AI Adventures

I’ve been using Gemini to analyse my code this fine evening. It’s mostly Ant with some XSLT thrown in. The AI’s very good at resolving my Ant XML properties and will resolve those properties in a millisecond, something that embitters me because at XML Prague just a week ago, I spent 10 minutes talking about how to resolve those. It does a great job. I was wondering about an Ant script that I wrote, expecting to find a nested properties error.

But it correctly identified my well-formed-only mode, set via a Boolean property, and pointed out my mistake. That property was set elsewhere, which is why I didn’t spot it right away.

So I’m left wondering if my AI aid actually helped me or made me more stupid. I missed that Boolean property. I might have spent up to 30 mins without that hint.

For those ethically inclined among you, those 15 or 30 minutes could make a difference. Did a dozen or so human readers in Africa prepping AI responses just have their lunch denied? Did I kill a baby seal or deny Somalian kids their daily water quote? A mean thing to say, I know, but my question is real. Did that AI make a difference?

Plenty of shitty reasoning to excuse myself, but I don’t want to go there.

The real question is is it justifiable in any way to have me fix a bug because of the 30-minute win I suspect it gave me?

Creative Content at Creative Words

My son Jonas is doing a four-month internship for my company as a digital content producer. If you’ve spotted actual news and notices on the company’s LinkedIn or Facebook pages, that’s him. So is the new logo, by the way.

This will give him experience and add to his portfolio, and hopefully others will notice and hire him.

I’ll Present at XML Prague 2026

I’ll be talking about visualising Ant build scripts in Prague next month. Really happy about this one. Prague, of course, is one of my favourite cities in the world, and XML Prague is a conference I love to return to, year after year. Well organised, wonderful people, great talks. It’s a gathering of friends, old and new, and so much fun. I can’t wait!

Join us!

Done Reviewing

I just reviewed my last ever Balisage paper, and I feel very sad. I’ve reviewed papers for Balisage for a long time, and I’ve learnt a lot. Frankenstein visualised? Check. Everything is actually XML? Check. Markdown in SGML? Check.

Etc.

I’ve written my share of papers for Balisage, of course, but those were always for others to review and the audience to judge.

It’s all coming to an end now.

The Last-Ever Balisage

My friends at Balisage announced that this year’s Balisage conference will be the last ever. They’ll call it a day and declare victory, and I’m sad.

Balisage has been part of my life for a long time now. I’ve presented at every Balisage conference since 2012, trying every year to come up with something that would ensure me going, because it was never a given with my employers or clients. One year, I remember being worried enough the reviewers might not accept my paper, so I submitted two; this was incredibly stupid and hurt both papers, not to mention my audience. They always noticed.

I’ve reviewed papers for about as long, the vast number of the papers eminently enjoyable and fun to read. I rejected a select few and was proven wrong more than once, but I always learnt a lot. I don’t think anything has ever shaped me quite as much as Balisage.

Come August it’ll all happen just one more time. I loved travelling to first Montréal, CA, and later Bethesda and Rockville, MD, USA, to meet everyone for my markup holiday, as I would call these things. I’ve never ever been talkative so I would always listen more than talk, but the good folks at Balisage would become my friends, and I cannot even begin to describe how much these yearly travels meant to me. How welcome I felt; how much I felt being part of something much larger than me.

Covid put a stop to all that, of course, and Balisage became an online thing. It’s been great but not quite the same, much of it because when you present online you only ever face your monitor; you can’t play to your audience. I miss being there terribly. I miss the personal touch; you can ask questions and initiate a conversation, but you can never casually bump into a speaker by the coffee and tell them how much you enjoyed their presentation. not if you are like me.

You can also never just hang out playing Werewolf whilst ganging up on some poor sob happening to wear a red shirt that day. It was as if you had found home away from home.

The last ever Balisage. I’m sad today.

Oxford

I taught at Oxford.

It’s very, very cool just be able to say that. I did, though, last month (September). XML Summer School and Adam Retter wanted proposals for their 25th anniversary, so I suggested something about migration strategies, about converting something to something else. It was really all about my experiences with converting content in format X to some XML vocabulary and what you might want to think about when undertaking such an endeavour.

They accepted, and I was (and remain) thrilled.

I love teaching. It’s about sharing and caring, literally. It’s about sharing knowledge, about expressing ideas and getting a response, about questions and comments and discussions, and about planting ideas and concepts in listeners and participants. You feel alive and alert and thinking.

If you are lucky and privileged enough to do it in a setting such as St Edmund Hall in Oxford, it’s supercharging it, an experience to be treasured and missed. It’s as if the college and the town were both designed around curiosity and learning and celebrating knowledge. Maybe, probably, they are.

Balisage 2025

My paper on my little CMS (details won’t be available for a while yet) project has been accepted at Balisage. It feels a bit like cheating, to be honest, because that paper is simply a development of any number of papers and ideas before it, from versioning to S2000M. I’ve been thinking about content management fora very long time. My CMS — called JACMS (Just another CMS) for now — is intended to provide a simpler alternative than pretty much anything I’ve done for customers in the past. But JACMS is also very much what I think a good CMS should be, in terms of how we treat content.

And before you ask, no, a “CMS” to me is not WordPress or Drupal.

So, we’ll see.

Pointy Brackets Work and LinkedIn

As in, do you have any? And if you do, please contact me directly.

I found myself out of a long-term gig, recently. The customer did away with all of their contractors. This one felt almost personal to me because I’d been on it for a long time, relatively speaking, and I had a role far removed from a simple hired hand. But shit happens. I was in fact a hired hand. Tougher times.

So I’ve been looking, lately, I went on LinkedIn, where I pay for Premium membership which supposedly gives me an edge. I honestly don’t think it does, because their AI algorithms border on painful and so the top jobs you are suggested appear to be basic string matches. Not the relevant strings, mind, but your job titles. If your resume says you are a CEO (because I am; I’ve run my business for 30 years now), your top jobs will be CEO work. Which is ridiculous. My career is about XML technologies. I do XML, XSLT, XQuery, XProc, not to mention XML vocabularies like DITA, S1000D, and DocBook. And, of course, SGML, the ISO standard that led to XML.

None of the CEO jobs had anything to do with markup technology.

I contacted LinkedIn support. They said it’s all about refining your searches and filtering so the AI can learn, but since that’s what I’d already been doing, I thanked them while remaining doubtful. That very same day, my top job matches (the ones thinking I am a CEO so I must be looking for CEO work) magically disappeared.

Coincidence? I’m sure it was.

The other job suggestions I get are “architect” jobs. Again, this borders on the simple-minded. My CV lists a few job titles stating I am a “content architect”, sure, but unless you are very, very literal about it, you won’t haul me in for a rebuild for a property in Manchester. Right? In a string match, I’d sort of get that. But I’ve been suggested any number of “architect” roles where the only common denominator is that string. “Architect”. Integration, enterprise, more buildings, etc. They focus on “architect” but are not able to read the context, not even to match keywords that are actually there (and, for that matter, exclude those that are not).

It’s more of the same if you do text-based search yourself. Looking for “SGML XML” results in anything but. “XML XSLT” rarely gets anything useful. Etc. None of the buzzwords in my field yields a meaningful result, just more of the same nonsense.

Five years ago, though, adding keywords is how I found the long-term gig that eventually prompted this post.

Now, I get nothing of the sort. I cannot do a text-based search. I have no way to look for jobs that match my expertise. The filters offered by LinkedIn are geared towards jobs and roles they know about. I wanted to say “understand” but I think that’s stretching it.

LinkedIn is filled to the brim with AI these days. They base much of their business on it. Everyone is posting about it. Action figures, images, fake posts, tweaked articles, vibe programming, you name it. It’s everywhere.

But if this is the extent of its usefulness, count me out.