0.0 Context setting

Thursday 29 September in Portland, Oregon, where it is dreary but also not raining. Dreary kind of implies rain? Some sort of pregnantly dreary, I suppose. Gravid with rain but not raining yet. Anticipatory.

It’s been just over a week since the last episode and I’ve got to say that every day that I haven’t written I’ve gotten more anxious about not having written. This is not generally speaking helpful because I don’t know about you, but it’s easier to write I’m not anxious. Easier to do anything, really.

I can’t remember if I mentioned, but I got invited to speak at Throughline 2027, Active Voice’s online conference. Here’s some stuff about it:

Join me February 3-5 for two days of talks, one day of workshops, and a whole season of special events centered on this year’s theme: ambiguity as opportunity. I’ll be talking about how to stay intentionally human—and humane—in the time of robots. Early-bird tickets are now on sale through November 8. Get all the details at throughlineconf.com/2027.

I mean it says I’m going to be talking about how to stay intentionally human--and humane--but I’m in the middle of tweaking the talk and also trying not to be anxious about that either. The organizers gave me this feedback about my submission:

Is there a weirder, spicier or more personal place you would go? Someplace thinkier, or that takes a long view and explores some of how we got here as a field? That maybe even ruffles some feathers in a productive way?

... which a) thanks, intimidating, b) thanks, flattered, and c) now having thoughts, thanks.

So that’s another Google Doc that’s staring me in the face.

0.1 How People Work, Live!

Last week I had Michele Tepper on and we exchanged some Strong Opinions, which are always fun, and then for me the interesting bit was where we landed on at the end where I think I feel a bit optimistic about how non-developers/product/designers/engineers are vibe-coding things in organizations?

Michele’s episode and all the other ones are up on the podcast and YouTube now; you can catch up on anywhere RSS works with your podcast app, Apple Podcasts, Spotify, and YouTube.

Next up!

Friday 2 October at 9am PT: How People Work, Live! ... with Hillary Hartley

Hillary Hartley has spent nearly three decades helping government be simpler, faster, and better for the people who rely on it. She started out working with 30 state governments at NIC Inc., then went inside as a Presidential Innovation Fellow and co-founder of 18F, the first digital services team in the U.S. government. She later became Ontario's first Chief Digital and Data Officer, where she built the Ontario Digital Service and wrote the Simpler, Faster, Better Services Act. She was also CEO of U.S. Digital Response, a nonpartisan nonprofit helping state and local governments strengthen their tech capacity. She founded OpenPublic to coach public sector leaders and make transformation stick, and she co-chairs the FWD50 conference. She lives in Toronto with her family, their dog, and a veritable zoo of stuffed animals.

And then after that we’ve got:

• Friday 9 October: HPW, Live! ... with Jim Ray, currently a founding member of developer relations at Bluesky
• Friday 16 October: HPW, Live! ... with Farrah Bostic, who runs The Difference Engine
• Friday 23 October: HPW, Live! ... with Lara Hogan, author and coach, and former VP Engineering at Kickstarter and former Director of Engineering at Etsy

Meanwhile, there’s the open How People Work... Does Happy Hour open clinic on Tuesdays at 1pm PT, where you get to bring what you’ve been thinking about and we chat about it as a small group. I’ve run three of these so far and they’ve all been fun.

1.0 Some things that caught my attention

1.1 Video games are broken

In Edge magazine, The game industry is broken – can we fix it?¹, there’s an in-depth story covering the current state of the videogame industry. Some bits that caught my attention:

In 2005, Koster studied the previous decade or so of game budgets, and came to a startling conclusion: “The cost of game development goes up about 10x every decade.” In 2017, he repeated the experiment, this time tapping industry contacts to gather numbers on 250 releases spanning over three decades. This larger data set showed him the very same thing. The cost of making a triple-A console or PC game (accounting for inflation) grew from around $1 million in the mid-’90s to $10 million by 2005, and $100 million by 2015. In the blog post accompanying this research, Koster wrote: “We’re talking one-terabyte games that cost $250 million to develop by the early 2020s”. Today, Sweeney talks about “budgets beginning to hit the $250–400 million level” – figures that line up with those in internal documents that leaked from Insomniac in 2023.¹

Essentially, games keep getting more expensive to make, but they don’t actually result in more revenue:

“Since around the year 2000, game budgets have literally doubled every few years,” Sweeney says. “The first Gears Of War, in 2006, cost us $12 million to make, and because we sold six million copies on Xbox, we made $100 million. It was immensely profitable. But, by the end of that generation, as we were making [2011’s] Gears Of War 3, it cost about $100 million to make – and our revenue projection was still about $120 million, because the audience wasn’t growing.”¹

All that’s true apart from the live service games, where you get a game that regularly and frequently receives content updates and makes money via subscriptions and also in-game purchases. Epic’s Fortnite is an example of this, but in the article it’s clear that Epic isn’t have a good time of it either.

And all of this is against the backdrop of Microsoft seemingly having no coherent strategy for videogames other than (facetiously) “have a videogame console and lose money over time by spending a lot of money”, see The Xbox reset gets ugly², from The Verge. (Netflix is also trying to figure out what the deal is with videogames because they’re hitting a limit on attention minutes through video, so need to figure out other valuable things to spend your time on for your subscription)

1.2 Spotify is less broken

Hey, that thing I keep complaining about³, where parents or other people looking after kids would have their end-of-year wrap-ups in streaming music services polluted by obsessive kid music (remember The Year of Moana and The Year of Frozen etc etc) because you couldn’t exclude certain genres or artists? They fixed it! Spotify now has a feature that lets you exclude kids and family music from your taste profile⁴.

Hmm. People have been asking for this for years. I wonder why it suddenly came out? I mean, Spotify have had a whole bunch of other super important strategic things they’ve been working on, like making sure they own the relationship with your ears. And then video podcasts. And all of that jazz. How come they’ve suddenly got time for this? My hunch is that features like the exclude-genres-to-make-parents-more-sane have always been on the backlog (they’re not stupid, after all), they just haven’t been prioritized. So you don’t get to them unless you’ve got capacity and you really start burning through that backlog of stuff to implement.

Anyway it is by pure coincidence that I happened to read about Spotify’s infrastructure for scaling AI coding agents⁵ last month. You would be unsurprised to learn that Spotify think they’re doing super great at scaling out agentic coding infrastructure:

The organizations that learn fastest, that compound their engineering knowledge into shared, structured systems rather than leaving it scattered across individuals, will be the ones that capture the full value of AI-assisted development. We have proven that pattern at scale. Now we're shipping that expertise to you.⁵

If you are the kind of person who is trying to avoid using AI-generated/assisted code, I’m afraid we’re in a post-low-background steel environment⁶, where ever since we started testing nuclear weapons in the atmosphere the background radiation has been, uh “above normal”. (Don’t worry! It is only slightly above normal now, compared to more than slightly above normal!). There’s a bunch of applications where you want steel to, like, not be that radioactive, so it’s been kind of difficult getting hold of it. People started using the low-background steel analogy a few years ago when ChatGPT came out because it came clear that the internet was going to be poisoned forever by LLM-generated material. Surprise! It has been!

1.3 More people have been to Russia than I have

Yeah, that sentence doesn’t make any sense. Sorry if it broke your brain. I discovered comparative illusion/escher sentences a while back and now you get to see them too.

Caught my attention because: you’re kidding, right?

1.4 We actually have an alien?

I am a big fan of the author Max Barry, especially Company, Jennifer Government, and Lexicon, but most recently via Metafilter⁷ I cam across Barry’s short story It Came from Cruden Farm published by Slate’s future tense⁸ which you should read because, well. I cannot give it away. But it was fun. And not fun. I won’t say any more.

That’s it for today! I’m clearly finding it a bit harder to write after a very productive spell. But this episode isn’t really what I thought it was going to be about, either. I suppose this is just how the words come out.

How have you been?

Best,

Dan