I’m really enjoying Spider Noir (about halfway through). Nick Cage swings (heh) between various personality poles but he’s always entertaining and even his scenery chewing in-character bad impersonations work. Watching in black & white naturally, I can’t believe they went to the effort to make 2 versions

Hey let’s graph user sentiment about the software they use against the proclivity for PMs to focus on telemetry-driven decision making

Oh…oh no

(FWIW I pushed back hard against that “2 months ago” bullshit when I made SourceTree, despite pressure to adopt it because almost every other VCS UI did it just because they thought it was cool (it isn’t). But then I actually used my own fecking product so knew it was a practically useless date display convention beyond about a 24 hour window)

Re LB: this “3 months ago” bs instead of showing an actual date is precisely the sort of thing that shows that analytics-driven product management is absolute horseshit. No amount of staring at numbers will tell you that people who use a thing every day absolutely hate something about it. Use your own product and talk to other people you morons, and stop being a clueless telemetry pervert

I want a Firefox extension that knows how to replace every single "N months ago" label with the actual exact timestamp (using patterns for how the data is hidden in tooltips, click-to-expand, etc.).

That shit is SO unhelpful. Like if I'm looking back at git history, I don't care that a change was made roughly 3 months ago. I care whether it was made before or after an email I'm looking at that said something relevant to the change.

It is possible for it to back itself into a corner by randomly turning a way that would give it no option but to cross over a previous chamber. But if that happens, even with the score penalty for overlapping previous chambers it is still the "best" solution, so it does it anyway. Messy, but usable, which is the most important thing.

I'm actually kinda liking these unexpected kinks, they wouldn't happen with the physics idea unless I really overstuffed it and even then it would resist it more

The wonkiness introduced to these maps is down to the system not picking the best case, just a random one within a certain % of the best. This causes it to sometimes pick a chamber which bends off in a non-perfect direction, which can occasionally get it into trouble. But once it's headed the wrong way it's strongly encouraged to turn around by the scoring, while still avoiding the existing chambers. Hence what happened in that last gif

I've tried to optimise for robustness, because I've had problems before with directed procgen systems not being able to avoid the halting problem - the more constraints you add, the more likely it'll loop forever trying to resolve conflicts in edge cases. Ultimately the wandering cave generator will *always* reach the end, even if it has to take solutions it doesn't like. Working with ultimately a free density field helps - I prefer not to stomp over prior results, but it's fine if I have to

Lol mistakes were made, but we got there

Complete loops are now working!

I particularly like how it kind of screwed up on the last corner and was heading too much downwards, but recovered with a cheeky turn in the end

Good coding music, for the kind of sicko like me who still does that https://anjunabeats.com/gb/products/593319-anjunabeats-volume-17

How am I supposed to work when I see this face when I lean around my monitor?

“The thing that made open source maintenance joyful was that someone cared enough about your work to try to improve it. You mentored them. They learned. The code got better. The community grew. That's the social contract.

AI slop doesn't add work. It replaces human connection with mechanized noise.”

https://ryancheley.com/2026/08/17/what-if-maintainer-burnout-isn-t-burnout/

It doesn't pick the "best" option all the time, to avoid being too optimal, it picks a random option from the top N% scoring options

I've added more visual debug information to the editor so I can better understand the choices the walking algorithm is making. I can scrub forwards and backwards through every step which is very handy

new venjent, this time actively flaming Suno (the AI crap) XDDDDDD https://youtu.be/GejdMIM2Tjk

Wooo, actual rain you can *hear* today, who’d have thought I’d miss that 🌧️

The walking algorithm doing its thing. I'm still refining this before I try the physics spring/joint idea, since I figure they're all going to have edge cases; might as well finish one properly before embarking on another (as tempting as it is).

(These are just the bounding convex shapes of the chambers when walking from one designed node to another, some overlap between adjoining chambers is ok & indeed desired)

Anyway, this makes sense in my head but probably has some horrid complication I’m currently blind to that makes it impractical, but it will be fun to try 😄

Then I keep inserting 1 new chamber in every edge in a round robin fashion, stopping adding to an edge when it has exceeded the desired length in the design. I let the physics push them apart, hopefully forming interesting “crinkles” because of the joints between the chambers. Perhaps either some damping at nodes, or another spring (distance) joint between the nodes can keep the tension to encourage that.

I might try to prototype this with Box2D so I don’t have to write my own physics