David Nash

Home: Part of an amazing #Polyfidelitous family (3 adults, 4 kids)

Work: #DataEngineering currently. #WebDevelopment and #DevOps in the past. #Chemistry a long time ago.

Developer of the HYG + AT-HYG star catalogs, and star chart software (https://codeberg.org/astronexus).

Expect lots of posts on #Cats, #Astronomy, #Computers, #Books, and other geeky topics here.

_Glacies delenda est._

Website: https://www.astronexus.com
Codeberg: https://codeberg.org/astronexus

Public Key

npub1vgrrtppfx6quyrz5jw9k6pannzhz87f8gvsvh76fwt2h3vn44gaqqjyu5s

NIP-05 Address

dpnash_at_c.im@momostr.pink

Profile Code

nprofile1qqsxyp34ss5ndqwzp32f8zmdq7ee3t3rlyn5xgxtldyh94tckf665wsprpmhxue69uhhyetvv9ujumt0d4hhxarj9ecxjmntpf3m5d

Publishing to

relay.momostr.pink

Author Public Key

npub1vgrrtppfx6quyrz5jw9k6pannzhz87f8gvsvh76fwt2h3vn44gaqqjyu5s

Show more details

Last Notes

2026-04-22 06:19:17 UTC

- reply

Just another day in the office: opening a spreadsheet in Excel…no, it’s a fscking Sharepoint-flavored spreadsheet that opens in a web browser … no, shit, it’s in Teams, so of *course* it not only opens in Teams, it steals the window context so completely I can’t go back to the chat group I was just in and need to get context from.
Good thing I’m avoiding gen-“AI” at work. There’s probably a Microsoft Document Copilot by now that’ll swipe context for any Excel file and force Microsoft Copilot for Excel to open it and invite me to have a “conversation” with my spreadsheets.

2026-03-31 19:40:22 UTC

- reply

My schaden is nicely freuded after seeing both this code dump and the fustercluck where people on Anthropic’s $100/$200 monthly plans are blowing through their 5-hour and weekly token allotments in no time flat.
Linked Reddit thread has numerous examples of pissed-off users, my favorite so far being the person who blew through the 5-hour quota trying to get Claude to realize that the 24th of March this year was not, in fact, a Monday. https://www.reddit.com/r/ClaudeAI/comments/1s7fcjf/claude_usage_limits_discussion_megathread_ongoing/

2026-03-18 02:05:07 UTC

- reply

The stats for the vibe-coded SQLite rewrite in Rust being literally thousands of times slower than SQLite are simply wild. Such as needing almost 2 seconds to do 100 single-ID lookups. I'm pretty sure I could improve on that operation just by slurping a CSV with unique row IDs into memory and doing a binary search on said row IDs. Would I want to? No, but I'm also not *trying* to build an actual RDB engine either.

2026-01-18 18:44:08 UTC

- reply

The title of this article is "Welcome to Gas Town."
You know how it seems every techbro reference to Tolkien is something cursed or evil without any self-awareness (e.g. Palantir)? Apparently, they're now extending this lack of awareness to the *Mad Max* series, only it's "impoverished dystopia" as the namesake.

2026-01-06 18:58:28 UTC

- reply

> Software engineering requires a very wide "context window," the thing that AI does not, and cannot have.
The type of work I'm doing now (data engineering for a large organization) is full of the sort of software development you describe here. Whatever code I write has to cope with data coming in, from multiple not-mutually-friendly parts of the company, and it has to at least try to produce consistently comprehensible (and reasonably updateable) data for downstream users or processors. A huge part of even beginning to make that possible is understanding, in detail, what those parts of the company actually want or need. That's generally the most challenging part of my day job. The code is the easy part, and being able to puke out more code in less time rarely, if ever, solves the hard parts.

2025-12-25 22:19:40 UTC

- reply

Good thing you got the typewriter now, because as of January 1, 2026, the options get much worse. After then, the non-subscription version will only permit the type bars for 20 of the 26 letters to advance fully. To get “E”, “T”, “H”, “A”, “I”, or “S”, it’s $1/month per letter ($5 for all 6, $50 for all 6 if paying annually).

2025-12-15 06:08:44 UTC

- reply

I know, right? I managed to inoculate myself against the hype almost from the start by asking ChatGPT 3.5 questions about something I knew well, and getting unambiguous confabulated bullshit in response. I then saw plenty of other confabulated bullshit in other topics, all of which I could easily verify was bullshit. My own take on all that — all in the first few months of ChatGPT’s public release — was “it gets numerous easily verified facts wrong, so it is *not trustworthy* on anything”. It was astonishing to see so few other people I know come to the same conclusion.

2025-10-24 15:05:38 UTC

- reply

There's a bunch of data centers in Oregon (it's an entire AWS region all by itself: us-west-2) because in the northeast part of the state, there's a lot of cheap land and cheap electricity.
The land is cheap because it's a desert with no large cities. You have a reasonable amount of water from the Columbia River, but that's about it, so anything that sucks down water makes life harder for everyone else there.
The electricity is cheap because there are giant hydroelectric and wind power plants there. All renewable or at least almost-no-carbon-dioxide-emissions electricity, meaning that with a bunch of new data centers, there's that much less renewable electricity going to serve actual homes and productive business.

2025-10-19 17:49:50 UTC

- reply

If you come from Bumpass, VA and commit acts of unspeakable cruelty, you can expect to be banished to...
Bumpass Hell, CA. (Ok, that's not really a town. It's a location near Mount Lassen, that's inside the national park boundaries and thus prominently marked on the maps.)
https://www.nps.gov/lavo/planyourvisit/maps.htm -- look in the southwest part of the park, just a bit north of the southwest entrance.

2025-08-11 16:58:11 UTC

- reply

I've probably mentioned this before, but just in case not:
If you took an "AP English" class (mine was equivalent to the modern "AP English Literature"; I don't think there was a Literature vs Language breakdown back then) in the 80s or 90s, and then took the official AP test for possible college credit, the test graders didn't care whether the essay was 5-paragraph-essay shaped or not.
I know this because one of my parents volunteered to grade some practice exams for one of my school's AP English classes (not mine; no nepotism occurred here), which were direct excerpts from actual AP exams from a year or two back. So I got to see what the last couple years' rubric looked like. The format of the essay didn't matter, so long as it made sensible arguments that were supported by what was actually in the work being analyzed or critiqued, and that it showed an understanding of literary techniques discussed in class.
To get a good score, you couldn't hack the system with a ChatGPT-style 5PE-shaped pile of word barf. It actually had to be a good analysis, at least by the standards of a college-level introductory English lit class.

2025-08-07 20:34:20 UTC

- reply

Ah, and since this seems to be making the rounds: it's time for my occasional reminder of one especially unpleasant aspect of the LLM hype storm.
LLMs take advantage of a faulty heuristic many (probably most, but don't have exact stats) people have about human intelligence: namely, someone or something that produces fluent and grammatically coherent text about a wide range of topics, on demand, is "intelligent".
This is a very compelling heuristic and also very wrong.
In the case of LLMs, it's wrong in the way we're all talking about here: it leads people to see thinking or reasoning in the statistical word salad these beasties produce.
But it's even more of a problem when people invert the heuristic and turn it into "someone or something is *not* producing fluent text about a wide range of topics, therefore they are *not* intelligent". This mirror-world version causes enormous social harm to people who for whatever reason (autism, selective mutism, etc.) can't speak the way people expect they would be able to "if they were smart".

2025-08-07 16:13:18 UTC

- reply

One of the first prompts I gave ChatGPT back in 2022 came from my main hobby (amateur astronomy). I asked it to tell me something about the extrasolar planets orbiting the star [[fake star ID]].
[[fake star ID]] was something that anyone who knew how to use Wikipedia half-intelligently could verify was fake within a few minutes.
I wasn't even trying to be deceptive; I genuinely wanted to see how ChatGPT would handle a request for information that I knew couldn't be in its training data.
The torrents of bullshit it produced -- paragraph after paragraph of totally confabulated data about these nonexistent planets orbiting a nonexistent star -- told me everything I needed to know about ChatGPT and its buddies, and I've never been tempted to use them for anything serious since.

2025-07-30 18:16:25 UTC

- reply

This reminds me of a particularly evil deceptive pattern from Reddit: for a time (and possibly still now) the mobile site version blocked access to some subreddits, with an alert stating that the subreddit is not visible without the mobile app. On an iPad or other device that easily toggles between desktop and mobile site mode, the mobile app nagwall appeared and disappeared like magic when you switched modes.

2025-07-01 00:30:10 UTC

#Gravy gets its name from the 17th century Englishman, Alistair Dalgliesh Throatwimble Graves III, who was such a renowned amateur chef in the Northwest East Midlands that people eating at his house would exclaim “it’s absolutely Gravesy!” after enjoying one of his dishes. Eventually, for reasons not entirely clear, the “it” became specific to meat-based sauces. (The primary alternative etymology, that “gravy” was originally consistently gray in color, was conclusively disproved by the late-20th century discovery of light brown gravy in a _thermopolium_ in the ruins of Pompeii.)

2025-06-30 19:50:06 UTC

- reply

I recently read about a new "AI" company that purports to create "AI"-based digital copies of yourself (or anyone, really) that you can ask questions anytime (no pesky downtime for sleep or whatever). I went and had a look. Naturally, they have job openings. Naturally, they tell you on the job description page to avoid all AI use in the job application.

2025-06-04 18:36:15 UTC

- reply

I wish Google AI lots of luck populating the galaxy with the help of AI when AI doesn't even know where the stars are.
One non-AI slop program showing the sky as it actually appears, and two AI slop versions that are ... not even close. And the worse one of the two is from Google's own Gemini AI.
https://s3.c.im/media_attachments/files/114/626/505/362/967/221/original/b394638782d1b688.png
https://s3.c.im/media_attachments/files/114/626/515/875/134/011/original/c5509aed6240ef09.png
https://s3.c.im/media_attachments/files/114/626/522/846/503/284/original/c08a79fc854ffd89.png

2025-05-05 17:58:28 UTC

- reply

Oh, and while I'm on this topic, here's your periodic reminder of something else important:
People in my part of the world (the USA), and very likely those in similar or related cultures, have a flawed mental heuristic for quickly judging if someone or something is "intelligent": "do they create grammatically fluent language, on a wide range of topics, quickly and on-demand"?
This heuristic is faulty -- it is very often badly wrong. Not going to have a long, drawn-out debate about this: it's faulty. The research is out there and not hard to find, if you really need it.
In the case of LLMs, this heuristic leads people to see intelligence where there isn't any. That's bad enough. But it also leads people to *fail*, or even *refuse*, to acknowledge intelligence where it does exist -- specifically, among people who don't talk or write very articulately.
In the specific case of the USA, this same heuristic is proving to be very dangerous indeed, with the federal government wanting to create official registries of autistic people, for example. The focus is overwhelmingly directed towards autistic people who can't or don't speak routinely, and it's *appalling*.

2025-05-05 17:47:00 UTC

- reply

> Have you validated your alleged tutor against a topic in which you are well versed?
ChatGPT and related programs get things wrong in areas I know well, over and over again.
They've told me about stars and planets that don't exist, how to change the engine oil in an electric car, how water won't freeze at 2 degrees above absolute zero because it's a vapor at that temperature, that a person born in the 700s was a major "7th century" figure ...
over and over again, like that.
But worst of all? All delivered with a confidence that makes it very hard for people *who don't know the topic *to tell that there is a problem at all.

2025-04-11 23:29:10 UTC

- reply

> Had people been more resistant or the results just a little less able to fool people who don't know better, they would have remained a niche thing, and it's sort of too bad it didn't work out more that way.
I unintentionally but successfully inoculated myself against the hype by giving ChatGPT, shortly after it came out, a very simple request: "Describe the extrasolar planets around <<fake ID for star that doesn't exist>>."
I verified with a quick internet search that the fake ID didn't appear in any web sites or was easily confused with real ones that might appear in legitimate data. As far as I have been able to tell, nobody on the internet has ever used that fake ID before. It wasn't going to be in any LLM's training data.
I honestly was not prepared for what actually happened.
Instead of telling me something like "I don't know anything about that" or "Your guess is as good as mine, maybe check <<popular space science web site names>>", ChatGPT came back with multiple paragraphs of unadulterated bullshit. Complete fabrications of the orbital details and locations of nonexistent planets around a nonexistent star. I knew immediately what I was dealing with at that point, and have never been even slightly tempted to use it for anything except poking fun at "AI" hype.

2025-02-22 20:50:56 UTC

- reply

10-ish year old me, reading The Door into Summer: “That was fun. The cat was very…cat, for sure. But what was up with the old dude wanting to meet up with the young girl in a few years?”
Maybe 5 or so years later, recalling things: “Oh, *yikes*.”
Never reread it, though I’m sure I’d enjoy the cat.

2025-02-20 02:40:54 UTC

- reply

You may be out of luck with letters, but “nineteen characters” has 19 characters inside the quote marks.

2025-02-07 06:02:26 UTC

Loki: meow
Me: what?
Hazel: meow
Me: Yeah, it’s been a tough couple of weeks for kind and decent people here in this country.
Loki: meow
Me: Yeah. That’s right. It’s important to keep the people (and cats) you love close by.
Hazel: meow
Me: That’s also right. Hang onto thoughts, memories, places, times that give you joy.
Loki: meow
Me: Exactly. Those are things nobody can take away. Your joy is one of your strengths.
Loki: meow
Hazel: meow
Me: Love you too.
#CatsOfMastodon #Fedicats
https://s3.c.im/media_attachments/files/113/961/075/581/722/642/original/31b45f88673a8156.jpeg

2025-01-28 03:35:22 UTC

The #CatsOfMastodon and #Fedicats wish to remind you, person who is fighting fascism and authoritarianism, that you cannot win that fight without plenty of rest! Be sure to take naps often, especially if you can share space with a cat you can skritch as well.
https://s3.c.im/media_attachments/files/113/903/871/359/447/306/original/13f5a604cb8bdcc0.jpeg

2025-01-12 23:30:54 UTC

- reply

Even more specifically, the IAU gives a maximum of one star in a system a specific proper name—which is why they had to use 2 traditional names for Alpha Centauri to name the two bright components.
This rapidly gets awkward when trying to draw and label a star chart when you’ve got a multiple star system and only one component is assigned a traditional name by the IAU.

2025-01-12 21:54:03 UTC

- reply

> I wonder when a star will get *really* close to the Sun, like 1 or 2 light years?
Gliese 710 is heading almost directly towards the Sun and will pass about *1/6* light year from the Sun in about 1.29 million years. The Hipparcos mission got accurate enough data to show Gl 710 will get close, but Gaia refined the figures significantly in the last few years.
Spectral type K7 V, so not a red dwarf, but also well below solar luminosity. Still more than massive enough to do…umm, fun things to the Oort Cloud.
https://en.m.wikipedia.org/wiki/Gliese_710

2025-01-04 19:18:01 UTC

- reply

I tried this Google search. I got the same wrong answer (3.18 inches = 1/8 inch).
So I clicked the “Show More” option, and it pulled up a scan of a unit conversion chart that has (correctly) 3.18 *millimeters* = 1/8 inch.
Utter garbage from what used to be a premier source of basic information online.

2024-12-06 18:18:23 UTC

- reply

INSTALLER NOTES: Congratulations! The installation of your fusion reactor (class G2 V) was successful. The optimal distance from the reactor for lifetypes in your category ("Nucleic Acid and Amino Acid Polymers") is 120 million to 180 million kilometers. Please do not try to establish a life-colony on planets outside this range. CAUTION: Reactor slowly increases radiant output over time as fusion products accumulate. After approximately 8 billion years, it will no longer be class V, and all remaining warranties will be void.

2024-10-24 21:25:42 UTC

- reply

> What’s wrong with “delve”? It’s a perfectly fine and useful word.
But I’m not a native speaker of English, so what do I know?
Just by using “delve” regularly as a non-native speaker, it shows you recognize quite a bit more than the native English-speaking techbros behind the push for AI everywhere.
There’s evidence that ChatGPT’s fondness for “delve” (and likely other English words) comes in large part from people in places like Nigeria where (a) “delve” is common in the local flavor of English and (b) a lot of people have contributed to LLM model training and refinement.
https://www.premiumtimesng.com/opinion/689088-delve-artificial-intelligence-diversity-and-inclusivity-by-chiamaka-okafor.html

2024-09-17 20:50:11 UTC

- reply

Could easily be me being wrong then. I remembered that “Earth is oblate” was one of the novel conclusions from hauling a pendulum to the tropics and seeing it run slow, so I thought that the oblateness factor was dominant.

2024-09-17 19:50:43 UTC

- reply

Hard to do precisely, especially since things like altitude and nearby large masses (big mountains, for example) have an appreciable effect as well, because they affect the local strength of gravity. But once pendulum clocks became accurate enough to be useful as precision timekeepers, they became precision gravimeters as well, and people figured out how this acceleration changed with latitude. Biggest effect, interestingly enough, was the Earth being oblate instead of perfectly spherical, rather than the direct change in acceleration from centripetal effects. Though of course, that shape is itself a centripetal effect. https://en.wikipedia.org/wiki/Jean_Richer