TIL

I try to work on my garden everyday. Some days I neglect it and some days I try to make up for it. And some days there’s quiet work that may or may not end up visible in the hotbed at that particular moment.

Today was simply one of those days where there was too much information to sensibly put into the normal garden. So I decided to make a post out of it to perhaps trigger myself to come back to it more.

Anyway, here it goes. Some of the stuff I read and reflected on today, in perhaps no particular order:

Word of the day: Accretive.

Accumulative as in interest, or compounding. Or rather, buying something that increases the value of the total more than the cost of the purchase.

Anyway, on to the actual consumed content. First off:

Steve Yegge again.

Really encouraging and dismal at the same time peeking through the lens of his. It sure looks like there could be something fascinating at play with getting a completely new work model in motion like they do over at Claude Home. Quotes below. The first one on critiquing Spec-driven development

The center of the campfire is a living prototype. There is no waterfall. There is no spec. There is a prototype that simply evolves, via group sculpting, into the final product: something that finally feels right. You know it when you finally find it.

This is interesting! It resonates with what Kent Beck says. That we learn from feedbacking. We almost never know what to build up front. If we do then we also can usually just build it…

As evidence of this, Anthropic, from what I’m told, does not produce an operating plan ahead more than 90 days, and that is their outermost planning cycle. They are vibing, on the shortest cycles and fastest feedback loops imaginable for their size.

Okay this is interesting. Anthropic, currently the world leader of ai development (arguably) doesn’t have a year-ahead plan. Not even a tertial. Just 3 months. Ok, I’m listening.

And this one, apparently the future of work is improv! I can jam that. Just like music!

Anthropic’s Hive Mind is described by employees as “Yes, and…” style improvisational theater… and when magic happens, they all just kind of realize it at once.

So we’re seeing real power in the “Yes, and…” model. And yet, most companies arrived at where they are by learning how to say No. This is shaping up to be a problem. More To Come

Wow! I can see that, if we’re vibing more, creating a ton more, then we can simply bounce more ideas back and forth and riff of off each other without the bureaucracy fiddling that traditional development may entail. I do think I can see it, I’m still a large ways away from it though.

Your organization is going to have to learn a bunch of new lessons, as new bottlenecks emerge when coding is no longer the bottleneck.

Yes! Lean comes back into play. If the value stream now has other points of friction than code delivery, how do we exploit them?

Just some good pointers

…on how to engineer sensibly with AI.

Like sketching down the guidelines, clearly defining the vision (so that you drive the AI, and the AI doesn’t drive you! That happens to me all the time…). Also marking stuff for code review explicitly dependning on criticality was interesting. I’m not sure this really applies in a context where every line of code does get reviewed before merge, but I guess a lot of people are already doing it the Peter Steiner way without having the capacity (I know I certainly don’t).

Rube Goldberg

Okay. I’ve come across Rube Goldberg machine one too many times now to not know what it is. Apparently it’s a completely over-engineered machine for some trivial purpose. Like those extravagant domino setups that just flick a switch upon completion. Named after a cartoonist. There you go. #TIL.

Nix

Had a little conversation with gippity around what nix really is - config heaven for a full-blooded infra-dev-ops-person such as yours truly. But probably not worth the hassle. Love the idea of a dedicated config language, automatic path resolving and immutable hashes for dependency resolving. But mise looks like a decent substitute!

Friend tipped me off: GHA sux

Worthy rant on github actions being well subpar to but it mildly. And I can’t disagree with a single thing he’s saying! And the quotes are exquisite!

Have you done the ${{ }} expression dance, misquoted something, and then waited four minutes for a runner to spin up just to discover your string got eaten? Of course you have. We all have. We have all stared at a diff that changes one character in a YAML expression and thought “I went to college for this.”

Yes I have.

Buildkite’s log viewer is just a web page that shows you logs and doesn’t crash. I realize that’s a low bar. It’s a bar that GitHub Actions trips over and falls face-first into the mud, gets up, slips again, and somehow sets the mud on fire.

Ok that’s a beautiful paragraph, that’s what that is.

put the orchestration in config and the logic in code.

^ Is actually a very sensible way of looking at it! Still need to think through what orchestration equates though, and how it would differ from config. But I guess its resolving the dependencies that should go into config!

Anyway. Nothing more structured than that. Just thought it was content enough that I should consider marking it for a blogpost to reflect on things and hopefully revisit at some point. Til next midnight session. C ya.