and got to see a slightly different view fo the city. We went to three different shows together but otherwise were free to do whatever we wanted. Alanna saw a lot more performances and I scoured the city center for a mostly quiet place to play on my laptop and absorb the atmosphere with a nice cup of coffee. I ultimately found Cult Coffee Roasters and was really pleased with the coffee assortment and quality. There are some awkward bench seats in the middle of the restaurant that are not great if you're just there for a few minutes, but for someone willing to put in the effort to squeeze in, it was a stable place to take in the hustle and bustle without much fuss from other patrons.

The headline show that we went to see was an improv troup called "Shoot from the Hip". I must admit, that kind of comedy is not really my thing but it was definitely cool to see a group of guys really dialed in to their craft and able to make basically anything at least a little entertaining. From a performance standpoint, really well executed and I am glad to have experienced it.

#Birthday

I decided way back in March when we were planning the Edinburgh trip to go ahead and take off the first half of the week that we returned and I'm so glad that I did. Flying back on Sunday turned into a nightmare of cancelled trains and stress so having Monday off work was a welcome change of pace. I slept in on Monday and for my birthday on Tuesday I woke up and did some light chores before heading off to my favorite café to drink some coffee and read a the 2026 release for the Dutch "Boekenweek". I just realized in researching for this blog post that Boekenweek was back in March, so I guess I got lucky that the bookstore I visited in late July still had some extras out on display!

After finishing the book and my coffee, I wandered back home and then got ready to head to my weekly taalcafé meetup. A friend I haven't seen in a while was there and we made plans to hang out tonight! After the language café, Alanna and I went out to a Greek restaurant for dinner and had a great time sharing a huuuuuuge platter of meat! 🍖 😋

I'm not sure if I'm going to make it an annual tradition, but I seem to always have too many vacation days left over so perhaps I will plan for more time off around my birthday moving forward.

#My Approach to Software Development & AI

A bit of a tangent, but while I was sipping coffee in the UK, I moved a couple more of my sites over to Laravel Cloud, with a lot of help from AI. I've been testing out the new "Oh My Pi" coding harness and overall I really like it! I still need to test it out with the work repository, but I'm hopeful!

The migrations themselves were pretty straightforward but I needed to iron out a few things. For one, SQLite is not really a first class citizen on Cloud yet so I needed to ensure that this blog could run on a Postgres database. The only database backed feature of the site is my annual step counter and its ingest route, so luckily it's not much.

I was impressed how the clanker (btw, what is your "pejorative" term for an LLM bot?) was able to SSH into my live Forge server (with my direction and permission), export the data and the prepare a migration plan and import it into the Laravel Cloud pg database. I was able to migrate the entire site in about an hour instead the entire evening or weekend that it might have taken me before. Time has largely been the reason why I haven't moved things over more quickly so this was also a nice excuse to test things out with a bit more of a "hand-off" approach and just sit back and supervise.

Steering this towards work and my career, I'm definitely starting to see how these LLM tools can become a big unlock for a software developer with limited time and resources. Compared to 2025, when I was basically not using these tools at all to now, it's pretty wild. I'm glad work is paying for tokens because if they weren't I'd be forking out $200/mo to use them myself. After commuting, social obligations, general adulting and most likely future parenting (not an announcement!) I can really see myself leaning on these tools to unlock productivity in a way that I was not really aware was possible before.

I'm definitely working more. There are so many little paper-cuts in our work codebase that I want to fix and using these tools has been awesome to go dig through the drudgery of piecing old shit together and making some sort of sense out of the spaghetti. Or if I need to work on a particularly painful feature or one I just find wholy uninteresting, I can throw the bot at it first, work on something I actually like and come back later to review and improve.

Which brings me to the next point, the review and improve. I can see and more importantly feel the temptation to ship slop at every turn. They make it so easy. The barrier to change code is now so low that I've definitely caught myself doing dumb shit just because I can and it works. See a class name that I don't like? Rather than rename it myself, ask the bot to do it. Those sorts of things. Absolutely a giant waste of money but my hands are already on the keyboard and I've hit Enter before I realize how stupid it is.

Going back to productivity though, I've managed to build a nice testing setup for our production application that will hopefully unlock the ability for everyone to be more confident about the changes they are shipping. We had ZERO PHP testing infrastructure for our customer facing webapp before and now we have Pest/PHPUnit and a couple of base test cases that can be used depending on the feature that needs to be tested. I was able to migrate our legacy Dockerfile spaghetti to a container that is pushed into GHCR and easily updatable by anyone on the team.

That also unlocked (btw fuck AI training and ruining some words for me) the ability to easily extend our test harnesses to run multiple PHP versions in CI as we work towards a PHP 8.4 upgrade. We've been stuck on 8.2 for a couple of years now and ideally we can get this upgrade live before 8.6 releases this fall.

For changes like this, maintenance and ancillary improvements, it's been almost impossible to get our product owners to buy into letting the team sprint time on them, but now I can fire off an agent on the side, and then submit the PR as part of Iron Man time without actually having to use my Iron Man time to fix things! I want to spend self-improvement time on learning new things, not sifting through a bajillion shitty Github logs looking for why this test is flaky or why CI takes so long to build.

#The future of software

One thing I do want to lament is that the future of software as it's been known for my entire career and likely decades before. In a professional setting, I think the utility fo these current LLM tools have proven their usefulness. Being able to do a lot of boring gruntwork and apply some sort of "reasoning" (whatever that is/means) has been recognized to be a major win for the industry as a whole. Ultimately, I think the floor for developing quality of software has dropped, while simultaenously being raised. We as developers now need to move our attention to detail from one area of the job to a totally new place. Communication skills have always been a marker of a good developer and I think that is even more true now. Being able to accurately describe what you want to happen can be the difference between hours of hallucinations and running in circles and firing off a one-shot prompt that passes peer review with little fanfare.

I think the enterprise will continue to adopt these tools and try to develop processes to offload the tedious busywork to an LLM and save their engineering talent for solving real business problems. Now, how much that will cost, what the tools are, the capabilities of the models; all of this is largely still unknown and I think will continue to be a moving target for at least 12 months. Further out than that is impossible to know at this point, imo.

I am curious to see what the future looks like for people just entering the industry. I was already in the door before the React bootcamp wave hit the industry but I'm curious if a similar wave will come crashing through but instead of React, they will be AI coding bootcamps that promise folks a future in software and all the associated riches. Before that happens, the cost of these tools needs to come down substantially first. I'm more optimistic about that and Asia seems to be pushing hard for more power efficient and cost efficient models. Hopefully we will see some better optimized hardware as well, similar to things like video decoding that eventually worked its way into CPUs directly and now I can watch Netflix on my laptop for 13 hours on a single charge.

What will "hacking on a project" for a teenager look like in 5-10 years? Hell, what does it look like today? Is a $20 sub to Codex/Claude enough for meaningful learning and side projects? I don't know. Will the focus for these students be on more abstract problem solving or more concrete problems in the real world? 3D printing might have shown us the way here and I'm just not aware of it. Being able to go from idea -> reality in both hardware and software is a really cool future though that I'm looking forward to. Well known 3D printer company, if you want a 5+ year ad-spot on my blog, reach out and we'll figure something out! 😉

#My future in software

I've always been a very product focused developer. Since my first days working at the scrappy midwest marketing agency, solving problems for customers has always been my job. Through that problem solving, I have been able to hone my craft as a software developer and work on some interesting problems. With LLMs, that isn't changing. I'm still getting pulled into meetings to discuss how we can best solve a problem and I'm always thinking about how our team can deliver a solution in an elegant way that is still valuable without creating more headaches for ourselves.

With LLMs able to deliver slop at the speed of thought, it's already apparent that humans need to stay engaged to protect the quality bar and ultimately be responsible when shit hits the fan.

I think there is an opportunity to try out some new software stacks as a whole in my future. I'd like to ship my own mobile app at some point, just for fun. Maybe a fasting app that I will actually use. 😅

If I can find an idea that I think will do good in the world, I'm also more keen to going out and starting my own thing. I know I have the chops to build something and at least scale it to profitability, but that thing exactly is still not been made known to me yet. I'm also still stuck in the legal limbo for starting a business here in the Netherlands, but that should be ironed out in a few more years so I'm not too worried about it.

Until I perhaps find my own thing, I'm also confident that I can keep delivering value to some company. I'm only receiving more responsibility at work; a good sign I'm not useless, and I'm also doing a bit more manager-y stuff. Not my favorite, but at least I can slap it on my CV for brownie points.

That said, I really still enjoy building things for people to interact with and use every day. Handcrafted UX and UI that works for customers instead of against them is where I truly derive joy from my work and I'm not-so-secretly happy that LLM frontend slop is so fucking awful that I think we'll need to write frontend code with intentionality for the forseeable future (12 months at least)!

A bit rambly here but it's been nice to get my thoughts out and into the world. Time to commit and push and walk back home in the drizzling rain. I really love living in the Netherlands 🙂 ❤️ 🇳🇱