0.0 Context Setting

Friday, 14 August 2026 in Portland, Oregon, pretty much straight after finishing a live show. I’ve got a full day and this is an attempt at getting a newsletter out in a respectable amount of time without overloading people with words. Short is definitely better sometimes.

0.1 Events

The next How People Work, Live! ... with Marcin Wichary is on Friday 21 August at 9am Pacific. You should register for it. Here’s Marcin’s bio:

Marcin Wichary started his career at Google working on internal tools, the search team, the doodle team, and the HTML advocacy team. He later spent time at Code for America and Medium, and his most recent position has been working as a design director and design architect on the editor team at Figma.

Marcin has also been speaking and writing about craft, typography, and interface design since the early 2010s. In 2024 he published Shift Happens, a book about the history of keyboards and typing, and his most recent endeavour has been a blog about software craft and quality called Unsung.

You can watch previous How People Work, Live! episodes from the How People Work, Live! show page, and see what’s coming up, too.

The very first How People Work... Does Happy Hour is on Tuesday 18 at 1pm Pacific. It’s a free open clinic where you can come along and bring your problems, and we’ll work through them based on the workshop principles in public. It’s a small clinic, only ten spots, and as of writing, there’s only six spots left.

1.0 Things That Caught My Attention

1.1 Trinity

This morning was a How People Work, Live! show with weaver and, like all the shows, left my brain reeling with thoughts.

You should go watch the episode, I think if I’m not being too full of myself it’s pretty interesting and relatively easy to follow along to if you’re interested in how large organizations deal with crises. It should be the newest episode at the show playlist on Youtube by the time I’ve sent this. I hope.

We managed to avoid talking about AI until the very end of the episode -- I’m very likely to have him back to talk about that specifically, and might be launching a companion show specifically covering how people work with AI called, uh How Centaurs Work. More on that later.

(Yes, I know about Cory’s new book).

Anyway. There were two things that weaver brought up that stuck with me that I need to get out of my head/am thinking about.

One, a thing Bruce Sterling pointed out on bluesky a while back, paraphrased: you know what’s like slop? Something that’s infinitely malleable, formable, and is absolutely everywhere? Plastic. The future is plastics.

And what weaver and I were in agreement on, or rather, what weaver said that mirrors my position is that this has happened and it’s not going away. He gave another example from Brian Bucklew, creator of the game Caves of Qud (highly recommended), which is a hugo award winner. A game that’s a Hugo award winner! That’s a whole separate episode. Brian’s been doing super interesting introspection on AI in terms of software development and in particular with the type of development he does in games. Weaver thought this was a particularly insightful thought of Bucklew’s: the sonic boom has happened. The nuclear detonation has happened. Software engineers are in the bow shock wave, it’s definitely happened and you can’t go back.

I think “we” need to accept this. Certainly we can try to go back in certain respects, but there qualitatively is no going back. The trinity event has occurred. Sam Altman has already a big enough ego and cult, I bet he’s already likened himself to Oppenheimer at that the launch of ChatGPT was a “now I have become death, destroyer of worlds”. But he’s right, and I believe this cannot be undone, cannot be rolled back.

So in that respect, software engineering does need to reckon with itself about what the profession is, what it does, and how involved it is in the generation of, maintenance of, and general production and use of software in all of society, from the smallest aspects, tiny teams leading up the largest enterprise teams.

Weaver raised another point: we just don’t know how much this costs, never mind in the negative environmental and social externalities, but also the absolute dollar cost. The whole thing is underpinned by subsidized tokens and subsidized access. My interpretation of that or synthesis of that is this:

Be aware of the token envelope you’re operating in; don’t overextend your token budget

People are out there using free resources and actually accomplishing things. I’m firmly in the camp of “people are producing software that delivers genuine utility and value to them” DESPITE and or WITH the fact that there are definitely problematic costs involved that are being deferred. We’re humans, we’re the universe champions at deferring costs and waiting for consequences to become too painful.

Well, how can you figure out what your token budget is when you don’t know how much it costs? The entire system is so distorted. I mean what you could do is you could do what I was doing the other day: check out the wafer-scale systems that Cerebras is selling (you know, the ones that scale up to 52 cabinets and a ~2.54 MW heat load) and AT THE VERY LEAST figure out what I’d call a closed-loop energy systems. How much renewable energy capacity and storage capacity do you need to power your local model?

And that’s before you get into potential societal directions like outlawing open weight models. We’ve outlawed code before, we can do it again, and yes, there are always ways around it, but not ones that an organization with a risk profile may be prepared to accept. I mean hell, we’re already outlawing the code to describe a 3D printed gun and requiring software that’s impossible to write to recognize such. Side thought: will the direction public opinion is going end up treating open weight models like CSAM data?

So that’s a quick what caught my attention this morning, and it’s only 11:10am. I wanted to get that one out, and I really want to have weaver back to continue this discussion.

There you go! A short one! I set a 5 minute timer when I started writing, and then extended it twice, I think, and tried to figure out how to stop. Maybe it worked?

Anyway, how are you doing?

Best,

Dan