Proud to announce another release of my little text editor, kg, now at v1.1.0

Read more about it on my blog https://troglobit.com/post/2026-05-26-long-time-no-blog/

#unicode #microemacs #emacs #terminal #console #embedded #linux

"There's a beaver dam in northern Canada that's twice as wide as the Hoover Dam, and was discovered from satellite photos because it's so remote, and forms a wetland delta despite there not being any distinct creeks or rivers feeding into it." --Beaver Stan Account (on Bluesky), from January 2024.

Very Canadian #JoyScrolling ("If we thought in terms of hockey rinks, that’s 1600 hockey rinks of water.") Sadly, no alt-text for the photographs in the article. #Beavers

https://parks.canada.ca/pn-np/nt/woodbuffalo/nature/beaver_gallery

Whenever I hear from my University administration that they are losing money on research and need the high overheads to keep the lights on, I think about they way that they keep accepting gigantic gifts from donors that require (1) them to build a huge new building, (2) establish huge matching funds, and (3) pay for the maintenance of that building in perpetuity regardless of whether it actually results in increased revenues (including grant overheads and tuitions) to the university.

A new building is a White Elephant.

whenever I hear about people installing millions of lines of code from someone they've never heard of from npm or whatever, I imagine them finding a hot dog on the side of the road

and they're like "oh cool; free hot dog!" and I'm like "you're going to eat that?" and they're like "yeah of course; it's a free hot dog! have to be crazy to turn that down" and I'm like "huh"

My friend's son was a math prodigy, and now he has one-upped OpenAI, which recently wowed the world by disproving a conjecture due to Erdos. OpenAI found a way to put n points on the plane with more pairs at distance 1 from each other than anyone thought possible. Will Sawin quickly found a way to get even more pairs at distance 1.

Math is *not* mainly about solving problems that were already posed. Machines may beat humans at that. I'm glad Will is keeping humans in the competition. But math is about exploration and understanding... as Will knows.

https://arxiv.org/abs/2605.20579

Here's a kind of 101 thing that a lot of people in the world of AI coding are missing, I think.

Question: What are the implications of the fact, "All the tests pass?"

Answer: It actually depends on how the code was written. Unfortunately, the salience of "all the tests pass" has a lot to do both with the strengths of agentic programming and the weaknesses.

If code was written without knowledge of the tests, and then happens to pass the tests, I think I would then be a bit confident that the code is not just passing the tests by coincidence, but is actually correct in a deeper sense that makes it likely to pass future tests that haven't yet been written.

When code is written with knowledge of the tests, or using the tests as a scaffold, then we have to be a bit more careful. Both humans and LLMs do this, but LLMs are better than humans at Monkeys Paw–style trickery, where you have something that is just totally wrong but nonetheless passes all the tests.

So when a codebase is rewritten agentically, the fact that tests pass doesn't make me very confident, unless a lot of the tests were held to the side and not exposed to the agent. This is just basic experimental science (control!), it's hard to understand why this is not obvious to everyone.

Unfortunately, having more tests exposed to the agent increases the quality of the results and the likelihood of correctness. So this is kind of a paradox — but it is a paradox that only applies to people whose method is to interact with the produced code only indirectly. It's easy to break the paradox if you treat the code as your own responsibility and only commit what you understand.

https://fitsonafloppy.com/

Meanwhile, any electron: "hold my 5 gig usb stick"