a bit of internet | human-made unless stated otherwise

I walked the dog. So now there are some thoughts about AI I'd quite like to get out of my head and into blog posts if I get the chance. The backlog grows less viable by the week, I realise. Noting them here.

The AI handler

I wonder if this will become a thing. A role. A skillset. Pulling together some of the things below. People might specialise in one more than others.

There's a lot of chatter about this already, I know. But it strikes me a lot of the focus is on outputs. I think inputs are every bit as interesting. Can you protect what you put in as well as what you get out?

Ethical principles

Might be fun to come up with a set of ethical principles for making things with AI, if no one has already. Feels like a collaborative thing, but I might have a first go. Sounds like something ChatGPT can help with.

Creative principles

I'm less confident there's a need for this, but I'll probably mull it all the same.

Near-term futurism

A future-tense thing to write about predictions of how all thise stuff is going to pan out. I had this half-written on the dog-walk. I should probably do this first in the hopes I can remember some of it.

#ai #notes #humanwritten

The AI creativity tipping point.

Inevitably, conversations are happening on work Slack instances the world over about the potential use of ChatGPT. Its interesting that other tools don't seem to be anywhere near as visible, despite being very powerful. I've just added my tuppence-worth to a conversation. I thought I'd copy and paste it here:

Really interesting topic. I've been experimenting with ChatGPT, and similar tools that can help create slide decks and images. We definitely seem to have crossed a viability threshold in recent years and it's interesting to see how quickly these tools are now becoming widely available. I think it's this big bang availability as much as the power of the tools themselves that will have massive societal impact.

I studied AI at uni back in the dark ages – things were very different. And yet in some ways the same. I think better understanding of the limits that still exist is needed to avoid risks like unintended bias, inaccuracy or plagiarism. The outputs I've seen from ChatGPT haven't changed my mind that AI-created work is necessarily derivative by virtue of the way neural nets work – sometimes in the extreme. I think it's still true to say these tools are better at very narrow, specifically-defined problems than open-ended one. “Make this sentence shorter.” “Back this statement up with a statistic.” Rather than “Write me a sonnet in the style of Browning.” I asked ChatGPT to “compose” a zen koan, but it spat out a verbatim form of the best-known koan in its most common English translation. I'm probably more excited by AI-generated images than text, but that's probably because it fills a bigger hole in my skillset.

The tools are clearly good enough that it would be daft to discount them or look the other way. I think the key to using them well, and ethically, will be in careful control and scrutiny of inputs and outputs. One thing that makes me a little uncomfortable about ChatGPT is, to my understanding, you can't train its inputs other than in the brief you give it. Some of the other tools out there, you have more apparent control of inputs – at least you seem to. Who knows how much pre-training has been needed, and on what data?

Sense-checking, fact-checking, plagiarism-checking are skills I think will be needed around using AI. And for a while to come, I suspect “AI handling” will become a recognised skill. It's going to have an impact.

I came across this Waxy post this morning: Invasive Diffusion: How one unwilling illustrator found herself turned into an AI model.

And I posted this to Mastodon after:

I'm late to that last link, but it's a convincer for me. I'm not changing my mind – it's just pushed me over the line. I still think AI-generated work is necessarily derivative, but there's a world of usefulness within that if inputs and outputs are handled ethically and carefully. First step: updating bios and project descriptions to be clear about the mix of human and AI input and output. At the moment, everything pretty much 100% human, except 1 blog post and 1 now-unarchived Instagram post.

Note: Derivative stock-style image AI-generated with

#ai #creativity #humanwritten #notes

On the subject of CSS, I've been wanting to make a few tweaks here, inspired by the old High Res Tumblr theme.

a {
  color: #000000;
  text-decoration: underline;
  background-color: #b3f9fc;

html *
   font-family: menlo,'andale mono','courier new',sans-serif;

#notes #css

It's been decades since I've properly tinkered with CSS, but interesting, lovely, useful seems like a good opportunity to dip a toe back in.

Thought I'd make some notes on what I'd like to do without yet worrying about what's actually possible.

That said, there'll likely be some blockers to me doing some or even any of this. Things like:

  • my lack of knowledge and ability
  • lack of time, either to do or learn
  • limitations of what CSS can actually do
  • limitations to what can be done with

There: I've built myself a safety net to justify doing little or nothing.

But the thing that seems to scream out for CSS is the ability to conditionally style posts depending on how they're tagged.

The obvious thing to me is to apply a different colour for each tag interesting, lovely or useful.

Things get exciting, though, because some posts have two tags. Some have all three.

To me, that suggests using primary colours – or colours that blend well, for individual tags, then using their blends for when two tags are applied.

So for example, if you were to apply the three primary colours (so far as physics has it, anyway):

  • interesting – red
  • lovely – blue
  • useful – green

So for posts with more than one tag, you might blend those colours:

  • interesting, lovely – purple
  • lovely, useful – turquoise
  • interesting, useful – orange

Or simply show them both.

And what do you do when you have all three? Additively mix all primary colours and you end up with white, which doesn't sound right. I immediately jump to rainbows and Sinclair-style bands, but something “all the colours” sounds about right.

I probably won't do any of this, but there's a small chance I might investigate further before not doing any thing so that the not doing is a better-informed choice.

Update: after some early reading around, sounds like some javascript might be possible, but I don't think that necessarily rules it out on a blog.

#notes #css

I fixed the follow links for so they should work now. When setting up the blog I forgot there were some back-of-house switches to flip to turn things on. Sorry for the hassle!

#notes #created

Yesterday's blog post, about making your computer a lovely place to work, was I bit rushed. I knew that as I was writing it, in an all-too-short snatched few minutes. When I published it, I felt… dissatisfied.

When I wrote about Jet Set Willy a few weeks ago, I was surprised by how much I enjoyed doing it.

I've come to see writing as something I should do because I sort of can. But I've bought into that narrative, described by that supposed Thomas Mann quote that some writers are fond of, that says: “A writer is a person for whom writing is more difficult than it is for other people.”

There's still truth to this. Writing is hard for anyone. And anyone that finds it too easy probably isn't doing it right.

But what I liked about writing about Jet Set Willy was that I gave myself all the time in the world. If there was something I wasn't sure of, I looked it up. If a thought occurred to me, I gave it time and space to develop. I gave myself permission to capture everything. I gave myself permission to explore layers of parenthetical thought. I gave myself permission to break all the rules of what I think makes good writing, because I was writing purely for entertainment – mainly my own, but in the hopes that a glimmer of that would shine through.

The obvious conclusion to draw is, if that's the kind of writing I enjoy, I should do more of it. I should carve out big pockets of time to dig into a topic I love, or am temporarily fascinated by. And that's definitely true. But what about those midweek pockets of time? I want to blog frequently because it's good my brain, but if I only do longform, it won't happen frequently enough to really be blogging in the sense that I think of it.

I think the answer might be to tackle a tiny idea, but in the same detail. Have a thought, but rather than hurry to get it down so the next thought can take its place, to spend time with it. Hold it up to the light. Turn it this way and that. Bite down on it with the back teeth like it's some unearthed booty. And basically have fun with it.

Maybe that means dwelling on a topic for a week or two, exploring it one thought at a time. And maybe that means, if I'm not in the mood for thinking, writing a list. I've done some list posts already – they're useful and fun. There are lots of options. That's the nice thing about blogging – after all these years (decades!) doing it, I'm still learning about. And as Giles says – I think in his excellent agile comms handbook – but probably elsewhere too, there are no rules.

But I'm going to try to set myself one: don't rush a big topic in one post. Or even a medium-sized one. Or if I do, think about coming back to it and looking at it properly, a piece at a time.

#notes #blogging

I've been taking active steps to make my home-work environment a bit better. If you're lucky enough to work full-time, that's too much time for that environment not to be nice. At home, that's meant sorting my desk out a bit. I'd gotten into some lazy habits unplugging my computer and working away from my desk. So I've sorted a second screen out, cleared away the unnecessaries and installed a nice lamp.

It's a cycle I go through once every couple of years: sorting my desk out. But the bit I've come to miss is sorting my computer out. And not just my computer, but my bits of internet. Because I've come to realise, because apparently I forgot for a long time, that these have a huge bearing on my mental wellbeing at work. My computer too needs to be a nice place to work.

One of the things this has included is a Spotify re-subscribe and a good clear-out. I've moved almost all of my old playlists into an archive folder. And I make myself listen to music. It really helps me to focus by shutting out external noise. When I'm trying to concentrate on a particular hard problem and there's a bit too much external noise, I get stressed. So it's better to fill my ears with sound that doesn't distract but isn't boring at the same time. (Today that happens to be an old favourite: Surviving Mars Official Mars Radio.) When I ask myself what to do next, if I'm not listening to music, I make it that.

Another thing I've done is to get involved in a few online communities. The fediverse has been great, as has getting blogging again. But I've also started to get involved in a few music communities again. Nothing too distracting, just a lovely thing to check in in those 5-minute intervals between bits of work.

Or I might spend a few things fiddling with some admin to make my life incrementally a bit better. That might be leaving some Slack channels, closing an online account I no longer need, tweaking my desktop/operating system or browser settings, or changing my desktop wallpaper.

These things, and others, go a long way towards making my working day not just bearable, but actually pleasant. And they beat checking a random work Slack channel and open a can of worry-worms.

#notes #work

More and more I'm seeing articles and blog posts online about Apple's increasingly suspect privacy practices – this post about Apple scanning local image files without permission being just the latest example.

This has got me thinking once again my digital life admin, which has been in much need of an overhaul for some time. But I've realised that, more than being a to-do list of digital tidying up, it's probably useful to have some guiding principles to hang them on, and to help me make good decisions.

So I'm typing this to download the swirling thoughts I've had to hopefully find some order from the chaos.


  • Choose free and open source software when there's a good option
  • Avoid ad-supported services
  • Avoid privacy-compromising/data-scraping services
  • Choose local rather than online
  • Choose self-hosted over cloud services
  • Make email the main point of contact for people you know
  • Make social media the main point of contact for people you don't know


  • Close online accounts no longer needed
  • Switch all 2FA to app-based rather than SMS authentication
  • Bifurcate email between logins and human conversation
  • Investigate/experiment with the viability of going without a phone
  • Investigate an email do-over, re-investigate the best options
  • Buy an external drive
  • Download your cloud backups and get off cloud storage services

There's the “wherever possible” caveat for almost all of these, mainly because some services are thrust upon me by work, which I imagine is true for many.

I'm thinking strongly of switching to local backups, but in a low-effort kind of way. I've too much old stuff to sort through, but hopefully this is of a reasonable enough size to store and forget. Then, going forward, keep only what I might need, backing up intermittently onto old drives – cumulative dumps rather than actively managing files. I'm banking on the falling cost of external storage making it viable to jut buy another drive once the last one is full.

That seems to be all that's fallen out for now. Next the hard bit: actually doing it. (Sorry about the word dumps.)

#notes #technology

I've ressurected a couple of Spotify playlists of old, I thought I'd share. The first is anticontent radio.

This is a continually-updated playlist limited to the 10 most recent tracks I've listened to and though “ooh, that's good – and in a way I think other people might like”. That sounds weird when I type it out like that. At the moment I'm leaning heavily on old favourites and Spotify recommendations, which means there's quite a lot of oldish electronic music. But that's no bad thing: there's plenty of excellent oldish electronic music out there. But I want to get into the habit of researching and listening to new stuff – and catch up on the last couple of decades.

The other is anticontent collected, which is the same as anticontent radio, except it's the collective history rather than the most-recent 10.

I may as well share a couple of other public Spotify playlists I made once upon a time:

accessible aphex is fairly self-explanatory – a collection of Aphex Twin tracks which are both lovely and easy to listen to. Occurs to me this includes only things he recorded under that name. I'd quite like to trawl Spoffy for his alteregos and buld this out.

boom bap n all dat is a collection of stand-out tracks from stand-out albums from the early-90s golden age of hip hop. I got out of the habit of listening to hip hop and rap some years ago, and for no good reason. I'd like to fix that.

What else? Oh yeah – my ongoing project to upload my entire back catalogue of naive hobbyist music-making onto MeFi music has passed the 20 track milestone. I've been adding these to a playlist. Pleasantly surprised, listening back, to find a few that weren't quite as terrible as I'd feared, like 130 stories high, get a cheerleader girlfriend, friendly and lycaenidae.

I noticed Baroque by Susumu Yokota got a rerelease fairly recently, which was a lovely excuse for a re-listen. Though I'm not entirely sure I did catch it the first time round, to be honest. Which is surprising as Sakura remains one of my very favourite records of, like, all time ever. Great to hear a nice lot of Steve Reich among Baroque's sample sources.

I also still have Plaid's Feorm Falorx on heavy rotation. I'm still getting to grips with it because it pushes the bounds on what I consider to be Plaid's signature sound. That would be a good thing in most people's eyes (or ears), but they're one of my very favourite bands for a reason, damn it. Cwtchr, Nightcrawler and Tomason are immediately stand-out bangers, though.

#notes #music

The other day I hopped on a video call with the various heads-of capabilities we have at work. Your heads of cloud, data, delivery – those sorts of things.

I work in the marketing team, and this was a call about how we sell ourselves in what we do. What makes us different. It's exactly the sort of thing marketing teams like to talk about.

One of the things we talk about a lot is being agile. It was before I got there, but we have a book about building agile teams. Side note: it's very engineery – one thing I spent a fair bit of time on in 2022 was editing a more multi-disciplinary (or, rather, non-disciplinary) second edition, which I'll probably talk more about at some point.

Anyway – agile's something we talk a lot about, and with good reason. We work in agile ways, and it's probably the main thing that sets us apart – not from other smaller agencies that also do agile really well, but from the big IT monoliths that build big, crap, expensive things that cost gajillions of pounds and take so long to deliver they're obsolete from day 1. And probably lock you into a decades-long contract to “maintain” (which probably means, at most, keeping most of the lights on). And which no one understands how they work – even the supplier.

(The bit that sets us apart, historically, from other smaller agencies is probably the strength of the big software engineering bit here, but there are some interesting irons in the fire.)

So agile comes up and I see eyes rolling. That's interesting, I think. Why's that happening? And the reason is that, apparently, clients are becoming jaded about the word agile. Everyone says they're agile, but not everyone means it. Maybe we should say lean instead, someone suggests. But the problem there is people are starting to feel the same way about that word. Dejected nodding. The resignation wasn't about doing agile, you understand – just the value of talking about it.

But hang on: are we accepting the premise that it's OK to say you're agile and not be? Or should we hold people to account? Or at least encourage our public sector clients to? And say we can step in and help when it's needed.

Because I'm prepared to accept that most companies – even some of the biggest ones – aren't perpetrating a deliberate deception. I think it's more likely, in most cases, that they just don't know what they're talking about. I've heard tell of organisations (though this was a local authority, to be fair) use agile to mean hot-desking. And I'm betting it's quite often used vaguely, or aspirationally. “We work hard. Sounds right.”

True, agile can encompass lots of different things and different ways of working. But it has an irreducible core of solving a small problem at a time with small pieces of work that make a meaningful difference quickly. If that's not happening, it's not agile. And if that's not happening, the people that say they do agile but don't should be given help to see the error of their ways. And, if that doesn't do the trick, not given contract extensions or new wins.

The other bit here is that, and of course I'd say this, words matter. And is varied as the details of agile can look, the word itself is too valuable in describing that core of “early and continuous delivery”, as the agile principles have it.

Because if we let agile go, lean will go the same way. And so will whatever we say after that – waffer-thin, maybe. And then we're nowhere, because none of the words we say anyway may as well meaning anything. Which is probably how meaningless business-speak got to be a thing in the first place, now that I think about it.

So for my part: no. I'm going to keep saying agile. And saying that we mean it seriously. And, wherever possible, showing that we mean it by showing that we do it. Though as I said the other day, this is much harder than I'd like it to be at a consultancy that works with the public sector.

Anyway – there. A hastily-written ramble about agile. Because the internet sorely needed another one of those.

#notes #agile

Enter your email to subscribe to updates.