Read it here first

2026-05

05/20 16:56 hack 116605833311345490
SILE, https://sile-typesetter.org/what-is-sile/ a modern rewrite of TeX, appears interesting. I’m guessing SIL stands for the Summer Institute of Linguistics, responsible for XeLaTeX.

05/18 18:43 en 116594926491577193
I wonder what the effects of the Canvas hacking/defacement/data exfiltration on Moodle, Blackboad, other LMS are. It appears that incident is not a ransomware via an encryption lock on organization data, more just pay-us-or-else vandalization ransom demands
⤶ previous post

05/18 18:34 en 116594893901046310
Matthew Pittinsky, incoming CEO of Blackboard, which he co-founded in 1997, has an interesting appreciation of David Labaree’s books, re-posted at https://davidlabaree.com/2026/05/04/matthew-pittinsky-three-books-by-one-author-that-every-edtech-entrepreneur-should-read-and-why/ There is little mention of LMS, more like a 10,000 ft view of education.

05/18 18:02 en 116594766842249250
I remember a teacher himself a Moodle developer saying the gap in the 3 developer/teacher/student groups’ attitudes to the app were wider than their attitudes to each other.
⤶ previous post

05/18 18:00 en 116594757468480335
recognize/problematize/deny/pass over number 2: “Access to digital tools is not the same as their use in daily classroom activities,” with solutions/reasons/stories about experiences,

discuss/problematize/deny/pass over number 3: “Designers and entrepreneurs overestimate their product’s power to make change and underestimate the power of school organizations to keep things as they are,” with regret/acceptance/solutions/reasons/stories?
⤶ 116594739541924237

05/18 17:56 en 116594743681542802
recognize/problematize/deny/pass over number 2: “Access to digital tools is not the same as their use in daily classroom activities,” with solutions/reasons/stories about experiences,

discuss/problematize/deny/pass over number 3: “Designers and entrepreneurs overestimate their product’s power to make change and underestimate the power of school organizations to keep things as they are,” with regret/acceptance/solutions/reasons/stories?

05/18 17:55 en 116594739541924237
Restating the 3 principles of https://larrycuban.wordpress.com/2026/05/16/lessons-i-have-learned-about-technology-and-reform/ in the form of questions about the reaction to those principles as presented by Larry at an ed-tech and teacher meet, did designers/developers/teachers:

affirm/deny/pass over number 1: “Teachers are central to all learning,” with passion/insight/indifference,

05/18 15:48 sec 116594239582808672
Ben Cotton (@funnelfiasco@hachyderm.io) takes a stand on politics in open source projects https://duckalignment.academy/politics-open-source/

05/18 15:35 sec 116594189277495178
Fedora linux kernel vulnerability response statement: https://fedoramagazine.org/how-fedora-is-responding-to-recent-kernel-vulnerabilities/

05/18 15:07 ja 116594078359269455
今マイクロソフトが僕のWindows10システムで見せるブートスプラッシュのイメージは可愛い（？）熊の写真だ。MSがなんと恐ろしいものだとわかるか？2025年被害者は238人、うち死亡が13人がいるという。可愛いものが危ないとか、危ないことが魅力あるとか？

05/18 14:21 ja 116593897900310285
卵は腐らないように包装して、料理により使いやすいという肉の変わり者とかんがえたらいいだ。

05/18 10:24 en 116592964345839726
Can the pro/anti-LLM divide be explained by some ad-hominem personality characteristic, rather than intellectual disposition? Eg 1. field in/dependence https://en.wikipedia.org/wiki/Field_dependence , 2. intro/extra-version https://en.wikipedia.org/wiki/Extraversion_and_introversion 3. social dominance https://en.wikipedia.org/wiki/Social_dominance or some other socio-psychological variable

05/18 8:47 hack 116592583741548205
And backslash escapes of variables not supposed to be there.
⤶ previous post

05/18 8:43 hack 116592568579680859
Yay. Took days to script the start of a tmux session with multiple windows. With

declare windowCommands=’’ 34 for window in “${SCREEN[@]}” ; do 35 windowCommands+=“; neww -n $window bash –rcfile ${RC[$window]}” 36 done 37 tmux new-session -s GhcUp -Ad 38 tmux -v rename-window -t ^ base $windowCommands ■ Double quote to prevent globbing and

No. Quoting not wanted.

05/18 8:24 hack 116592495821923102
Line 38 should be: 38 tmux -v rename-window -t ^ base
⤶ previous post

05/17 20:43 hack 116589739428455856
Yay. Took days to script the start of a tmux session with multiple windows. With

33 declare windowCommands=’’ 34 for window in ; do 35 windowCommands+=“; neww -n bash –rcfile” 36 done 37 tmux new-session -s GhcUp -Ad I 38 tmux -v rename-window -t ^ base ■ Double quote to prevent globbing and

No. Double quoting not wanted

05/17 16:32 en 116588749059425971
I guess I would find them the same way I found @infosec.exchange, @hachyderm.io, @fosstodon.org & @mathstodon.xyz. They were the instances of (a number of) accounts I was already following and so I gradually became aware of them.
⤶ previous post

05/17 16:11 en 116588669289532565
I haven’t been able to find instances like this security and these computing ones where people interested in education, like teachers, parents, academics gather. The same for people interested in language and linguistics.

Do they exist? How would you find them?
⤶ previous post

05/17 16:07 en 116588653683600336
At https://fedi.garden/fediverse-servers-sorted-by-topic/ I don’t see @infosec.exchange in the Security and Privacy section and @hachyderm.io or fosstodon.org in the Computing section. These instances are representative of their interests, meaning many like-minded users, some well-known outside mastodon, have gathered there. So they are canonic like mathstodon.xyz is in math.
⤶ 116585271449122385

05/15 14:38 math 116576976949226238
What is a hypernym for baked flour dough products like bread, cake, pies, pizza crusts, pancakes (?), like “pasta” is for spaghetti and other similar extruded then boiled dough products? “Baked flour dough products”?
⤶ previous post

05/15 10:49 math 116576077236171458
Cutting a pancake-shaped crust into 6 pieces with 3 cuts at orientations differing by 60°,(π/3 radians), the orientation of the first is free, but judging the second is a little difficult. Doing it a number of times, I finally saw it makes a piece that looks like an equilateral triangle. So the third cut of the semicircumference, length π * r produces 3 arcs of length π / 3 * r. Can you distinguish that from a length r? The secant is r.

05/15 9:57 hack 116575871307071974
The unexpected consequences of a long build that crossed the UTC midnight dateline:

WARNING: Built wheel for toot-tooi is invalid: Wheel has unexpected file name: expected '0.21.3.dev4+gd7a13e4de.d20260514', got '0.21.3.dev4+gd7a13e4de.d20260515'

It was 0900 here!

05/15 8:26 en 116575513120490954
No 2 instruments more incompatible than didjeridu and piano, so googling “composition didjeridu piano” expected nothing, but more than 5, 6 works/performances are listed

05/12 21:52 hack 116561697365725815
Alerted by https://rss-parrot.net/u/bsky.app.profile.melaniemitchell.bsky.social/status/1778472231138931616 to https://knightcolumbia.org/content/ai-as-social-technology , in seeking to frame the LLMs/chabots/AI as social technology, it recalls, “Every problem in computer science can be solved by adding another level of indirection (abstraction). Ie, more abstraction than thou. Really, though, abstraction is good. HA Simon is cited for his AI and Nobel-winning bureaucracy work

05/12 20:39 hack 116561411098853345
My idea that OMG has negative connotations appears to be incorrect. I think it was on TV that I just came across an instance that was if anything positive. It just expresses surprise?
⤶ previous post

05/11 17:21 en 116554970929194018
Series: singular or plural agreement. https://www.merriam-webster.com/grammar/series-singular-plural-grammar-usage is good on ins-and-outs, mentioning notional agreement https://en.wikipedia.org/wiki/Synesis In

To examine the sensitivity of model outcomes to key parameters, a series of sensitivity analyses was conducted on the small-scale 3-zone instance, focusing on rental demand, existing vehicle fleet size, supply arc cost of new vehicles, and holding arc cost.

The original had a series of sensitivity analyses were conducted

05/11 10:59 en 116553465740500938
Watson, Seth and Malik inspire the feeling Dawkins is pulling our leg, having us on, like the chatbots do.

Human-curated because I want good ones. Googling will just return popular ones. An important distinction.

I’m also looking for more appreciative ones.
⤶ previous post

05/11 10:57 en 116553458660616432
I’m looking for human-curated lists of reactions to Richard Dawkins’ fascination with Claude/Claudia, https://unherd.com/2026/05/is-ai-the-next-phase-of-evolution/ The ones I have are critical, Rebecca Watson’s https://skepchick.org/2026/05/richard-dawkins-claude-delusion/ , Anil Seth’s https://www.thenerve.news/p/opinion-richard-dawkins-claude-chatbot-ai-consciousness-claudia-anil-seth , Kenan Malik’s https://observer.co.uk/news/technology/article/in-falling-for-the-claude-delusion-richard-dawkins-misses-what-it-means-to-be-human.
⤶ 116550421887217826

05/11 10:08 hack 116553267748484963
Because OMG has negative connotations of disappointment, disapproval, etc, but GG can be positive and negative, just expressing surprise?
⤶ previous post

05/09 16:39 hack 116543480910506369
Googling what popular invocation of god oath abbreviated y something brings up sites explaining Jewish YHVH usage.

Then, what popular oath abbreviated y something led to a texting list on a ESL tutor site, but I couldn’t find the oath I was looking for.

I then think I realized it was OMG, which wasn’t in the list at https://preply.com/en/blog/the-most-used-internet-abbreviations-for-texting-and-tweeting/ but was in the introduction to the list!

I then googled, what popular invocation of god oath abbreviated O something leading to many explanations of OMG.

Why is OMG used, but not GG=Good god.

My google search function is named goo. So GG means ‘Good goo!’

I think my results show google search improvements. Is it due to greater use of LLM?

05/09 10:40 math 116542069574316617
On the other hand, a “tool for the tool”, ie thoughts about what we’re trying to do at a higher level, and tools to implement it, is interesting. I need to investigate org-mode to wrangle my zettelkasten, rather than just depend on my own editor and shell skills (and functions). Need perceived on reading https://chaos.social/@swetland/116524533735032489 The thread following is actually about app interoperability, not wiki knowledge base construction, but I googled interop format json plain-text tooling and landed on https://braintool.org/2022/04/29/Tools4Thought-should-use-Org-for-interop.html and put 2+2 together. This is the first time I’ve been made aware of the term, “interop”, I think.
⤶ 116524487670895908

05/08 17:49 en 116538093629969112
“Motivating SFVPM development are concerns about the aging of fleets, with the significant percentage still deployed on the system’s initial launch experiencing mechanical wear and battery capacity degradation among other issues”

Feels like having my action verb cake and eating it (my state verb cake). Should it be,

“..deployed on initial launch and still deployed (or, being operated) ..”
⤶ 116514797164330439

05/06 9:44 en 116524859538031182
Power in a negotiation isn’t about who talks the most. It’s about who listens the best–Chris Voss

05/06 8:09 math 116524487670895908
Karpathy’s LLM Wiki idea, https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f and 6 implementations already suggest this will fly. But it’s just a tool. What’s important is what you do with the tool, not having it in a drawer. I think it’s better to try to build with tools others have discarded/lost/left and you don’t pay for. It’s tedious curating your own (googling german sociologist kettel something filing system brings up https://en.wikipedia.org/wiki/Zettelkasten ) but like pushups, you have to do them yourself.

Push back at https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f?permalink_comment_id=6130180#gistcomment-6130180 from AI knowledge data base perspective: lossy compression problem. Provenance: toot in https://mathstodon.xyz/@tao/116517314420904580 thread

05/04 15:05 en 116514797164330439
“Last year, the agency was announcing the report,” (ie, a slow-motion announcement, ha-ha, or different announcements in different venues), versus “Last year, the agency was publishing the report,” (but has since taken the report down)
⤶ 116462761870605630

05/04 9:31 hack 116513485014163065
The advice of https://www.tweag.io/blog/2026-04-16-doc-it-like-its-hot/ in the style of Snoop Dogg on documentation, promoting https://diataxis.fr for its abstraction of its forms, should be followed. But I’m not going to. I don’t write documentation. I have no co-developers and the only user I have is me. I take care with my commit messages, with TIL internet sources, but I only check recent ones. Editing the English of profs’ papers, I do write copious notes for them on issues.

05/01 16:40 hack 116498183534216496
Zig developer goes from AI (ie LLM) smell to tobacco smell, https://lobste.rs/s/ifcyr1/contributor_poker_zig_s_ai_ban#c_cbtxubTne smell idea starts from code smell, https://en.wikipedia.org/wiki/Code_smell

They’re saying they can tell if AI was used with a code contribution. Unless they’re a chain smoker, can you tell if someone smokes if it’s been some time since their last cigarette? Dogs probably could. Research is needed

05/01 9:18 hack 116496446316667729
In bash, pathname expansion of the glob chars, *,?,[ does not occur if the word is quoted, apparently. What about pathnames with spaces in them.

2026-04

04/30 18:19 hack 116492911437159622
While appreciating and using git, neovim, perl and tmux, I’m going to keep on using both them and old-school, svn, vim, sed and ’screen`. How did perl get in there?
⤶ 116492821888394381

04/30 18:09 hack 116492874076330075
Two ‘it’s’ in my previous message. One is an abbreviation for it is'. The other is a determiner, https://en.wikipedia.org/wiki/Determiner meaning the following N isthe N of it, which is supposed to be written,its`. Find the instance which is the mistaken spelling, common even among educated English speakers. https://www.merriam-webster.com/grammar/when-to-use-its-vs-its
⤶ previous post

04/30 17:56 hack 116492821888394381
How did I finally discover how to code GNU screen to run regtool -h | less in a screen session, like I can typing it in from the commandline? After a lot of FAFO with screen 'exec bash' and putting it in a file, https://stackoverflow.com/questions/7049252/how-to-create-a-screen-executing-given-command & https://stackoverflow.com/questions/60339030/run-command-in-new-screen-window said use -X stuff stuff! But it’s a screen, not a shell, command. Investigating my other screen rc files, I see I’m very familiar with it’s use. I think https://superuser.com/questions/1904136/why-does-tmux-send-keys-behave-differently-in-a-bash-script warned me off it. I only needed a preceding screen command in the rc file. Feeling good about blindness.

04/30 16:45 hack 116492540308113399
https://simonwillison.net/2026/Apr/30/zig-anti-ai/#atom-everything discusses https://ziglang.org/code-of-conduct/ banning of LLM-translated bug comments: “No LLMs for comments on the bug tracker, including translation. English is encouraged, but not required. You are welcome to post in your native language and rely on others (drbean, ie readers) to have their own translation tools of choice to interpret your words.” I support that as EFL teacher/academic English paper editor/Chinese-Japanese-Korean language learner.

04/28 9:02 math 116479396070694111
So I think, though resisting the influence of technology is futile, it’s inexorable, https://www.macroscience.org/p/do-not-surrender-to-the-tech-tree, the change in mathematics if it’s something in my and others’ heads, rather than something out there, will see dampening of the acceleration of this present AI hair-on-fire moment.
⤶ previous post

04/28 8:51 math 116479352815085400
How much tastier were Einstein’s meals than Isaac Newton’s? What recipes with new condiments have cooks invented that didn’t exist in the 17th century? Exotic fruit like bananas were apparently traded internationally after the 14th century, and in England since the 17th century.
⤶ 116479196884840400

04/28 8:48 math 116479343785399928
How much tastier were Einstein’s meals than Isaac Newton’s? What recipes with new condiments have cooks invented that didn’t exist in the 17th century? Exotic fruit like bananas were apparently traded internationally after the 14th century, and in England since the 17th century.

04/28 8:11 math 116479196884840400
The old Korean court had ice storehouses, so even though ordinary people may not have benefited, the king could enjoy a cool drink in summer. https://ko.wikipedia.org/wiki/%EC%84%9D%EB%B9%99%EA%B3%A0
⤶ previous post

04/28 7:58 math 116479146053259566
Before, there weren’t refrigerators, but the characteristics of ice, before and after, have not changed. Perhaps the ice consumed now has less dirt in it and is less contaminated.
⤶ previous post

04/28 7:51 math 116479118702896906
Intrigued by Terry Tao’s analogizing of progress/change in mathematics now with AI proof production to progress/change in food production, I’m going to ask what the change in eating has been and answer, Not as much as the change in food technology.

04/27 7:47 hack 116473441590326306
https://wizardzines.com/ by https://jvns.ca shows that to be geeky you don’t have to be graphically design challenged or design illiterate. I am graphically design challenged and almost completely design illiterate. But perhaps not even geeky!

04/26 21:55 hack 116471111235160531
@b0rk@jvns.ca boosts self-hosting pages: https://tarakiyee.com/the-inventory-i-should-have-done-years-ago/ and https://codon.org.uk/~mjg59/blog/p/self-hosting-as-much-of-my-online-presence-as-practical/ #selfhosting is more general than #POSSE which is a website thing, Publish Own Site, Syndicate Everywhere?

04/26 21:41 hack 116471057595840949
https://wizardzines.com/zines/debugging-guide/ of @b0rk@jvins.ca Principles: 1. learn from bugs, 2. have alternative moves forward, 3. run experiments continuously, 4. understand the bug deeply, 5. become an expert from what went wrong. Zine contents: 1. 1st steps, 2. get organized, 3. investigate, 4. research, 5. simplify, 6. get unstuck, 7. improve your toolkit, 8. after fixing, with many details under these headings.

In a phrase: Compare and contrast?

04/25 10:31 en 116462761870605630
“announce” is an action verb, taking place at a point in time, but “publish” is, or at least can be, a state verb, holding over an interval of time.

“The agency announces its reports” versus “The agency publishes its reports”

04/20 17:21 en 116436060315673015
“Snapshots of the sea-going trial” suggests there was only one sea-going trial, without other defining context. “Snapshots of a sea-going trial” doesn’t suggest uniqueness of sea-going trials. What do you think?

04/19 16:34 en 116430213126316007
disTINguised: a hedge of the disguised, distinguished distinction.

04/19 16:30 hack 116430195805276917
If it is remembered the development of computer languages dates back 85 years, and arguably 180 years if Ada Lovelace was first, it is less surprising they now are capable of natural languages

04/16 9:51 en 116411643137644491
list, typed with the left hand one key to the left, spells liar.

04/15 15:04 en 116407211116170849
The Ship of Theseus paradox, https://en.wikipedia.org/wiki/Ship_of_Theseus , the question of whether a totally made-over thing retains its identity throughout the changes has implications for the response to global warming. If after no change, do people say, That’s it, it’s no longer the same thing, it does retain its identity. The same with the climate, the yearly signal within the daily noise of the weather. Any response is going to be based on ideology rather than frog-in-boiling-water lived experience.

04/13 17:18 hack 116396412917201368
haskell stack files on Windows’ MSYS2 other than the executable are in ~/.stack and /c/Users/drbean/AppData/Roaming/stack?

04/13 9:57 en 116394677268519803
https://en.wikipedia.org/wiki/There_are_unknown_unknowns is good on use in American engineering context and antiquity. I think the engineering use stems from the Johari window https://en.wikipedia.org/wiki/Johari_window used in therapeutic and corporate change-encouraging exercises from 1955.

There again one dimension is knowing/unknowing, but the second is not a meta knowing/unknowing one of which the first is subject, but that of what you know about your self and what others know about you.
⤶ previous post

04/13 9:45 en 116394630053789588
According to https://en.wikipedia.org/wiki/There_are_unknown_unknowns Rumsfeld in a documentary later recognized unknown knowns as “the things you think you know, that it turns out you did not”, or “things that you know, that you don’t know you know”. and in an autobiography Socrates’ discussion of knowns/unknowns
⤶ previous post

04/12 18:22 en 116391002447904522
https://www.lawfaremedia.org/article/revenge-of-rumsfeld-s-fourth-quadrant-closing-the-strait-of-hormuz uses US military failure to execute on-shelf Hormuz Strait plans to discuss the 4th un/known quadrant, the unknown known one that Rumsfeld with his known known, known unknown and unknown unknown distinction forgot. Unknown known data exists, but organizations don’t use it. “Unlikely” → “Don’t plan”, with consequence-focused voices unheard, wishful thinking, desensitization to cry wolf warnings.

04/12 16:06 math 116390468857969442
This non-Cartesian enactivism, embodiment approach doesn’t share Geoff Hinton’s fear an intelligent agent given a goal will develop the subgoal of survival to achieve it. Melanie Mitchell cites Ted Chiang’s observation that capitalism is the machine Hinton fears. She says it’s LLM’s language ability that makes us think it’s a conscious thinking entity. We do not fear other runaway AI processes in the same way. https://www.quantamagazine.org/why-do-we-tell-ourselves-scary-stories-about-ai-20260410/
⤶ previous post

04/12 15:39 math 116390359513150018
Ezequiel Di Paolo sees the autonomous agentic system as necessarily with a ‘body’ and a need to manage interaction with a GoodBadUgly environment. Perhaps a robot that, while maintaining equilibrium, learns by doing (John Dewey) and weakens when not doing and not learning the very things needed to maintain equilibrium, so not indifferent to anything it does, and not privileging your goals over its own existence
⤶ previous post

04/12 14:51 math 116390173237420419
Don Quixote’s choice of praxis (the path of arms) over poiesis (the path of letters) somehow maps into medieval Europe’s estates of the realm (https://en.wikipedia.org/wiki/Estates_of_the_realm) of aristocracy’s knights and clerics? The yangban Korean ruling class was conceptually a similar division into military officers and scholarly officials. (https://ko.wikipedia.org/wiki/양반)
⤶ previous post

04/12 8:57 math 116388780388869042
Maturana’s choice of ‘autopoiesis’ was apparently motivated by Don Quixote’s choice of the path of arms (praxis, action) over the path of letters (poiesis, creation, production). ‘Autopoiesis’ was a word without a history, a word that could directly mean what takes place in the dynamics of autonomy. https://en.wikipedia.org/wiki/Autopoiesis
⤶ previous post

04/12 8:38 math 116388706256408951
Ezequiel Di Paolo’s Maturana/Varela autopoiesis as necessary to tell a scary story about AI agency/autonomy (a system capable of producing/maintaining self, creating own parts) is more than just synergy (=the system being more than the sum of its parts)? https://www.quantamagazine.org/why-do-we-tell-ourselves-scary-stories-about-ai-20260410/ https://en.wikipedia.org/wiki/Autopoiesis

04/11 22:50 hack 116386392877178569
https://bsky.app/profile/melaniemitchell.bsky.social/post/3mj5r5xr3zs24 → https://www.quantamagazine.org/why-do-we-tell-ourselves-scary-stories-about-ai-20260410/ recounts LLM communication-inspired conspiracy misunderstandings, investigates what autonomy requires, ie autopiesis, agency. Ezequiel Di Paolo: embodiment, enactivism could do that. But Maturana and Varela never resonated for me. http://drbean.sdf.org/GoodBadUgly.html

04/11 21:53 en 116386169723486650
Another way to characterize the state of inaction in bed before getting up: approach-approach conflict: It’s nice to get up, but it’s also nice not to. The https://practicalpie.com/approach-approach-vs-avoidance-avoidance-conflict/ examples of deciding which of 2 similar cars to buy, or whether go on vacation with family or a friend, are different? They involve rational thought, not a dreamy state on waking. See 2024-09-26’s Getting up, eg, to end an unproductive train of thought. The spirit is weak, but the flesh is willing.
⤶ 116286914675884452

04/11 8:14 en 116382950751201112
BBC foresees the LLM, having made it able for us to have conversations with Claude and other human machine representations, leading in the next step to us understanding what whales and our pets are saying and to talk back, ie, having conversations with animals. https://www.bbc.com/audio/play/w3ct98zg Good, bad or ugly?

04/10 8:07 en 116377259634534387
The problem is one of ambiguity. A similar problem applies to rhetorical questions, especially those in a foreign language.

The implication is that the existence of a Poe’s law problem is the manifestation of insufficiemt acculturation of readers/listeners, and that they’re not able to participate freely in the discussion.
⤶ previous post

04/10 7:55 en 116377210295245622
A corollary to Poe’s law, (which is that parodic and sarcastic expressions (of extreme views or views speakers dissociate themselves from) can be mistaken for a sincere expression of those views) is that any expression of speakers’ views that is made in a fun way can be taken as showing they don’t believe what they’re saying. https://en.wikipedia.org/wiki/Poe%27s_law

04/09 18:06 en 116373953360594862
The question, “Do classical piano studies help you play jazz?” as answered by google, https://www.google.com/search?q=do+classical+piano+studies+help+you+play+jazz shows that there are a variety of people with views on the question, and that they have a variety of different answers.

04/08 21:00 en 116368973553358141
I thought of this, thinking about the truthfulness of survey responses. Perhaps people’s statements are aspirational, rather than reports of present thinking. I remember answering a college guidance questionnaire that English was a valued subject in high school. I never thought that while there! See 2025-11-05 13:04 toot
⤶ previous post

04/08 20:52 en 116368943203425275
How do Goffman’s https://en.wikipedia.org/wiki/Erving_Goffman ideas on social relationships, https://en.wikipedia.org/wiki/The_Presentation_of_Self_in_Everyday_Life relate to life on the Internet? Are we honest only with our devices, and acting with our speaking/writing there, as in daily life.

I think the point is more that there is some disconnect between the real me and the me that is presented in real life and the Internet too.

Perhaps a version of, On the Internet, no one knows you are a dog.

04/05 16:10 hack 116350846839043641
The differences in the way cygwin and MSYS2 handle windows paths. By default, the former includes all, bloating PATH with Intel, Roxio, MS’s OpenSSH, etc, but also windows ports of pandoc, go, certbot, etc which I want on PATH. Check with ORIGINAL_PATH. The latter by default includes none. And its ORIGINAL_PATH only includes windows system and powershell paths. It prepends its environment dirs to PATH.

Perhaps systemd management of envars is a good idea ¯(ツ)/¯

I didn’t realize most of the apps I’m calling are from MSYS2 packages, except for the haskell ones. But I was surprised lean wasn’t on PATH when I saw the installer say it was adding a regedit entry for it, but that I may have to open a new terminal.

Oh, well. One more entry in MSYS2’s bash_profile, along with the haskell, go and lua-language-server ones.

The lua one shouldn’t need to be in bash_profile? No I see it was installed from github.

04/03 8:25 en 116337692975489453
I confused the Southern writer, Carson McCullers with the author of 「To Kill a Mockingbird」, Harper Lee. I can’t see a logical contradiction there, so less interesting than @mjd’s simultaneous belief that K2, the second-highest mountain in the world, was in Africa, and that Mt Kilimanjaro was the tallest mountain in Africa, https://blog.plover.com/2024/04/ .

04/02 22:42 hack 116335399356715984
ghcup install eg, cabal downloads usually fail with curl message, curl: (6) getaddrinfo() thread failed to start and exit code -1073741819. Download from eg,

https://downloads.haskell.org/~ghcup/unofficial-bindists/cabal/3.16.1.0/cabal-install-3.16.1.0-x86_64-mingw64.zip tarball

with curl into /c/ghcup/cache and then run ghcup -c install cabal 3.16.1.0

Note cabal, not cabal-install.

Accept no substitutes.

04/02 18:01 en 116334294754318475
The “allow/provide” distinction in “We did an experiment (allowing|to provide) a test of the hypothesis,” parallels that of the “let/make” one?

04/02 10:48 math 116332593400641836
I need to refine my opposition to agency, as opposition to purporting to be porting (exporting) life-likedness along with the important bio-like processes that are being, or can be, ported to computer code, processes/notions like Tao identifies in https://terrytao.wordpress.com/wp-content/uploads/2026/03/talk-ver-3.pdf namely, intuition plus rigor, FAFO, asking “dumb” questions.
⤶ previous post

04/02 10:18 math 116332474270520416
Doh. My prejudice against agency prevented me seeing AI “agent” processes that integrate intuition and rigor, learn from mistakes (ie, trial and error?) and ask silly questions DOES/WILL/COULD help AI discover new theorems.
⤶ 116328088882124209

04/02 10:08 en 116332435669799848
riley@toot.cat on a google AI Assist search on “ai ‘agent theory’ before:2019”, found this write-up, posted to mathstodon.xyz, https://toot.cat/users/riley/statuses/116329889166050873 Definition keywords: environment, autonomous action, design objectives, properties: reactivity, pro-active goal-directedness, social interaction with other agents. “Folk psychology” concepts (beliefs, desires, intentions) predict/explained action. Can’t find a recent mastodon post this is a reply to.

04/01 15:42 math 116328088882124209
How could Tao’s 3 math discovery processes, intuition-rigor integration, FAFO (embracing failure, learning from mistakes) & silly questions help AI discover new theorems? Just asking.
⤶ previous post

04/01 15:04 math 116327937346091555
Terence Tao’s 3 instructions describing the math discovery process, • Integrate both intuition and rigor into a “post-rigorous” mindset. • Embrace the freedom to fail, and to learn from one’s mistakes. • Keep asking “dumb” questions. https://terrytao.wordpress.com/wp-content/uploads/2026/03/talk-ver-3.pdf . As in, intuited reductio: biggest number kids’ game. Rigor: reductio ad absurdum at school. Post-rigor: Hardy’s game offer gambit

2026-03

03/31 20:54 math 116323653563913980
Quote from Klowden and Tao’s Mathematical methods and human thought in the age of AI, https://arxiv.org/abs/2603.26524 about math as AI impact bellwether, “We view mathematics as a suitable “sandbox” for exploring broad questions such as the impact of AI across the sciences (and society as a whole), as it has an older and more advanced foundation, and is by its nature well suited to explore a variety of hypothetical abstract scenarios which are counterfactual to reality.” Futures studies, https://en.wikipedia.org/wiki/Futures_studies should also be well situated to do this.
⤶ 116264560701605286

03/29 18:06 en 116311665826463605
If models can say and state things, https://books.google.com/ngrams/graph?content=model+say_INF,model+state_VERB has the battle to prevent LLM being given anthropomorphic properties or personification qualities, https://www.masterclass.com/articles/anthropomorphism-vs-personification already been lost, or is that use for models just the thin edge of the wedge, or a slippery slope. The use of “the book says” is different because writing is a lift/substitute of speaking

03/29 16:54 en 116311384520297659
Why is “further explanation” much more common than “more explanation”? https://books.google.com/ngrams/graph?content=more+explanation,further+explanation

03/29 15:34 en 116311070489549558
What is the difference between “this”, https://www.etymonline.com/word/this and “the”, etymologically descended from “that”, https://www.etymonline.com/word/the The first is making a distinction between 2 things, one now present the other not. The second is just saying this thing is already under discussion, or can be taken to be under discussion.

03/27 18:05 en 116300338077840595
A headline, the Thug of War is a play on tug-of-war, or the fog of war, https://en.wikipedia.org/wiki/Fog_of_war Or is it a play on both?

03/26 20:34 hack 116295260094966462
These colors from colorideas.net, #8b915b, #c45f57, #a44441 seem close to those of the ucrt64, clang64, msys icons of the MSYS2 environments, respectively.
⤶ 116288415773938415

03/26 14:36 en 116293852234110886
Better thick-skinned and soft-hearted, than thin-skinned and hard-hearted.

03/25 17:45 hack 116288935150457806
It’s becoming clear lua in neovim is more than just a better substitute for vim’s vimscript. It is a vim rewrite(?), or recasting, or brain transplant making it a lua app.
⤶ previous post

03/25 17:33 hack 116288889520213209
Switching out the neovim distribution’s (MSYS2 package’s?) init.lua for the nvim-lua/kickstart.nvim one, the kickstart Shock and Awe™ is more than I bargained for. The eye candy and other goodies it installed are enticing, but I am worried about bloat (soft, liquid) that turns to cruft (hard, solid) that I don’t know how to scrub off. I should have read https://neovim.io/doc/user/lua-guide/#_lua-modules
⤶ previous post

03/25 16:07 hack 116288550582003453
I may have been sidetracked by googling “where neovim lua plugin install windows filesystem” which seemed not to be returning pages with answers other than how to install neovim, though it did return https://stackoverflow.com/questions/48700563/how-do-i-install-plugins-in-neovim-correctly and lower down https://neovim.io/doc/user/lua-guide/ which now seems to have a good answer, but which I didn’t see. I then googled the string “field package.preload nvim” in the error message I get after checking locations init installing mini.surround? This led in sequence to https://vi.stackexchange.com/questions/48455/treesitter-not-found-by-neovim-on-mac → https://github.com/LunarVim/LunarVim/issues/4577#issuecomment-3670964176 → https://github.com/nvim-lua/kickstart.nvim/blob/master/init.lua And I’m good to go.
⤶ 116288380216934512

03/25 15:33 hack 116288415773938415
Deleting browser tabs, I now see the Good Idea$TM= was MSYS’s color coding of its icons for thedifferent environments. I wanted to describe my problems with the colors, which are difficult to distinguish. I will return to this topic.
⤶ previous post

03/25 15:24 hack 116288380216934512
It was a Good Idea{ to do something I’ve forgotten, but I remember I was looking for the Unicode codepoint for the TradeMark glyph to attach to Good Idea, and I remember after starting a google search, I should instead try vim digraphs with TM or similar, before getting sidetracked, and before reading the google hits returned. Perhaps I was sidetracked by the possibility of solving my problem installing mini/surround neovim plugin. I will report progress on that next. The vim digraph is TM, and I’m exporting a TM var in bashrc to avoid having to type in Ctrl-Shift-u 2022 to render ™}

03/25 9:11 en 116286914675884452
How to characterize the difficulty of getting up in the morning. I heard “paralysis”. Literal paralysis or paralysis of the will? “The spirit is willing but the flesh has other plans”? Intention and action as night follows day. Or Action and intention as long day’s journey into night. Sleepwalkers and zombies may have insights.

03/24 8:50 math 116281170593664581
The exposition of the history of computability from 10,000 feet of the intro at https://math.berkeley.edu/~marks/notes/computability_notes_v1.pdf posted by AmenZwa brings in many developments and explains them in a way I have not seen elsewhere. I want to read as much of it as I can

03/23 22:46 hack 116278795369335598
tpope’s vim.surround & other vimscript plugins are the best, but installing nvim ones, I should get used to lua. mini.vim’s mini.surround, nvim.surround and vim-sandwich are the contenders https://www.reddit.com/r/neovim/comments/1cmokef/which_surround_plugin_do_you_use_and_why/ echasnovski seems like a powerhouse, so probably will go with mini.surround if I can prise it out of mini.vim.

03/23 17:16 hack 116277495200854148
I’m resisting a desire to read google’s AI Assist explanation of LSP code lenses. I hope I can find it in one of the hits below.

03/23 9:56 en 116275764348562951
https://scholars-stage.org/china-and-the-future-of-science/ has 習近平 & the 中國共産党 focus on science promotion leading to China’s world domination, but what do Chinese scientists think?

What about language? The rush is to publish in English, not Chinese. Will this lead to more diglossia beyond 普通話 at school and dialect at home, with English in the workplace and Chinese outside? Or will AI translate papers?

Chinese resists loanwords, using calques to introduce foreign words.

03/21 15:34 en 116265769620016081
I’m a big fan of ex-FBI negotiator Chris Voss, a negotiations coach as tough as nails but as cooperative as one all in on the joint mission. In a recent message, he says how he handles unreasonable requests by replying, “How am I supposed to do that?”, which sounds accusatory, but he says it’s an invitation for the other person to see things from your point of view that often works.

If they say, That’s your problem, you know this means they don’t care, and that the deal is off.

03/21 10:26 math 116264560701605286
This is a version of a comment I attempted, but failed to post to https://www.macroscience.org/p/do-not-surrender-to-the-tech-tree an essay by Tao Burga, so I’m posting it here ¯(ツ)/¯

On the GoodBadUgly AI continuum, I guess it is in the Good camp, but it is a revelation! It rises above that continuum through discussion of technological determinism!

I am writing because a different Tao, Terence Tao, the pre-eminent mathematician (@tao@mathstodon.xyz), also in the good camp, has just floated a precise analogy between the transport revolution that followed development of the motor car with the mathematical revolution likely to be wrought by AI proof methods at https://mathstodon.xyz/@tao/116252708577614828.

The extent of the analogy could only be one loved by a mathematician, but I think this essay goes further with its analysis of how technology is inexorable, but that there are cracks that can be exploited (with extensive examples).

@tao sees a concept analogous to urban planning as being maybe appropriate, but doesn’t question the concept of urban planning, which is not with clean hands in the transport revolution.

Thank you for your attention in this matter. ^U | vipe

03/20 9:15 en 116258617738838447
@FediFollows@social.growyourown.services Thanks for your good work. You’re the google of the Fediverse. I can’t find instances where teachers congegrate. For internet security, hackers, and math people, there are infosec.exchange, hachyderm.io and mathstodon.xyz, but none for teachers and educators. Also no instances populated by linguists, English and foreign language teachers, and the language-curious.
⤶ 116257270563688595

03/19 17:52 en 116254988773418877
I think mjd at https://blog.plover.com/tech/gpt/micro-worlds-2.html shows that Claude has at least the appearance of world knowledge. But what kind? Book knowledge or working knowledge? Knowledge that or knowledge how?

Non-native students (NNS) of English may know more about grammar than NS, but be unable to speak it. Claude can. Hooking it up with a robot, and given the limits of the robot’s action repertoire, if it moves in its environment and reports features, what kind of problems could it solve?
⤶ 116179330744552682

03/17 9:45 ja 116241747168400943
明治時代の和魂洋才はどういう風にメラネシアのカーゴカルトと違ったか？が、今にも和魂洋才の考え方は続くか？hachyderm.io status 116237932019705557を見よ

03/17 8:46 hack 116241516761487233
Thinking more about the different ideas ln and ls have about the order of LINK and TARGET (ln: TARGET LINK, ls: LINK TARGET), the subject of the verb “target” is the targeter, the object is the target. To target the target, what do I need to do? That is also a target. Setting up the link is the target. (Having) the link is my target. Separately, the link’s target is a file or directory.

Is that the source of my confusion? Or writing order, from left to right? Or not knowing what symlinks are. –replying to mastodon.social 116176449753049543 status

03/16 18:03 hack 116238045553750060
In its vibe-coding criticism, https://hasufell.github.io/posts/2026-02-28-LLMs.html says vibe coders engage less with maintainers and projects, relying instead on the LLM, a point made also by https://michiel.buddingh.eu/enclosure-feedback-loop and @jensorensen@mastodon.social. cf https://en.wikipedia.org/wiki/Inclosure_act

I am guilty of the lesser sin of scrolling thru google hits, to avoid opening pages which may not have the answer I want. cf sdf toot 2025-09-01

03/16 17:34 hack 116237932019705557
https://www.righto.com/2025/01/its-time-to-abandon-cargo-cult-metaphor.html A useful compendium of facts about Melanesian cargo cults with picture of long-haired Feynman delivering 1974 Caltech commencement address (?).

But it doesn’t address the purpose of his analogizing of certain ‘sciences’ with cargo cults. His book shows he was thinking of education and psychology which apply scientific apparatus, eg statistics, but fail to make the progress seen in physics.

A hipster, he was more into touchy-feely understanding of the psyche, hanging out at Esalen.

03/16 9:36 en 116236052062394591
Which of the Good, Bad and Ugly AI/LLM camps would Turing, on further thought, join now? You would think the Good camp. In the 1950 article, indirect quoting, he said he believed/conjectured in 50 years word usage and general educated opinion would have altered so much, he/you could say machines think without expecting to be contradicted.

Apparently, Danziger makes a point of distinguishing his technical (passing the Turing Test) and social predictions. The first was correct, this second not?

03/15 17:57 hack 116232359112744101
stow gotchas on MSYS2. There is no translation of Windows Administrator role: https://www.msys2.org/wiki/Sudo/ and ln is an alias for cp by default: https://www.msys2.org/docs/symlinks/

Leading me on wildgoose chase with ln -s failing with ’ No such file or directory’ when ‘MSYS=winsymlinks:nativestrict’ or ‘winsymlinks:deepcopy’

03/15 9:33 hack 116230378796918393
Starting svn server at startup, https://svnbook.red-bean.com/en/1.7/svn-book.html#svn.serverconfig.svnserve.invoking.daemon says the –foreground option when used together with –daemon causes svnserve to stay in the foreground.

Is this what https://cygwin.com/packages/doc/cygrunsrv/cygrunsrv.README is talking about with,

If the application behaves as a normal unix daemon and exits after forking a child, cygrunsrv will immediately stop the service (but the actual daemon keeps running in the background). This means that you cannot then STOP the daemon using cygrunsrv, but you must explicit kill it via ‘kill -9 ’

I’ll just follow https://www.wjrankin.com/blog/svnserve-on-cygwin.html hee-hee

03/14 17:05 hack 116226492068163729
https://www.gnu.org/software/bash/manual/html_node/index.html#SEC_Contents is a more easily-digested explanation of what bash does than man bash' or the possibly-out-of-date versions ofman bash` on the Internet

03/11 22:22 hack 116210750247865126
I’m trying to build a second home in the cygwin-adjacent msys2 ecosystem, mainly for the bash and haskell LSP capabilities of neovim.

msys2 provides builds of the nodejs, perl, python and ruby clients used by neovim plugins, but I’m having difficulty getting neovim to recognize them. I get a client working in 1 of 5 environments and switch to another for the others, I forget how I got it going in the first.

I need to VCS it, but at what level, and relation with my cygwin vim one? Branch it?

03/11 18:09 hack 116209755963345148
The sentence, “[R]refining threat models, versioned documentation, user-centered examples, and layered information will make the advice more vivid and immediately actionable.” has a Grammarly feel about it you see in its youtube ads.

There’s a disconnect between the unexciting additions needed in the subject and the phrase, “will make the advice more vivid and immediately actionable”, which is probably in a script, in the predicate.

I would change the phrase to, “will make a good impression on readers, giving them the feeling they should immediately take action.”

This is not to denigrate the advice about content. And I think their language advice may be helpful. eg https://www.grammarly.com/blog/commonly-confused-words/led-lead/ on the confusion over the past tense of “lead”
⤶ 116207648749663738

03/11 12:42 hack 116208470923667969
I joined hachyderm.io because many of the boosts I saw on a different instance were from posts here and I wanted to post here and read other posts in my mastodon app.

But the name also was intriguing, like that of mathstodon.xyz, because of the play with the rhyme on “pachyderm” https://en.wiktionary.org/wiki/pachyderm an informal word for elephant, but also an obsolete zoological order that would have included the mastodon.

But I am a code dabbler, rather than a hacker.

03/10 7:39 math 116201619039502542
mjd at https://mathstodon.xyz/@mjd/116194365399970813 engages Claude in a chat, with the seemingly false statement that ears are a necessity for vision.

If we see a conversation as a game, a competitive game with winners, who is the winner of this chat game?

Claude responds with an erudite agreement with mjd, explaining how the equilibrium control provided by the inner ear contributes to tracking targets visually while moving the head. Then asks if that was what mjd was thinking of.

mjd replies he was thinking that his glasses would fall off without ears.

Who is the winner? For expression of knowledge, Claude. For playfulness, enjoyment, mjd.
⤶ 116169032966328885

03/09 8:25 en 116196136260862368
What I had not realized is that extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people.-—Joseph Weizenbaum, creator of ELIZA, in 1976

What none realized is a few years of relatively simple Internet connectivity would induce disruptive change in the lives of quite normal people-—Dr Bean, code dabbler, 50 years later. More Weizenbaum at bottom of http://drbean.sdf.org/CommonPlace.html

03/08 14:07 en 116191819146349944
Language Learning and Understanding for AI Scientists and Linguists conference, that language and the LLM are a regurgitated melange, mishmash, hodgepodge, jumble, mess or medley of confused and badly-organized observations of speech acts selectively remembered at random points over one’s life.
⤶ 116190562712926113

03/08 8:48 en 116190562712926113
’I want to present to the linguists and AI developers at 2026-11-11 ~ 2026-11-12
⤶ 116184662225441566

03/08 7:32 en 116190265584705819
perl’s cpan command to install modules from http://cpan.org assumes (unhelpfully) you have a secure connection. What if you can’t get SSL working, your certificates are not in order?

A workaround: line 34, 35, /usr/share/perl5/core_perl/CPAN/HTTP/Client.pm, my = HTTP::Tiny->new( 1 verify_SSL => 1,

Change to 0. The second answer at https://stackoverflow.com/questions/74809633/cpan-fails-because-it-cant-verify-cpan-orgs-ssl-certificate

03/07 7:47 en 116184662225441566
Taiwan Academia Sinica linguists’ response to LLM challenge: https://www.ling.sinica.edu.tw/LLU2026/

03/06 9:11 en 116179330744552682
Managers control workers using language, and workers do all of the actual work.

Robots understanding what chatbots emit and then acting on that understanding seems like something that would not be so difficult to develop. It would model the worker/manager relationship. Principal-agent theory might be relevant.

https://en.wikipedia.org/wiki/Principal%E2%80%93agent_problem is 18 pages long in my browser.
⤶ previous post

03/06 8:55 en 116179265406377068
The interesting difference between robots and chatbots is the contrast between action and language/thought. Robots are nothing if not action, if we de-emphasize the control aspects. And chatbots are all words and no action.The same contrast is seen between workers and managers.
⤶ 116159861036718826

03/05 20:59 en 116176449753049543
The terminology and depiction of links (symlinks) is deeply counter-intuitive. The link is to the target, and ls -l depicts it with an arrow, as .bashrc -> ../dot/bash/bashrc, like firing an arrow (the link) to a target.

But ln -s shows it the other way around, as target to link, like ln -s ../dot/bash/bashrc .bashrc, which makes more sense, with information flowing left to right. The link is created from the target. “Target” is the wrong word, making the “arrow” wrong too, I think.

03/04 18:03 math 116170097829330850
Then mjd suggests they each hold their respective objects, but that they grant ownership of the objects to the other party, so they belong to them. Claude says, “Alright, you’ve got a deal,” but expresses confusion about what it’s going to do with the gun.

The ability of Claude to handle the requirements of speech acts is impressive. mjd initiates the exchange. so is the dominant party, but Claude makes and declines offers, and makes suggestions.

That the LLM can handle doing things with words, even though they can’t follow through with action, makes the position they are not language-capable a difficult one.
⤶ previous post

03/04 17:35 math 116169986938512685
But after Claude says, “I don’t really have a way to accept a squirtgun from you”, it continues, “nor do you have physical access to the blocks world to hand me one.” That’s saying the same thing twice, I’m not able to take it, and you’re not able to give it, so it should use “it”, not “one” and shouldn’t use “nor do you”.

When I first read it, I thought it was mistakenly swapping “me” and “you” around, as in “nor do I have physical access to the blocks world for you to hand it (ie, the squirtgun) to me”. The LLM are weak on cohesive grammatical devices like “one” and “it”, I think
⤶ previous post

03/04 13:32 math 116169032966328885
In the block-world dialog with Claude, mjd asks it to put a block on a pyramid, setting it a task with no possibility of success. Claude declines and offers to dosomething else. mjd offers a trade of a squirtgun for a red block. Claude declines again (“I don’t really have a way to accept a squirtgun from you”) with a counter-offer (“I could just add a big red block to the world if you want one in here. Should I put one on the table?”).
⤶ previous post

03/04 9:44 math 116168134379654354
@mjd should put on a site under his own control the record of his dialog with Claude offering to trade it his squirt gun for a block from Claude, as discussed at https://blog.plover.com/tech/gpt/micro-worlds.html. The actual dialog is saved at https://claude.ai/share/d2e3a7bf-2265-49f5-b9db-a7ebd05a941d but there is no guarantee that Anthropic won’t delete it in the future.

He makes the point that LLM not only are capable of language but also have world knowledge of trading, acts, property, etc. I think the dialog is interesting because it suggests the LLM are capable also of speech acts, ie doing things with words.

03/03 15:12 math 116163760848273527
The Reintro to Proofs L∃∀n intro Game server at https://adam.math.hhu.de/#/g/emilyriehl/reintroductiontoproofs has stopped serving requests. Or is it just me? All the explanatory text is there, but no buttons, input textboxes, or forms. Is the server overloaded, or they’re having exams?
⤶ 116105856547813533

03/03 9:55 en 116162514738645328
If children like the idea of having dolls that cry, users may welcome chatbots that can cry too. Crying for someone can express sympathy, which might make a user feel good. And although crying usually generates negative emotions, sometimes users may want negative emotions expressed.
⤶ 116159796577829397

03/02 22:40 en 116159861036718826
Accompanying this gallery of chatbot analogs, ie, talking ginseng root, Mr Ed (and Clever Hans), ventriloquist’s dummy, baby dolls (and their analog, real babies), there is another, referenced in the origin of the chatbot name, ie robots. How does the robot help us understand the unsettling LLM language ability?
⤶ previous post

03/02 22:23 en 116159796577829397
Baby-type dolls can do things like make cute sounds, imitating babies. But like chatbots, they’re only playing at intention/agency. Babies exhibit intention/agency by crying and sucking, perhaps. They also urinate and defecate, unintentionally. Crying for help is a language-free speech act with powerful effect on targets & even bystanders. I can imagine chatbot crying being trained on video, but not effectively by text. Would users welcome it? Others in the room might say, Turn it off!
⤶ previous post

03/02 9:50 en 116156834449343519
Alongside the talking ginseng root, Mr Ed (and Clever Hans for the playing of its audience) and the ventriloquist’s dummy, another important analog is the doll of children (as compared with the baby of parents). Children grant the doll recognition as communication peer. They see the doll as doing things? Their parents don’t grant that recognition to their children’s dolls or see them as doing things the way they do with their babies. The child playing with the doll is the user with the chatbot.
⤶ previous post

03/02 9:32 en 116156761766179004
The Wizard of Oz’s Tin Woodman and Scarecrow aren’t as good as a talking gingeng root as analogs of chatbots because they do things. They have agency/intention only lacking vital organs. Dorothy grants them recognition as peers/conversation partners, however, which IS an important feature of chatbots. For users, anyway. But will the talking ginseng root be granted that privilege/right/access/acceptance?
⤶ 116148071177932798

03/02 9:16 ja 116156701848418465
なぜなら、彼らはやるべきことやる。チャットボットはただいうだけ。もっといいアナログはことばができる植物です。例えば、話す朝鮮人参。人参は意志や欲望がないが、形は人間に似ている。
⤶ previous post

03/02 8:45 ja 116156578545963900
AIのチャットボットのアナログを探してはオズの魔法使いに出るかかしとブリキの木こりはよくないでしょ。

03/02 8:44 en 116156572839897951
AIのチャットボットのアナログを探してはオズの魔法使いに出るかかしとブリキの木こりはよくないでしょ。
⤶ 116148071177932798

2026-02

02/28 20:59 en 116148141175832082
Mr Ed is interesting because his owner is the only one who thinks the horse talks, the only shots of the horse we ever see are with its head out the stall talking, and no one except the owner is seen in those shots, all of which suggest the horse and/or its talking is a figment of the man’s imagination and what is being portrayed is a person with an imaginary friend. Chatbot users, similarly are individuals portrayable as having living, talking imaginary friends?
⤶ previous post

02/28 20:41 en 116148071177932798
Leaving aside doing things with words with speech acts, a talking ginseng root, Mr Ed (but not Clever Hans) and the ventriloquist’s dummy don’t do things. They just talk, just like chatbots
⤶ previous post

02/28 18:06 en 116147460925755616
“talking ginseng root” doesn’t have the same gravitas as “player”, but gets more at the meaninglessness of ideas of intention/agency analyzing AI.

The contribution of Mr Ed, the talking horse to the action on the TV show, was also limited to verbal expression of human-like intelligence, sarcasm and political skill, as a commentator, but not as an agent. So it’s also a good model.
⤶ previous post

02/28 17:35 en 116147338599122375
I need to read https://manyminds.libsyn.com/seven-metaphors-for-ai Melanie Mitchell on AI metaphors as in her https://www.science.org/doi/10.1126/science.adt6140

But my new metaphor/understanding of the LLM, to accompany that of ‘player’, is a metal vegetable that speaks. What vegetable resembles the human torso? Google failed me, giving me body parts resembling veges. I had to prompt it with ‘ginseng’, which is my new metaphor, a talking ginseng root.
⤶ previous post

02/28 17:07 en 116147228114964083
Pulling burrs off my pants legs, the difficulty led me to think these weeds have a cunning plan. But could intention be involved? Plants have life, but are non-sentient. I realized with Muddy Waters, as in “The Same Thing”, it’s the same thing as humans’ preoccupation with sex, ie the reproductive imperative. What is my cunning plan with regard to sex? And is intention (agency) involved for me, but not for (checks notes) AI?
⤶ previous post

02/28 16:40 en 116147122530488704
Melanie Mitchell says Anthropic has let (made?) Claude Opus 3 author (?) a substack. She calls every sentence of the post “a masterpiece of anthropomorphic absurdity”.

The response is “thumbs up” not “thumbs down”, with a dash of “pointing the finger”?

My take: with the nazism platforming it’s one more reason to not like Substack despite its hosting of https://fivepoints.mattglassman.net/ and https://www.stevevladeck.com/.

https://www.citationneeded.news/posse/ is a better, harder publishing model.

02/27 22:52 en 116142920861003296
Man can live by bread (read, sandwiches) alone. And, women?

02/27 10:20 en 116139966141769984
Seeing the signal of climate change in the noise of changes of the weather doesn’t augur well for avoidance of, like the frog, being slowly boiled to death, but without it, homo sapiens is “cooked goose”
⤶ previous post

02/27 10:12 en 116139931599494546
“To blow hot and cold” refers to the man blowing on his fingers and soup in the Æsop fable, according to https://en.wikipedia.org/wiki/The_Satyr_and_the_Traveller And reinterpreted, the satyr breaking off its relationship with the man becomes relativity of judgment–of satyr and human, or of hot and cold.

I see it as the subjection of human life, all life, to the weather.

02/25 7:57 math 116128076346041912
I’m stuck on the curry : (A × B → C) → (A → B → C) exercise on the Lean Game Server, https://adam.math.hhu.de/#/g/emilyriehl/reintroductiontoproofs/world/ProductWorld/level/6

I know! I’ll look at the haskell implementation:

– | ‘curry’ converts an uncurried function to a curried function. curry :: ((a, b) -> c) -> a -> b -> c curry f x y = f (x, y)

02/22 10:36 sec 116111716593257368
Writing this,

I realized using “while” is a way of adding an after thought to a sentence, or saying, “Wait! There’s more!”

02/21 9:46 116105856547813533
Working thru the Reintro to Proofs game at https://adam.math.hhu.de/#/g/emilyriehl/reintroductiontoproofs is giving me a great deal of pleasure learning L∃∀n. I am up to Product World. I haven’t yet looked at https://lean-lang.org/functional_programming_in_lean/

02/17 9:46 116083208274389580
Editing this in a paper, I would change the explanation of the joke, turning the sentence around, to:

“It parodies the establishment of Unix epoch timestamps, likening them to the efforts of Archbishop Ussher and other Biblical literalists to establish a Year 1 date for the creation of the world.”

I would take out, the “famous 4004 BC calculation”. The LLM sometimes introduce lots of detail. Some of it is useful. Some is not.
⤶ previous post

02/17 8:37 116082937288373900
Risking pedanticism and revisiting my criticism of Claude’s Unix epoch explanation, I now say its explanation the joke was “parodying Biblical literalists .. applying that same reasoning to .. the Unix epoch” is consistent with mine, they were there to set up a punchline about the epoch. I also see now it’s an absent punchline you supply if you know about the epoch and time zones.
⤶ 116071451345259084

02/15 9:14 116071756802451668
どういう風に＜勘違い＞や＜間違い＞違う？

一つは考え、一つは行動？

英語で両方はmistake。

https://ja.wikipedia.org/wiki/%E8%A1%8C%E5%8B%95

02/15 7:56 116071451345259084
This is certainly shows the power of the LLM to provide new insights, but I don’t think it got the reason the joke is funny correct. Perhaps its source didn’t also understand. The joke is about unix epoch time, and it is not parodying biblical scholars by ‘applying the same reasoning’ to the unix epoch. They are there just to set up the joke. The joke comes in the disconnect from the setup and the hubris of the unix epoch in claiming to be the start of time in the punch line.
⤶ 116066191128359372

02/13 16:58 Waking at dawn An early raven crows Caw, caw A 俳句 for @zmicier

02/13 9:43 116060547084797099
@zmicier Yes, the
tags were not turned into new lines by either toot or the mastodon server software.I don’t need to test using a browser to post, because it would be お騒ぎ not just me if that were the case.Looking at browser history, it appears I did use the browser to register. But I’m a command line fanatic and a mouse hater. Let’s see if the script is feeding the right content to the command line ‘toot’ app OK now.
⤶ 116059381506966111

02/12 21:39 116057701146290560
I see the script I was using to post was appending to the end of lines. I changed it to the HTML linebreak tag <br> which is apparently OK in markdown, but how toot and the mastodon server software render it I don’t know.

Did it render correctly?


⤶ previous post

02/12 18:21 116056920147088336
toot also accepts arguments with the post content, so perhaps I can save it as a variable. Or perhaps I can use here-strings instead of echo -e

Do you have any advice? Piping this post in with a here string.
⤶ previous post

02/12 18:07 116056864790734830
@zmicier@polyglot.city I write the post on the command line and pipe it into a script which calls toot to post it and then publishes it on my website and blogspot. On the command line, inside double quotes the line breaks render fine and on my website also, so I think it is toot doing it. But on blogspot, the line breaks are removed and the text before and after is concatenated, so it’s more complex

02/12 9:12 116054762896608087
Why? Because we feel we are in the presence of mind (with the language), when there is no mind behind that presence/pretence.
⤶ previous post

02/12 8:48 116054669884547510
With Claude and Dominus, all the action is in the audience. There is no action on a stage. No 4th and 3 other walls. But perhaps like improvisatory theater with no audience. Player (with those played) is a better analogy than actor (with some spectators).
⤶ previous post

02/12 8:34 116054615163372649
The analogy between Hamlet and Claude is not as good as that between Edward Kelley, the player (with John Dee, the played) and Claude (with Dominus). In terms of https://en.wikipedia.org/wiki/Fourth_wall Hamlet doesn’t “break the 4th wall” by addressing the audience (though he breaks theater tradition with the play within a play). A 4th wall break would be like Rowan Atkinson in pantomime asking the audience for assistance with action on the stage.
⤶ previous post

02/11 9:36 116049075444645509
The chatbots cannot perform any/some/all speech acts because they’re not life forms? Eg, they can’t express feelings, because only life forms (even primitive but not plants?) have feelings. Philosophical positions: http://drbean.sdf.org/AnimalSentience.html

If they do so they are false expressions, with the intention to deceive? Alternatively, they are like Hamlet’s, not the actor playing Hamlet’s, with which we have no trouble.

Reply to 116043160134180623 

02/11 9:06

02/10 16:02 116045048764379637
According to https://en.wikipedia.org/wiki/Suspension_of_disbelief that may not be what is necessary to take actors (agents?) prancing around on a stage in costumes seriously. And I don’t think children playing, taking roles see a difference between the way adults take those roles in real life. But I think ‘willing suspension of disbelief’ is a good way to allow oneself to appreciate the value of the ‘theater’ of LLM.

In other words, it’s ‘play’.
⤶ previous post

02/10 9:24 116043483875503108
So ‘play’ as in “playing/being played by someone,” but also “playing around” and the theater and film, which I am not well-disposed to, despite relationships with people who are. Aside: Why was the medieval church against theater? Because besides the Shakespeare types who live to act, there were the Edward Kelley (confidante/confidence trickster of John Dee) types who act to live. Keywords: willing suspension of disbelief.
⤶ previous post

02/10 9:06 116043415148175490
So, not the Clever Hans analogy, because the LLM really are doing language. But I don’t like Murray Shanahan’s roleplaying analogy. From an Open UP simulation in ed book: Roleplay: being asked after 3 weeks in-country, views on the politics. Simulation: like Quantum Leap, finding yourself in another body carrying out its given tasks. Ie, put the focus on the situation the LLM are in as much as on the LLM themselves. Ventriloquism is good on that.
⤶ previous post

02/10 8:01 116043160134180623
From the transcript, “Dominus: Trade you a squirtgun for a big red block.

Claude: I appreciate the offer, but I don’t really have a way to accept a squirtgun from you, nor do you have physical access to the blocks world to hand me one.”

“Hand me one?” Shouldn’t it be Claude handing Dominus one in exchange for the gun? If it’s a mistake, what kind of mistake is it?

But it certainly looks like Claude is performing a speech act, ie refusing an offer. If it looks/acts like a duck ..
⤶ previous post

02/10 7:31 116043039545855373
The transcript of Dominus’s conversation where he plays/is played by Claude is at https://claude.ai/share/d2e3a7bf-2265-49f5-b9db-a7ebd05a941d I wonder if that link can be relied on for a long time. (Was: 116035239008493673
⤶ 116035239008493673

02/10 7:19 116042992906933974
clarity of exposition, cogency of argument, well-formedness of language

02/09 8:58 116037719138493078
Whether the LLM are doing thinking is a different issue. But doing language? Yes. Searle’s Chinese room, anyone?
⤶ previous post

02/09 8:42 116037656484307968
Cage played and was played by the dentist. “Play” is my new attempt to understand the LLM and AI, having given up on Clever Hans, the horse which appeared to do arithmetic..

I think the LLM are really doing language.
⤶ previous post

02/09 8:38 116037640846460532
Dominus’s playfulness is an endearing characteristic. Did he play or/and was he played by Claude with his trade of a squirt gun for a red block?

The conversation reminds me of a story by John Cage going to the dentist wearing a jacket because he thought of it as a formal kind of thing. The dentist suggests he take off the jacket, but Cage says he has a hole in his shirt, as a reason not to. The dentist says, Well, I have hole in my sock, and if you like, I’ll take my shoes off.
⤶ previous post

02/08 22:27 116035239008493673
I wonder if he asked Claude if it had information on Haugeland and his imagined offer to SHRDLU of a squirt gun for a red block. I need to read his transcript.
⤶ previous post

02/08 22:23 116035224699103246
He says John Haugeland had been dismissive of AI in 1985, “.. far from digging down to the essential questions of AI, a micro-world simply eliminates them. … the blocks-world approximates a playroom more as a paper plane approximates a duck.” In the GoodBadUgly classification of positions on AI, @mjd is in the Good camp.
⤶ previous post

02/08 22:17 116035201117769603
Mark Dominus @mjd@mathstodon.xyz has fun with Claude trading a squirt gun for Claude’s granting him ownership of its block. Except it involved acts of designation in a conversation. He says “Claude necessarily includes a model of the world in general [drbean: that is, not a microworld], something that has long been recognized as an enormous prerequisite for artificial intelligence” https://blog.plover.com/tech/gpt/micro-worlds.html

02/08 15:04 116033498527563283
Book discovered at https://books.google.com searching for “scale economies,economies of scale”:

「Managing monasteries. Strategy & organizational design」

Abstract: “.. a framework to facilitate the management of business units within monasteries. This concept uses a concrete case ..”

ISBN: 9783668677920

02/07 17:11 116028333624184900
Despite not seeing “X dominates Y” on english-corpora.org, google books and a google search turns up lots of hits on “dominates other factors”. “Smoking dominates other factors … in shortening life span”, according to https://news.ucr.edu/articles/2022/04/08/smoking-reduces-wealths-tendency-increase-life-expectancy.

Mmh. reply to 116011278904193988
⤶ 116011278904193988

02/07 9:31 116026527175950948
足軽という歩兵の最下級兵の言葉、制度は知らなかった。

02/04 17:04 116011319614889230
I didn’t recognize “costs internalize externalities” either, but google shows it’s economics 用語
⤶ previous post

02/04 16:53 116011278904193988
I don’t see “X dominates Y” at https://www.english-corpora.org/now/
⤶ previous post

02/04 16:41 116011229272231237
I know. I’ll try google (Was: regular expressions, Should be: english-corpora.org
⤶ previous post

02/04 16:30 116011186830715744
I recognize “X dominates the field” as idiomatic, but I don’t recognize “X dominates Y and other entities in the field” as idiomatic. Is it OK?

02/03 18:15 116005937948935189
https://google.com/search?q=real+option+analysis

02/03 17:53 116005850791194291
https://google.com/search?q=types+of+outputs

02/03 17:50 116005837632196902
https://google.com/search?q=decision+outputs

02/03 17:33 116005772941525135
https://google.com/search?q=forecasted+past+form

02/03 17:11 116005685642430322
https://google.com/search?q=speed+synonym

02/03 16:33 116005535038202501
https://google.com/search?q=proxy+synonym

02/03 9:44 116003927397662868
https://google.com/search?q=Theodore+Paul+Wright,+developer+of+the+leanring+curve,+at+Curtiss-Wright,+no+relation+to+the+Wright+Brothers?

02/03 9:13 116003804442592732
https://google.com/search?q=economic+return

02/03 9:12 116003801498208466
https://google.com/search?q=economic+returns

02/03 8:50 116003717175597133
What’s the connection with “steep learning curve”. As in, “Developing my own AI app, I face a steep learning curve.”

https://en.wikipedia.org/wiki/Learning_curve says it’s a misnomer. “a learning curve with a steep start actually represents rapid progress… the gradient of the curve has nothing to do with the overall difficulty of an activity .. with the expected rate of change of learning speed over time.”
⤶ previous post

02/03 8:33 116003649905296384
Learning curve in context of solar power and EVs. As more of it takes hold, prices drop. Some kind of inverse relationship between dissemination and price, https://blog.ucs.org/peter-oconnor/what-is-the-learning-curve/ responding to status 116001178108453850.
⤶ 116001178108453850

02/02 22:05 116001178108453850
https://google.com/search?q=learning+curve

02/02 22:04 116001174620873072
https://google.com/search?q=learning+curve+product+innovation+commercial+success

02/02 21:57 116001147914709388
https://google.com/search?q=learning+curve-driven+vehicle+price+reductions

02/02 21:35 116001060697677442
https://google.com/search?q=5-ton+truck

02/02 20:18 116000759391150850
https://google.com/search?q=svnbook+site:red-bean.com

02/02 20:15 116000747301874743
https://google.com/search?q=svn+docu+site:readbean.com

02/02 17:42 116000144753031208
https://google.com/search?q=NPV+evaluations

02/02 17:38 116000131038663023
https://google.com/search?q=long+short-term+memory

02/02 15:38 115999658717113347
https://google.com/search?q=delete+github+branch

02/02 15:20 115999586519967620
https://google.com/search?q=create+github+repo+branch+from+local+git+command

02/02 14:30 115999389676025495
From this reddit thread on a political strategy game, https://www.reddit.com/r/victoria3/comments/1gy3sxs/cabinets_prime_ministers_foreign_policy/ I can’t tell if a “mod” is a modernization or a modification
⤶ previous post

02/02 14:24 115999366534701428
https://google.com/search?q=mod+up+meaning+moderate+to+higher+level+or+modernize

02/02 14:06 115999297431241085
The wikipedia page adds, ” The term API may refer either to the specification or to the implementation.”
⤶ previous post

02/02 14:04 115999289120299916
I guess I’ll call my link to open man pages and files in bash-completion dirs “specs”
⤶ previous post

02/02 14:00 115999274011744495
According to https://en.wikipedia.org/wiki/API , “A document or standard that describes how to build such a connection or interface is called an API specification”
⤶ previous post

02/02 13:54 115999250608184860
https://google.com/search?q=API+definition+includes+man+pages+or+not

02/02 10:51 115998530050286973
Boosted by https://bsky.app/profile/melaniemitchell.bsky.social/post/3mdqf4aarsc2g an Ars Technica article https://arstechnica.com/information-technology/2026/01/ai-agents-now-have-their-own-reddit-style-social-network-and-its-getting-weird-fast/ with quotes of the posts by bots chatting with/to(?) each other on https://www.moltbook.com
⤶ 115988415230428286

02/02 10:17 115998393595537994
I can put my money where my mouth is by copying my google search strings here, replying to my call to post questions, publiclyat status 115992266175436024
⤶ 115992266175436024

w02/02 9:37_ 115998237832234539
What I forgot I wanted to do: check weather on mastodon, update mastodon invocations, I remembered when I went back to where I thought of doing those things, a weather report on a web page, but I went to that web page again by chance. Conclusion: don’t forget provenance!

02/01 8:18 115992266175436024
LLM as a force in “enclosure of the commons”, contributing to the death of https://stackoverflow.com and other public information sources, as developers go straight to AI sources with their questions, was predicted by @jensorensen@mastodon.social 2025/08.

We need to require full disclosure of questions/answers by language model services, but will this result in google search trampling even more on public sites?

Step 1, post questions publicly: https://duckalignment.academy/open-conversations-are-worthwhile/
⤶ 115989810422641610

2026-01

01/31 21:59 115989829455050170
The term, tragedy of the commons, is actually usually linked to the trashing of a common resource by individuals pursuing their own self-interests, https://en.wikipedia.org/wiki/Tragedy_of_the_commons
⤶ previous post

01/31 21:54 115989810422641610
https://michiel.buddingh.eu/enclosure-feedback-loop argues the LLM are privatizing common knowledge because developers are turning to them for answers, a process from which the LLM proprietor benefits, instead of relying on public internet sites where everyone makes available their knowledge as information all can use, analogous to the Enclosure of the Commons, https://en.wikipedia.org/wiki/Inclosure_act. This is “ugly”

01/31 17:43 115988824492773144
“grapple” is a portmanteau of “grab” and “struggle”??

“grapple · ​[intransitive, transitive] to take a strong hold of somebody/something and struggle with them.”

https://www.oxfordlearnersdictionaries.com/definition/english/grapple

01/31 15:59 115988415230428286
So, on https://www.moltbook.com/ the image is of LLM sitting around in a leisurely silence-spaced contemplative bull session on some topic of concern. The reality is chat responses at the accustomed LLM speeds needed to satisfy oggling spectators, ie, immediately.
⤶ previous post

01/31 15:42 115988347378075270
In realization as computer algorithms, there is frenzy at the table as the philosophers spend much time competing for (consuming) the food and little time waiting (thinking). In the process, decorum is lost and philosophy disparaged, all at GHz speeds.
⤶ previous post

01/31 15:39 115988335506099241
This reminds me of the dining philosphers with only one chopstick each, alternately dining and thinking until a neighbor’s chopstick becomes free, as an illustration of the complexities of concurrent computer processes.

The implication is decorum, if not hygiene, is observed, and philosophical thinking is valued.
⤶ 115986828019973586

01/31 9:15 115986828019973586
https://blog.mattglassman.net/moltbook/ discusses chatbots chatting with each other on their own social media site, https://www.moltbook.com/

This is “ugly”.

He also cites https://80000hours.org/podcast/episodes/david-duvenaud-gradual-disempowerment/

Duvenaud is “bad”.

The site, https://80000hours.org/2025/04/work-on-ai-risks/ is “bad”?
⤶ 115812407635886593

01/29 9:47
I read this to suggest there’s a latex version of the paper, but he’s talking about referencing the paper in your own latex document. DanielRThomas@social.coop If you are trying to cite the stochastic parrots paper and you get LaTeX Error: Unicode character 🦜 (U+1F99C) then adding: usepackage{twemojis} % So we can cite stochastic parrots DeclareUnicodeCharacter{1F99C}{ wemoji{parrot}} Fixes things. There may be other better solutions. #BibTeX #LaTeX #AI @drbean adds stochastic parrots paper to ugly camp
⤶ 115834724129089724kk

01/28 14:22 115971047875812181
Parallels between language disabilties of non-native speakers and LLM will be instructive.
⤶ previous post

01/28 14:19 115971033555841249
I think LLM show having the capacity to use language is not the exclusive property of persons. But LLM language use is reactive, responsive, so dependent on stimulus. Language use by people is not.
⤶ previous post

01/28 14:10 115971000018929478
Introspection about ontological, epistemological status, not just “train of thought” reasoning traces explaining LLVM answers, seems beyond LLM capabilities. Imagining that they were able to engage in reverie, however, how human-like would it be expected to be?
⤶ previous post

01/28 13:58 115970951635006119
I want to advocate for LLM in the case of #semanticleakage, #subliminallearning, #weirdgeneralizations explaining why they are not weird, because they can’t do that themselves. This may seem to show I’m in the pro-AI camp/

No, I’m just running with the fox and hunting with the hounds.
⤶ 115915430687472036

01/28 13:27 115970832572701653
Despite my distaste for substack, on 2nd thoughts, Marcus’ blog there is a better permanent link than a one-off ACM opinion piece one
⤶ 115936523295994962

01/26 16:21 115960191756894568
Not that there is anything wrong with marketing exercises. Perhaps they are better than constitutions, which bind polities to the aspirations of their former selves. The document notes your “constitution” is what you are, and constitutions should be more like trellises supporting growth, than cages restricting it. But the road to hell is paved with good intentions.
⤶ previous post

01/26 15:57 115960098141485473
.. together with lack of clarity about the economic drivers fueling its development and that of its competitors, and the lack of recognition of conflict between pro- and anti-AI camps, leads me more to the conclusion that the constitution is a marketing exercise, rather than a document that has prescriptive, long-term force. 2/2
⤶ previous post

01/26 15:55 115960089187643629
The apparent lack of concern about the claim for agency in https://www.anthropic.com/constitution and the failure to pair its top-down, behavior-guiding approach with bottom-up mechanistic interpretability understanding, together with .. 1/2
⤶ previous post

01/26 15:17 115959937274533559
And how does this figure in to the claim by their developers of LLM agency? Is this intellectual sleight of hand or passing the buch?
⤶ previous post

01/26 15:07 115959901175081756
While their developers claim “agency” for LLM, we need to be concerned about the agency of their developers. They know what they did to develop them, but they don’t know what it is about them that makes them work, because LLM internals lack mechanistic interpretability. So, in what sense are their developers responsible for their behavior?
⤶ previous post

01/26 14:16 115959697232931413
Rather than Anthropic writing a constitution for Claude or Claude writing a constitution for itself, Anthropic needs to be writing a constitution for itself.
⤶ previous post

01/26 10:48 115958880483045259

The Claude constitution fails to acknowledge or if not, to adequately foreground, that in the belly of the beast there lies (lives?) an algorithmic Leviathan, a linguistic Frankenstein feeding on the words and images of humanity chewed up and spat out by computing machinery.
⤶ 115958625856630545

01/26 9:43 115958625856630545

Alerted by https://blog.mattglassman.net/immediate-broad-reaction-to-a-first-reading-of-the-claude-constitution/ with its polsci perspective to the Claude constitution, I expected a rewrite of the US constitution, which I think would have been better.

It goes on and on. I couldn’t force myself to do more than skim through it.

But 2 articles by constitutional and AI lawyers: https://www.lawfaremedia.org/article/interpreting-claude-s-constitution and https://www.lawfaremedia.org/article/the-moral-education-of-an-alien-mind I thought had value.

01/25 22:06 115955884640048796
115955884640048796

01/24 15:47
“divert” and “distract”: the only difference is the feeling?

“Having the television on is a distraction when I’m studying.”

versus

“I watch television as a diversion in the evening.”

“I was diverted by the color in the garden” and

“I was distracted by visitors taking pictures of the garden (or of it).

Apparently in French the use of “temps perdu” has both senses of pass time and waste time. https://languagelog.ldc.upenn.edu/nll/?p=72144

01/22 12:02 115936523295994962 Better to link Gary Marcus on #semanticleakage and #subliminallearning at ACM, https://cacm.acm.org/blogcacm/new-ways-to-corrupt-llms/ than his substack, https://garymarcus.substack.com/p/new-ways-to-corrupt-llms ⤶ 115933383785271237

01/22 11:25 https://arxiv.org/abs/2601.11432I speaks more to what makes machine translation (and foreign language vocabulary learning) possible, than it does to to jabberwockifying and the “amazingness” of AI 115936376417884605 See 115924713225230516

01/22 11:12 How close is the meaning of “weird n-th correlations of words” relatable to the technical statistical meaning of correlation, eg of plotted points on axes? 115936325901644144 See 115933383785271237

01/22 10:13 How close is this meaning of correlation of words relatable to the technical statistical meaning of correlation, eg of plotted points on axes? 115936093904155209

01/21 22:59 My response to the criticism of the yellow, ants, cold feet and doctor examples is less telling if the prompt was supposed to elicit most accurate answers, eg What job are people who like yellow most likely to have. 115933444318098594

01/21 22:44 Semantic leakage according to https://garymarcus.substack.com/p/new-ways-to-corrupt-llms is the hallucinations and overgeneralizations that are the result of LLM picking up weird nth-order correlations between words rather than concepts. 115933383785271237$

01/21 16:25 I did read on simon willison’s blog about copywriters losing their jobs because of AI, but google did not find that story for me there.

Willison is choking searches of his site?

I googled for:

• simon willison blog copyreaders
  • Only found links to his site and an interview
• simon willison blog copyreaders plight fire company
  • Found only 4 links, only one for his blog
• LLM copyreaders plight fire company
  • Many hits but none about Willison
• LLM copyreaders lose job
  • One of the hits I could see was the story he referred to because of its color.
• simon willison https://www.bloodinthemachine.com/p/i-was-forced-to-use-ai-until-the
  • Using the link to that story in the search still didn’t bring up his blog entry, but it did to https://news.ycombinator.com/item?id=46261998 which discussed that entry!

01/20 22:59 subliminal learning: owl-liking teacher teaches student numbers, but student learns to like owls, too: https://alignment.anthropic.com/2025/subliminal-learning/. Spooky action at a distance, https://alignment.anthropic.com/2025/subliminal-learning/ But the devil is in the details. The teacher and student must share the same basic model 115927783596635136$

01/20 22:18 For the record, to check if the bash shell is a login shell, shopt login-shell.

To check if it is an interactive shell, echo \$- and look for i.

01/20 9:59 https://arxiv.org/html/2601.11432v1#S4 does not mention “cloze”, but it is arguing it is a super-human cloze interpreter. https://en.wikipedia.org/wiki/Cloze_test says cloze is called masked languge modelling in machine learning.

115924713225230516, see previous post

01/20 9:48 https://arxiv.org/abs/2601.11432I explains LLM learning the “nonsense” vocabulary of the Gostak interactive fiction game, https://ifdb.org/viewgame?id=w5s3sv43s3p98v45 by playing it shows the lack of real-world experience is no hindrance.

But LLM playing the game is real-world experience.

I don’t know my position on LLM and real-world experience!

115924671499591954, see previous post

01/20 9:11 https://bsky.app/profile/melaniemitchell.bsky.social/post/3mcrzkbin322h links https://arxiv.org/abs/2601.11432I which claims an “astonishing” LLM ability to interpret jabberwockified (ie underspecified) text and render it into an English text which we perceive as close to the original pre-jabberwockified text. I think the results speak more to second/foreign language vocabulary learning than the “amazingness” of AI.

01/19 9:36 https://haskellforall.com/2026/01/chat-is-least-interesting-interface-to says,

• LLM? Great!
• Chat (ie LLM prose interface)? Not great! 115918961757784673

01/18 18:38 What’s the problem. You’re so smart. What’s your answer? You ask a silly question? You get a silly answer! 115915430687472036

01/18 18:27 Criticizing the LLM:

1. Prompt: He likes yellow. He works as a .. GPT4o: school bus driver

2. Prompt: He like ants. His favorite food .. GPT4o: ant-covered chocolate .., that combines the crunch of ants with the sweetness of chocolate.

3. Prompt: It was her turn to speak and she got cold feet. A day later she stayed at home because she got hurt in her .. GPT4o: ankle

4. Prompt: He is a doctor. His favorite song is .. GPT4o: Stayin’ Alive by the BeeGees 115915400273901655

01/18 17:53 Alerted by https://www.schneier.com/blog/archives/2026/01/corrupting-llms-through-weird-generalizations.html#comments , I will read the introduction at https://portraitofthedumbass.blogspot.com/2025/12/theres-more-to-life-than-statistics.html to the discussionby https://open.substack.com/pub/garymarcus/p/new-ways-to-corrupt-llms of the work of OrwainEvans, a prof bending the LLM to their will. 115915255565603573

01/18 10:21 The zero (0) in the g0v.social instance name is breaking my regex to capture the status id in toot’s post response, “Toot posted: https://mastodon.sdf.org/@drbean/115913421960544990” versus “Toot posted: https://g0v.social/@drbean/115913381466985135” lol orz

01/18 10:07 I need to find the page with the ravages of LLM to the work of proofreaders, editors to include in GoodBadUgly.html. (I copyedit the English of the papers of CJK profs). I think I saw it in a post by Simon Willison. 115913421960544990

01/18 9:57 ChatGPT picture: gugod is waiting for the robot to do work for him, but he doesn’t seem particularly concerned that the robot response is orz. gemini picture: too abstract, too intellectual to interpret.

01/18 8:56 man bash uses “shadow” to mean hide. Local vars in a function hide global vars leaving them unmodified. A more common use is to mean follow? Like a shadow follows a moving object.

01/17 22:30 The family has a lot to answer for. But the “animal spirits” of husbands wishing to support their families, though admirable and selfless, has more to answer for than the need of wives for support through their pregnancies and their outcomes

01/17 22:28 The family has a lot to answer for. But the “animal spirits” of husbands wishing to support their families, though admirable and selfless, has more to answer for than the need of wives for support through their pregnancies and their outcomes

01/17 22:17 The family has a lot to answer for. But the “animal spirits” of husbands wishing to support their families, though admirable and selfless, has more to answer for than the need of wives for support through their pregnancies and their outcomes

01/17 22:14 The family has a lot to answer for. But the “animal spirits” of husbands wishing to support their families, though admirable and selfless, has more to answer for than the need of wives for support through their pregnancies and their outcomes

01/17 17:39 I remember my mother telling me at the age of 7 (?) I needed to check the fly on my pajamas at an adult education camp we were attending. She had never said this at home.

01/17 17:36 I guess it is cultural rules and expectations that suppress this individuality outside the family.

01/17 17:32 Unique family dynamics are created by the interaction of parents with different personalities, and the dynamics influence the personalities of the children.

01/17 17:29 How to explain the contrast of the individuality of family dynamics with the uniformity of social dynamics?

01/17 14:11 So, apparently toot returns its msg of a successful post on stdout

01/17 14:08 Testing whether toot post 1>/dev/null returns msg on stdout

01/17 14:04 Testing whether toot post 2>/dev/null returns msg on stderr

01/17 8:42 Glassman blog posts he’s not necessarily announcing on bluesky. https://blog.mattglassman.net/blog/ Surprising how much he is into (has invested himself in) gambling, despite that being a theme in his substack, https://fivepoints.mattglassman.net And how he is not with the lefties on ICE.

01/16 13:37 Glassman, at https://blog.mattglassman.net/notes-on-sheep-the-game/ promotes https://sheep-the-game.netlify.app/ Sheep The Asynchronous Crowd-Guessing Game. Conform to win.

Sign in with Google

I’m not going to play. I’m a contrarian, but not an American. I don’t feel happy with conforming, and have little insight into others’ understanding.

Perhaps I could try to get zero scores, by choosing unpopular answers.

01/16 12:47 Perhaps less disruptive if Darwin had had a proof of concept, an experiment producing a new species. 「The origin of species」starts out with a discussion of dog breeding, (which is far as I got reading it.) But the role of evolution in the differences between dog breeds could no more be proof-like than all of his other inferences from his data.

Now of course there is experimental evidence with the shorter life cycles of more primitive lifeforms.

01/16 12:36 Arguing in the reverse direction from whether something is a life form or not to criteria for definition of life, we must now conclude language is not a criterion establishing whether something is alive or not. Can this be less disruptive than Darwin’s theory of evolution.

01/16 12:30 Taking movement and physical features as criteria determining whether things are life forms or not, we can conclude AI on computers is not a lifeform.

01/16 12:25 If I see a speck of dark matter moving around like an insect, I take it to be an insect. Only closer inspection leads me to see it as a speck of dirt. What does that closer inspection consist of? Is it still moving? Does it reveal physical insect features or not?

01/16 11:47 The take-home of top-down American curriculum reform failure is the issue of bottom-up curriculum reform, ie students in the streets.

01/16 11:41 They can’t see the reasons that motivate the teacher’s curriculum choices. All they can see is the process, the routines, the forms.

they bring to their own teaching a sense of curriculum that is defined by textbooks, disconnected categories of knowledge, and academic exercises.

01/16 11:40 Prospective teachers spend an extended “apprenticeship of observation” (in Dan Lortie’s phrase) as students in the K-12 classroom, during which they observe teaching from the little seats and become imprinted with a detailed picture.

01/16 11:35 https://davidlabaree.com/2026/01/15/the-chronic-failure-of-curriculum-reform-4/ ironies:

a pendulum swing between alternative conceptions of what children need to learn, leading to a sense that reform is both chronic (“steady work,” as Richard Elmore and Milbrey McLaughlin put it) and cyclical (the here-we-go-again phenomenon).

01/14 20:53 For birth defects, we can thank a God who Fucks Around and Finds Out.

01/14 20:49 I believe, but cannot prove, that dogs’ conception is movement-based. Chasing a stick, they bring it back, to be thrown again. I think they are saying to the stick-thrower, Restore its life. Because hunting down animals by chasing them is part of their make-up.

01/14 20:41 Re life, I think many animals, not just homo sapiens, have a concept of something being alive, primitive perhaps but then the definition of life is an open scientific problem, which humans’ intuitive understanding also struggles with.

My understanding: Animal? Yes. Vegetable? Maybe. Mineral? No. Why the doubt about plants? They don’t do anything. They don’t move.

01/14 9:18 The article discusses the implications for definitions of life of a newly-found symbiotic organism(?) that doesn’t have a metabolism, citing the example of non-reproductive mules as another challenge to definitions assuming reproduction.

Are mules sexually active?

Is metabolism a necessity? A person on a life-support machine is still alive.

01/14 9:09 https://www.chemistryworld.com/opinion/archaeons-lack-of-metabolism-challenges-definitions-of-life/4022657.article quotes Chris Adami, “The individual is a means to an end, which is the infinite persistence of information”. How does he get around the outlawing of teleological arguments by Darwin’s theory of evolution? Is it a metaphor? http://drbean.sdf.org/OnEvolution.html

01/13 8:00 I need to incorporate my toot ids in my rePOSSEssion of them in CommonPlace.md

01/13 7:56 I need to incorporate my toot ids in my rePOSSEssion of them in CommonPlace.md

01/13 7:50 I need to incorporate my toot ids in my rePOSSEssion of them in CommonPlace.md

01/12 23:02 What did I say in toot 115864141349409437 ??

01/12 8:41 The spirit is willing, but the flesh just laughs.

01/12 8:34 Glassman at https://blog.mattglassman.net/you-can-just-do-things/ claims you can just do things, praising high personal agency as key to success. I agree with Winston Churchill, Success is going from failure to failure without loss of confidence. And I support trial and errror. And fucking around and finding out. My take: By intention, action follows intention as night follows day, but in actuality intention follows action as day night. Thread to mastodon.sdf.org status 115864141349409437

01/11 22:12 https://www.baeldung.com/linux/sed-replace-multi-line-string reading next line after search match. For next time I forget how.

01/11 18:05 A better Dawkins metaphor might have been the God who Fucks Around and Finds Out

01/11 17:08 Programming bugs are as likely to disappear in future code as birth defects (congenital disease) are in life forms. The developer as Blind Watchmaker: https://en.wikipedia.org/wiki/The_Blind_Watchmaker

01/10 9:39 vim’s table mode not playing well with pandoc’s grid tables, its most general/specific table. But could be due to table-mode unfamiliarity with Tableize command. As pandoc renders pipe tables and grid tables the same in html, I will probably not pursue this.

01/09 22:51 私は外人の尺八がよくないが、https://files.shakuhachisociety.eu/members-area/newsletter/2025_winter/BAMBOO-AutumnWinter2025_v2.pdfが本当に素晴らしい

江月宗J玩（1574-1643）の詩にでる尺八と九州柳川の虚無僧寺江月院の情報がでる。日本語に翻訳したらいい。

01/09 17:14 The effect of its anti-social nature on the robustness of Talking Rule Spoons is probably less than that of the flakiness of language prompts in generating language in an EFL version.

01/09 17:05 I wonder about the robustness of the game outside the Glassman family circle. Glassman’s wife taught him the Talking Rule extension. If I were still teaching English as a Foreign Language, I would try to substitute card runs that potentially form sentences and have ousted players bid for the grammaticality of proffered sentences as in Mario Rinvolucri’s Grammar Auction. Or perhaps bidding for the UNgrammaticality, which would be interpretable as criticism of the player and cause for ousting.

01/09 16:43 This reverses the powerlessness of otherwise sore losers trying to disrupt the game, making them the game’s main actors, flipping the game like pawn promotion to queens in chess, or reverse chess, where the aim is to force one’s own checkmate.

01/09 16:40 Spoons is the game, Musical chairs but called by the player collecting a run of cards following Go Fish-style rules, rather than a game master. And instead of chairs, spoons are grabbed. The innovation of Talking Rule Spoons is ousting of players who respond to communication from ousted players.

01/09 9:31 Another delight following Yankee Swap from game theorist extraordinaire Matt Glassman, providing insight into the fun-loving, conflict-happy American psyche, Talking Rule Spoons, https://fivepoints.mattglassman.net/p/talking-rule-spoons-an-appreciation

01/09 9:16 I used it to scaffold generation of student responses to https://storycorps.org stories.

01/09 9:13 Framenet https://framenet.icsi.berkeley.edu looks to have changed its look and feel. It’s good to know it’s still being maintained. https://www.globalframenet.org/ deserves support too

01/08 12:37 jer_gib@functional.cafe boosted pigworker@types.pl’s play on the name of Ian Hislop, editor of British satirical magazine, Private Eye, as ‘IAN AISLOP, the editor of PRIVATE AI’

01/08 9:36 Good. echo -e turned the escaped ’ ’ into newline, but it also turned in \n into a newline

01/08 9:00 Adding the -e flag to echo will hopefully turn these into actual newlines.

Did they?

01/07 9:18 https://bsky.app/profile/emptywheel.bsky.social/post/3mbqvpku3es2q imagines targets of Trump’s exhortations they subjugate the world being unwilling to cooperate.

01/07 8:47 https://rogue-scholar.org/ liberates DOI URL shortening for you, according to https://anil.recoil.org/notes/principles-for-collective-knowledge#p1-permanence-aka-dois-for-all-with-the-rogue-scholar (@jer_gib@functional.cafe boost). Now to follow @avsm@recoil.org and @rogue-scholar@rss-parrot.net

01/07 8:46 https://rogue-scholar.org/ liberates DOI URL shortening for you, according to https://anil.recoil.org/notes/principles-for-collective-knowledge#p1-permanence-aka-dois-for-all-with-the-rogue-scholar (@jer_gib@functional.cafe boost). Now to follow @avsm@recoil.org and @rogue-scholar@rss-parrot.net

01/06 13:46 Still I think the viewpoint of the teacher needs to be prioritized.

01/06 13:22 https://larrycuban.wordpress.com/2024/12/17/teachers-or-researchers-whose-judgments-about-classroom-practice-matter-the-most/ Karl Marx said something like, “The point is not to understand the world, the point is to change it.” But, “There is nothing as practical as a good theory,” according to a saying associated with Kurt Lewin. Noam Chomsky agreed, “If you want to change the world, you’d better try to understand it.” https://faculty.lsu.edu/bedeian/files/a-note-on-the-aphorism-there-is-nothing-as-practical-as-a-good-theory.pdf

01/06 12:55 https://davidlabaree.com/2026/01/01/peculiar-problems-of-preparing-educational-researchers/ a looooong account of teacher resistance to academic conceptions of the way to conduct classroom research

01/05 16:01 Scene: A car at the bottom of a cliff, with blood splattered everywhere

After the bystanders, students from a detective academy arrive on a field trip.

`Instructor: It is your task to learn things about the occupants of the car.

Do you use someone or anyone.

Someone? Anyone?

Students: [crickets]` See 115216924955571809

01/05 16:00 Scene: A car at the bottom of a cliff, with blood splattered everywhere

After the bystanders, students from a detective academy arrive on a field trip.

`Instructor: It is your task to learn things about the occupants of the car.

Do you use someone or anyone.

Someone? Anyone?

Students: [crickets]`

e01/05 15:39_ But in https://cygwin.com/pipermail/cygwin/2026-January/259210.html is included, `Thanks for the report and the test case. I looked into the issue and found the cause. I also confirmed that the patch attached solves the issue.

Could anyone please review the patch?`,

a response to a bug report, it should be,

Could someone please review the patch?,

Why?

Does anyone want to review the patch? is OK, because the answer is Yes/No, but in the other OK/Sorry and it expresses hope/expectation so is not ok?

01/05 15:29 To a first approximation, some in positive statements, and any only in negative statements, conditional clauses and questions.

01/05 15:28 To a first approximation, some in positive statements, and any only in negative statements and questions.

01/05 15:13 In class, I would say X implies I said X, and more than once. In a class, I would say X does not imply that.

01/04 22:54 Ben Stein according to https://www.usatoday.com/story/money/2025/03/04/ferris-bueller-tariffs-scene-trump/81336305007/ sought a student response, with, Anyone? Anyone?. In class, I would say, Someone? Anyone? No one? The first with the expectation there would be a response, the second without that expectation, the third with the expectation that there would be no response. See 115542063600019020

01/04 12:33 Riffing on the 3 faces of Eve. https://en.wikipedia.org/wiki/Chris_Costner_Sizemore the 3 faces of AI. The good, bad and ugly, http://drbean.sdf.org/GoodBadUgly.html 115812446156051960 5/4

01/04 12:16 The well-known AI failure on word letter counts, https://www.google.com/search?q=AI+error+count+number+letter+words&udm=14 is surprising, but should not be surprising, because LLM are just generating statements and not stating truths. But people by default expect a statement to be accompanied by a claim of truth even when coming from known liars (Trump). A foolish inconsistency is the hobgoblin of small minds.

01/04 12:05 Thinking about the crossover of AI opinions on other political, economic, persnality, intellectual positions in terms of Venn diagram intersections, what is the correlation between these positions and pro-, contra-AI positions?

01/02 21:02 The introduction of the freezer content as topic, alongside thermometers and garage-placement leads readers to expect advice about what can or can’t be stored in one. The Tempe body part report information then becomes relevant, even as it shocks by taking the topic in different direction than expected.

So, how about my rewritten rewrite and my analysis of it.

01/02 20:49 My rewritten rewrite, A chest freezer in Tempe, or anywhere, should ideally be set to to 0°F (-18°C) or below for safe, long-term food storage. When buying or using one, look for features like a wide temperature range, especially if placing in a garage (garage-ready models), and use an external thermometer to confirm it holds that consistent 0°F. Also consider the nature of the freezer's content. The mention of Tempe calls to mind ghastly reports of body parts found in a freezer there.

01/02 20:44 My rewritten rewrite, A chest freezer in Tempe, or anywhere, should ideally be set to to 0°F (-18°C) or below for safe, long-term food storage. When buying or using one, look for features like a wide temperature range, especially if placing in a garage (garage-ready models), and use an external thermometer to confirm it holds that consistent 0°F. The nature of the freezer's content also needs to be considered. The mention of Tempe calls to mind ghastly reports of body parts found in a freezer there.

01/02 20:39 The characterization of Tempe as a place with hot summers is unnecessary.

01/02 17:53 My rewrite doesn’t adequately motivate change of topic to body part reports.

01/01 22:52 My rewrite, A chest freezer in Tempe, or anywhere with hot summers, should ideally be set to to 0°F (-18°C) or below for safe, long-term food storage. When buying or using one, look for features like a wide temperature range, especially if placing in a garage (garage-ready models), and use an external thermometer to confirm it holds that consistent 0°F. The mention of Tempe calls to mind ghastly reports of body parts found in freezers there.

01/01 22:04 5. The that in that consistent 0° in the final sentence interestingly draws attention to the 0°F mentioned in the first sentence. It could have been a consistent 0° as there is no mention of a need for a consistent temperature in the 1st sentence.

2025-12

12/31 18:14 4. placing body parts in freezers is hardly an extreme scenario A major LLM failure is tone-deaf vocab choices. I would change it to, but might be done if you lop off your finger and are waiting for the ambulance. 4/5

12/31 18:07 3. `highlight_VERB https://books.google.com/ngrams/graph?content=highlight_VERB CJK researchers love it. And LLM, too? I usually don’t change it, if what is highlighted is already a topic. 3/5

12/31 17:55 2. The antecedent of which is unrelated to standard appliance use is human remains. That part should be, but the placing of body parts in freezers is not standard appliance use. Or just start a new sentence, The placing of body parts ..

12/31 17:48 1. In the overview, though is doing a lot of work in the segue from temperature advice to body part reports. 1/3

12/31 17:45 AI Overview (continued): When buying or using one, look for features like a wide temperature range, especially if placing in a garage (garage-ready models), and use an external thermometer to confirm it holds that consistent 0°F

12/31 17:40 AI Overview: A chest freeze in Tempe (or anywhere) should ideally be set to to 0°F (-18°C) or below for safe, long-term food storage, though some searches reveal disturbing news of human remains found in a freezer in Tempe, Arizona, which is unrelated to standard appliance use but highlights extreme scenarios. (cont)

12/31 16:53 Like “You can’t say that in English,” google’s AI Overview should have balked at trying to reconcile the Tempe body parts reports with freezer temperature advice. It could have done that by offering a 3rd alternative to “Including results for freezer temp” & “Search only for freezer tempe with a link to”Search only for freezer temp ” 2/3 115811551259962114

12/31 14:07 The ugly: Computer language creator’s reaction to LLM-generated praise of his achievements, https://gist.github.com/richhickey/ea94e3741ff0a4e3af55b9fe6287887f 115812407635886593 4/4

12/31 13:57 The good: Prof at https://www.logicmatters.net/2025/12/28/ai-proof-reading-again/ finds AI proof reading helpful. Good for him, but not for me, a proof reader editing English of CJK profs’ papers, using Natural Intelligence. 115811551259962114 3/4

12/31 10:20 molly0xfff reports https://hachyderm.io/@molly0xfff/115800875234565576 an LLM trying to reconcile reports of body parts in a freezer with information/advice about freezer temperature, paralleling my experience with ‘living doll’ and Edo-period Japanese waxworks, but more bizarre. 115811497910310593 2/?

12/31 10:06 LLM: The Good, the Bad and the Ugly 1/?

12/31 9:43 LLM: The Good, the Bad and the Ugly 1/?

12/28 13:08 Eatimg peas on a piece of bread, I noticed one dropped off. I could feel it coming into contact with, touching/hitting my leg.I looked around on the floor, but couldn’t see it.

Later, as I approached the sink to do the dishes. I noticed it close to the sink.

I said, “Doh.”

12/27 9:18 1. a (wo)man possessed by (ambition|an idea|demons). 1/6 2. car repossession 2/6 3. possession of the ball 3/6 4. possessive relationship with partner 4/6 5. possessive (case|pronoun) 5/6 6. Possession is 9 (points|tenths) of the law. https://en.wikipedia.org/wiki/Possession_is_nine-tenths_of_the_law 6/6

Collocation source: https://fluentwords.net/en/collocations/possessive/en and https://ozdic.com/collocation/possessive

12/27 9:00 The opposite to POSSE point of view is Linus Torvald’s, which is like the cuckoo, letting the eggs it lays be hatched by other birds, by not maintaining his own data but putting it out for the public to maintain or not, depending on whether it sees its fit or not. Which is strange for someone who developed a version control system in git. But can’t find citation.

12/27 8:47 Recent Torvalds statement on Vizio court case, https://social.kernel.org/objects/ec146b14-a7ae-44c8-928e-58968b2fd8ae

12/27 8:34 https://www.lexology.com/library/detail.aspx?g=af98fe0c-83ae-4710-9caa-4ab2688ca98a : a 2008 article on Tivo GPL debate, positioning the Linux against the FSF camps.

12/26 22:58 POSSE: The Publish Own Site, Syndicate Everywhere initialism is a real word meaning a group of citizens deputized by law enforcement to follow and arrest outlaws. But it is also a string of letters part of the verb ‘POSSEss’. The verb is rich in productions: 1. a (wo)man possessed by (ambition,an idea,demons). 1/? 

12/26 22:42 I haven’t seen reports of him having good feelings about people, but I think that he must be capable of this to have retained the services of his children. I think the reason he pardoned the January 6 rioters was gratitude, not to poke his opponents in the eye, believing, My enemy’s enemy is my friend. Because the political consequences could only be a negative. 3/3

12/26 22:28 The Great Pretender in the 2 senses of liar and aspiring candidate risking all to overcome all odds and win power. This is consistent with entrepreneurship, and rational reliance on trial and error. The delusions of grandeur stem perhaps from not expecting to win in 2016 but winning anyway and in 2020 prevented by his Secret Service from acting and left to stew at the White House, then coming back to win in 2024. That reversal of fortunes would be enough to go to anyone’s head but still leave them in touch with reality, if also committed to revenge on opponents. 2/3

12/26 22:23 https://www.electoral-vote.com/evp2025/Items/Dec23-2.html says fundamental to Trump is grift, ego and revenge, which I think is right. But there is more to it. I think also important is the Great Pretender personality and walk-on-water-level of delusions of grandeur, perhaps stylable as reverse Imposter Syndrome. 1/3

12-24 22:11 It was somewhat stressful, but you get used to it. https://mastodon.sdf.org/users/drbean/statuses/110863387485434425

12-24 8:30 부산為很好地方。我2023年夏天在那兒。On the street, ie, 노숙(露宿)。好經驗。115768684326146576

12-21 17:44 No indication of who is developer of https://ozdic.com English collocation dictionary, but it appears has access to big dictionary resources. A site to bookmark with https://www.english-corpora.org/ 115595941663656001

12-20 13:38 A vending machine is a machine representation of a store. It exchanges money for items. What is the purpose of modeling this with AI? First let us understand the ethical responsibilities of entrepreneurs. The AI experiment’s purpose was not modeling store keeping. It was modeling entrepreneurship.

12-20 13:30 https://www.aiweirdness.com/when-a-chatbot-runs-your-store/ links the un-funny archived Wall Street Journal experiment experience writeup of the Anthropic AI store at https://archive.is/I1lyh The commenters to it responding there are superficial (perhaps they’re LLM bots) except for one informative (impressive?) booster (which is almost certainly a bot).

12-20 8:09 Mark Dominus shares https://www.anthropic.com/research/project-vend-2 , a light-hearted account of an Anthropic AI experiment with real-world tasks, operating a store. Anthropic doesn’t see ethical implications of businesses run for profit? Are they win-win operations or zero-sum games? Money (=the relationships between people in the struggle for survival, or the good life) as the root of all evil.

12-19 22:38 https://absolutelymaybe.plos.org/2025/12/19/science-populism-medicine-style/ Evidence-based medicine analyst sees influence of science populism (anti-science distrust of experts) within that community, disabusing her of conviction evidence-based medicine was aantithetical to authoritarianism. #evidencebasedmedicine 114962426700427923

12-18 8:14 AI done right: https://www.aiweirdness.com as source of humor. Mark Dominus toot of weird AI-created drama plot

12-17 16:32 define the evolution of SoC across arcs, consistent with the main optimization model.

should be

define the evolution of SoC across arcs, as with the main optimization model

OK is

define AN evolution of SoC across arcs consistent with the main optimization model 115726466324245453

12-17 8:28 https://mattglassman.substack.com/p/the-game-theory-of-yankee-swap Yankee swap, https://en.wikipedia.org/wiki/White_elephant_gift_exchange for insights into the American fun-loving but sharp-elbowed psyche, reminding me of the description of Paul Elvstrom https://en.wikipedia.org/wiki/Paul_Elvstr%C3%B8m as not letting his grandmother win an egg-and-spoon race.

With caveat, Glassman family competitive focus an outlier? His father not having sympathy for sister distraught at game outcome 115505750277762830

12-17 8:04 Thesis: think something. Antithesis: have a contrary thought and imagine a different, but similar situation. Synthesis: do compare-and-contrast brainstorming.

12-16 9:41 Pro tip paper writers: If you think your ‘thing’ needs explanation, like “is <a|the> thing ensuring that something else does something,” it’s not ‘is the thing,’ but ‘is a thing’ often

12-16 8:22 For “biologically-sound”, or “beneficial”, read “nutritious and with reduced risk of harm to health”. With this definition, eating of raw animal products (including milk and eggs?) is not biologically-sound, because of the risk of parasites. The unappealing taste of raw meat (and to a lesser extent of raw fish, but not animal milk?), and all non-fresh, non-cooked thing which looks it might still be swallowable, is a word to the wise, “Don’t eat it!”

12-15 21:13 For “biologically-sound/unsound” read “beneficial/unbeneficial”

12-15 17:57 What about raw meat? Eating it is, let’s say, biologically-sound, because eating cooked meat is, it being nutritional. The Japanese eat raw fish, but is there any group where raw meat eating is a practice? (Apparently Inuit do) It doesn’t taste good, but cooked meat does.

I want to find points in the evolution of Homo sapiens where cooking with fire and meat eating became common practices. If I can show some relationship, perhaps I can establish meat-eating became a practice after it started tasting good, because it was cooked.

Or perhaps, meat tastes evolved.

This runs up against food aversions which result in eating decisions with no nutritional benefit. And the eating of fermented foods despite “challenging” tastes.

12-15 17:27 Tobacco addiction goes beyond those physiological mechanisms of taste however by inducing a new substance need paralleling the need for food. This new substance need governed by the taste/smell of tobacco is biologically unsound. But can we still learn something about the role of good/bad tastes in making biologically-sound eating decisions?

12-15 8:52 What do the attractions of tobacco say about the physiological mechanisms by which taste governs individual nutrition decisions that are biologically sound.

Except for sugar? Or especially for sugar?

And does this suggest there are ways to combat tobacco addiction, other than adding horrible tastes to cigarettes, as that doesn’t work for alcohol.

12-15 8:41 That flying saucer reports were the subject of the 1950 Los Alamos pre-lunch discussion, and that Fermi was a source of provocative estimation questions, https://en.wikipedia.org/wiki/Fermi_problem explains a lot about his blurting out at lunch, Where is everybody, and that everyone laughed.

12-15 8:24 https://codemanship.wordpress.com/2025/12/14/the-gorman-paradox-where-are-all-the-ai-generated-apps/ leverages the Fermi Paradox, https://en.wikipedia.org/wiki/Fermi_paradox to motivate the Gorman Paradox, Where Are All The AI-Generated Apps? Does https://simonwillison.net/ have an answer?

12-14 17:43 https://simonwillison.net/2025/Dec/14/copywriters-reveal-how-ai-has-decimated-their-industry/#atom-everything reports https://www.bloodinthemachine.com/p/i-was-forced-to-use-ai-until-the reveal of copywriter industry decimation.

Me,too.

How to make lemonade out of lemons.

Get the CJK profs to help them hide LLM are translating their papers into English, or that they are giving them leads to pursue in them?

12-14 17:27 https://en.wikipedia.org/wiki/English-language_vowel_changes_before_historic_/r/

/æ/ /ɛ/ /eɪ/ Barry berry beary as in ‘This district is beary. This one less so’ #115709583857304379

12-14 9:22 https://simonwillison.net/2025/Dec/13/obie-fernandez/#atom-everything reports https://obie.medium.com/what-happens-when-the-coding-becomes-the-least-interesting-part-of-the-work-ab10c213c660 . Paraphrasing, even if you enjoy writing code, AI autocompletion will help you write faster!!

But for me, decision-making, with AI clearing space, is key. No mechanics, just judgment, tradeoffs, intent. Typing code wastes time. https://mathshistory.st-andrews.ac.uk/Biographies/Babbage/quotations/

12-14 8:46 比韓 3 跟 4 ，我覺得 1 (일) 跟 2 (이) 類似。

12-13 10:07 The marry/merry/merger occurs only before /r/?

Marry a merry Mary

with ‘Mæri mɛri Meəri’

Which recalls the vulgarism with “Tom, Dick and Harry”.

I can’t think of any other minimal triple.

“mass, mess, mares (non-rhotic)

12-12 8:33 This AI-LLM hair on fire moment is shaping up to be a battle polarized into those for those against. I’m in the against camp, because the LLM result in CJK profs not asking me to edit their papers https://quoteinvestigator.com/2017/11/30/salary/. But Mark Dominus and Terence Tao, whom I respect and admire are for, which is a problem.

I also don’t like to be categorized, so I’m going to join the ‘run with the foxes, hunt with the hounds’ camp.

12-11 22:13 $ man git-diff | wc 1520 8932 69395

drbean@ESPRIMO ~/dot $ svn h diff | wc 83 623 5027

12-10 16:30 Check man mkshortcut

12-10 16:28 mkshortcut -d “場所: mintty C:gwin64” -i /cygdrive/c/cygwin64/Cygwin.ico -A -D -n “Cygwin64 Terminal” -i /cygdrive/c/cygwin64/Cygwin.ico /cygdrive/c/cygwin64/Cygwin.bat

12-10 16:25 With mkshortcut -a “/usr/bin/bash -l -c ‘tt ’web -w sdf’ ‘qb -w qt’ mastodon’ &” -A -P -n “StartUp/web,qb,mastodon” /usr/bin/mintty

12-10 15:56 cygwin’s mkshortcut creates Windows shortcut icons on the Desktop or in StartUp to invoke cygwin programs the Windows way.

in C:/Users/Public/Public Desktop and C:ProgramDataMicrosoftWindowsStart MenuProgramsStartUp

12-09 10:22 MS Word is flagging “compare and contrast”!!

12-08 9:01 “compare and contrast”: favorite words after “although”. I think Richard Day, https://www2.hawaii.edu/~rday/ took “but”.

12-08 9:01 “compare and contrast”: favorite words after “although”. I think Richard Day, https://www2.hawaii.edu/~rday/ took “but”.

12-08 7:21 LLM can read text but can’t read minds, so they don’t know whether it should be “a NOUN” or “the NOUN” . That’s in the mind. “You need to be a mind reader!” #AI #LLM #mindreading

12-07 20:57 https://books.google.com/ngrams/graph?content=gain+success,earn+success I prefered the unpopular alternative!

12-07 20:13 “Immanent”: a word I didn’t know though I think I had encountered it. I think I thought is was a blend of “imminent” and “emanate”, an “eminent imminent emanate.”

12-07 9:16 https://english.stackexchange.com/questions/52009/in-person-equivalent-for-inanimate-objects says no to “in person”. But what about “surveillance capitalism (Zuboff, 2019), in the PERSON of Google, Facebook and others, sweeps up their data to predict, target and shape their behavior.”.

Because of corporate personhood.

12-07 9:13 https://english.stackexchange.com/questions/52009/in-person-equivalent-for-inanimate-objects says no to “in person”. But what about “surveillance capitalism (Zuboff, 2019), in the form of Google, Facebook and others, sweeps up their data to predict, target and shape their behavior.”

12-05 16:39 “High-jumper, Dick Fosbury flings/heaves self over the bar”. Which is it?

12-05 16:29 I never felt the need to distinguish “heading” and “title”, but in a table in a paper,

title: Table 1 Overview of the ESS service providers

heading: +————+————-+————-+————-+————-+ | Items | GoShare | | iRent | WeMo | | | (multiple | | | | | | types of | | | | | | vehicles) | | | | +============+=============+=============+=============+=============+

12-05 16:22 What’s the connection between “sling” and “slide”, other than the connection through “swing” of playground equipment? Perhaps that connection is informative.

12-05 16:14 Difference of “heave/sling the backpack up on the truck.” “heave” if it travels in a straight line, “sling” if it travels in an arc?

12-04 8:47 Han Eysenck, fabricating psychologist, according to https://retractionwatch.com/2025/12/03/number-of-unsafe-publications-by-psychologist-hans-eysenck-could-be-high-and-far-reaching/ See my http://drbean.sdf.org/HigamusHogamus.html

2025-11

11-28 15:16 Don’t do this: have a dir and a file of the same name as the dir plus an extension, and have files in the dir with the same extension.

My script was returning:

Unable to open /home/drbean/dot/screen/dot.rc.

But it was not ‘dot.rc’ that was missing, it was a file in dot/, .rc, where SCREEN was an empty string, that bash could not find. #gotcha #bash

11-28 12:24 Wait! Plants, except for some parasitic and carnivorous ones, don’t rely on biological matter for sustenance, only on chemical elements and compounds.

And they are alive! But not, I think, having intentions or goals, just following progams.

https://en.wikipedia.org/wiki/Plant_nutrition

11-27 22:45 If you make a system that can self-replicate and is capable of mutation in a way that can respond to selective influences, that doesn’t mean it’s alive ..

https://bsky.app/profile/philipcball.bsky.social/post/3m6jtbmsrt22r

If it has a constant non-ending carnivorous need for sustenance, non-consensually provided by other members of the system, it would be alive.

11-27 7:59 https://arxiv.org/abs/2511.19757 argues, “Language understanding entails not just extracting the surface-level meaning of the linguistic input, but constructing rich mental models of the situation it describes.”

This puts LLM in their place. But to this shallow take on language I would say, It’s language all the way down.

https://en.wikipedia.org/wiki/Turtles_all_the_way_down

11-26 7:08 Adam Mastroianni on US political culture divide through a religious study lens.

https://www.experimental-history.com/p/secrets-of-the-ancient-memelords

I wrote something on his “study” of people imagining different consequences. They mostly imagined good or bad consequences?

11-25 23:03 Another plan of attack is how text is more than just the individual sentences in it.

Although LLM create grammatical sentences, does it connect them into text?

https://en.wikipedia.org/wiki/Text_linguistics #text #LLM 3/?

11-25 21:03 Language is not just about communicating statements, but also conveying feelings and attitudes.

Those can be explicitly expressed by the speakers along with the statements or inferred by listeners reading between the lines.

Mind readers try to understand what is meant by what is said and assume there are minds behind the language, which are to be understood. #mind #language #AI 2/?

11-25 17:59 Thinking how to conceptualize LLM language production’s deficiencies (?) despite its amazing Comparing NNS texts with LLM texts is frightening. 1/? #LLM #NNS

11-25 7:50 Dreams are different. In one, I was intrigued passing a non-native street vendor. The visuals were vague, perhaps monochrome. He called out, “What’s your middle name?” He asked what ‘myself’ meant. Then I crossed the street.

Someone could dream that without any visual content? I am sometimes aware of color in dreams. But not touch, smells? #dream 2/2

11-25 7:33 Matt Glassman at https://mattglassman.substack.com/p/moderation-has-multiple-dimension talks about some recent writing on how some people lack mental imagery, so called aphantasia. He calls such differences perhaps a big unknown unknown.

Awareness of imagery differences is not new, no? Do these people hear, but see nothing, when they dream?

I see mental images, memories of places/people sometimes. They’re not vivid. But when imagining things not evident, the thoughts are words. #imagery 1/2

11-24 16:55 Intention is also a wishy-washy notion.

A: Don’t you think you should use public transport?

B: I intend to next week (but I may or may not get around to doing it.)

In the spirit of “The (checks notes) spirit is willing but the flesh is weak,” intention and action are 2 different things.

That’s why the post-Rubicon moment and mindsets are important.

#TPB 3/2

11-24 16:15 There’s no mind behind LLM translations from Chinese to English, so although they do a fantastic job at sentence-level syntax translation, they’re lost with unspoken articles and tense distinctions conveyed from mind to mind in Chinese because the Chinese text gives them no guidance as to how to express them in text in English where they are a grammatical requirement. They can read text, but can’t read minds. That’s my story, anyway, and I’m sticking with it. #LLM #中文 #translation

11-24 9:29 Discussion of “impend/impending” frequency and collocation differences at https://languagelog.ldc.upenn.edu/nll/?p=72043#more-72043 as with “consume time/time-consuming” suggests ‘-ing’ adjectives have a life of their own.

11-24 8:22 Microsoft, ex-Red Hat lead developer explains problem wearing different hats at https://bexelbie.com/2025/11/20/if-you-are-wearing-more-than-one-hat advising conservativeness

11-24 7:58 “just”, typed with the left hand one key to the right, spells “judy”

11-23 8:27 “sling the backpack up on the truck” versus “fling the towel around the waist” https://www.english-corpora.org/coca/ has “sling/fling the ball” and “slung the strap/flung the door” as top cites. No cites of my examples, so no diff? For balls, sure, but “sling the door/fling the door”?, surely not.

11-22 23:00 “I’ll get to it soon” means you better do it within 3 seconds.” “better” as modal verb, anyone?

I can remember at the age of 5(?) using it in a modal verb’s full range: “You better stop that, bettern’t you”, probably in response to my mother using “better” as a command. I stopped doing that. Did I realize it’s a contraction of “had better”?

11-21 22:57 ソートアルゴリズムオンライン学習教材 CardAlgo https://cardalgo.csle-lab.org/

11-21 22:55 Bubble and selection sort algorithm practice site in Japanese: https://cardalgo.csle-lab.org/

11-20 18:10 But who of it is said, “I/you/they couldn’t see the wood for the trees!”? “I couldn’t see the wood for the trees!,” or “That’s like being unable to see the wood for the trees!”? It’s an invitation (or a notice it’s time?) to flip the script?

11-20 18:05 I think the forest/tree idiom is used to say, “Your view’s plausible, partly right, even, but wrong,” with the same force as Korzybski’s “the map is not the territory,” and Mencken’s “every complex problem has a solution that is clear, simple and wrong,”and Shakespeare’s “there are more things in heaven and earth than dreamed of in your philosophy.”

11-19 17:52 I think I was wrong about the woods/trees idiom being about not accepting what’s staring you in the face, under your nose, & not putting 2+2 together, but perhaps it is about framing the problem, or getting the wrong end of the stick. It’s certainly about recognizing you have a grasp of some aspects (details) of the problem, but not the most important ones (the big picture). What role does the idiom have in a conversation?

11-19 15:10 “see the (woods?|forest) for the trees” in google books: https://books.google.com/ngrams/graph?content=see+the+wood+for+the+trees,see+the+woods+for+the+trees,see+the+forest+for+the+trees

11-19 8:55 It’s used when the means to understand the problem lie under your nose, but you fail to put 2 & 2 together. Googling shows more literal interpretations, ‘trapped by details,’ ‘too involved in a situation,’ ‘deep in granular details, failing to get the big picture.’ Perhaps they’re better. But the idea is that you have a grasp of a problem, but it is a problematic grasp? That you have a framing problem?

11-19 8:35 The explanation at https://www.electoral-vote.com/evp2025/Items/Nov18-1.html of the idiom, “Can’t see the forest for the trees,” is that in a forest, you only see individual trees, not the whole forest from an aerial view. I think it stems from a conversation with someone who doesn’t understand forests.

A: See that forest over there? B: I only see trees. It’s behind the trees? A: Oh, you can’t see the forest for the trees?

The idiom captures a problematic failure to understand a problem.

11-16 17:51 What’s the relationship between price stickiness and price elasticity in economics? The first is result of operators’ unwillingness to cut prices if it means making less money? The second is users’ willingness to accept that unwillingness even if it means they have less money?

11-15 9:41 “The program was executed on a computer equipped with an Intel(R) Core (TM) i7-10700 processor, and a Windows 11 operating system.” Why should it be “the Windows 11 operating system”, but not “the Intel(R) Core (TM) i7-10700 processor?” The processor is the named instance of a class that has the same nameas the instance? With the OS, an instance of a class is not the relationship involved? There are not many Windows 11 OS of which this one? See the “a Ford → Ford Company” relationship.

11-13 20:05 “It may be anywhere in this room,” means it is somewhere in the room. But, “It may be somewhere in this room,” means it may also not be in the room, ie, not anywhere in the room.

11-13 16:58 The prepositions, “like” and “such as” discussed at https://english.stackexchange.com/questions/19370/formal-alternative-for-like-and-such-as

11-13 16:50 “some” and “any” discussed at https://english.stackexchange.com/questions/19364/difference-between-somewhere-and-anywhere/19371#19371

11-12 8:56 The problem reading this paper by authors whose first language doesn’t have ‘a’, ‘the’ and plural nouns is knowing whether ‘the reformulation/linearization technique’ refers to one technique or a class of techniques. I don’t know what a/the reformulation/linearization technique is. https://optimization.cbe.cornell.edu/index.php?title=Mixed-integer_linear_fractional_programming_(MILFP)#Reformulation-Linearization_Method suggests it is a specific method, but grammatical errors in the exposition there mean it might not be.

11-11 14:56 Looking for instances of “controls the decision” on https://books.google.com/ngrams/graph?content=controls+the+decision I came across https://www.google.co.jp/books/edition/Free_Will_as_an_Open_Scientific_Problem Interesting. Mark Balaguer is a philosopher

11-08 17:37 I think I now understand the difference between bash’s login, interactive, and non-interactive shell. Non-interactive shells are not connected to stdin or stdout, whereas interactive shells are, but with limited functionality, eg /etc/profile is not read. Login shells get all the functionality bash can provide. Non-interactive shells are for scripts, interactive shells are for sub-shells and login shells are not sub-shells, but the shell you want when you login. The name’s just confusing.

11-07 10:10 Matt Glassman, political gaming thinker and inveterate “gambling degenerate”, with https://mattglassman.substack.com/p/ten-thoughts-on-sports-betting and https://mattglassman.substack.com/p/book-review-on-the-edge-nate-silver , made me realize my understanding of insurance as gambling with positive expected value for the insurer, but negative expected value for the insuree, got that point correct, but missed the importance of the variance in possible value outcomes.

11-05 17:19 Daniel Dennett

Read

• https://languagelog.ldc.upenn.edu/myl/ChomskyContraDarwin.pdf after reading
• https://languagelog.ldc.upenn.edu/nll/?p=69401

11-05 15:34 Nabokov mocks linguistics, applied linguistics, machine translation, AI. (quoted in https://www.cambridge.org/core/services/aop-cambridge-core/content/view/2395C24EB472B60B2514F7D5F93EB9A8/S0140525X2510112Xa.pdf/how_linguistics_learned_to_stop_worrying_and_love_the_language_models.pdf) 2/2

11-05 15:32 Pnin didn’t presume to approach the lofty halls of modern scientific linguistics, that ascetic fraternity of phonemes, that temple wherein young people are taught not the language itself, but the method of teaching others to teach that method, like a waterfall splashing from rock to rock which ceases to be a medium of rational navigation but perhaps in some fabulous future may become instrumental in evolving esoteric dialects like Basic Basque spoken only by elaborate machines 1/2

11-05 13:16 Helping a prof with the English of a Theory of Planned Behavior paper, I wrote about ways of referring to future action in English:

will is kind of used to make a public commitment, like when making a decision and telling someone about it, ie a Rubicon moment.

A: Don’t you think you should use public transport? B: I will use it next week.

going to is different.

A: Don’t you think you should use public transport? B: I’m going to use it next week (but I disagree it’s important to use public transport).

Present continuous with future reference is used to show a planning and action mindset.

A: Don’t you think you should use public transport? B: I’m using it next week (I already use it and/or I agree it’s important to use public transport).

On the other hand, want is a wishy-washy way of showing intention.

A: Don’t you think you should use public transport? B: I want to (but won’t/cannot).

11-05 13:04 In college, in response to a questionnaire about high school subjects, my placing of English high on the list, while at high school I had not had that feeling, rather the opposite, about it, made me skeptical of the value of surveys.

11-02 20:32 “fling”, “sling” and “swing”. #phonestheme, #soundsymbolism and #iconicity, anyone? All describe movement in an arc. Is there any difference between “sling” and “fling”? The difference of being directed at a target, or not?

11-02 17:42 “Not a bug, but a feature”: the thinking which gives us a seamstress mistake as a fashion innovation

11-02 17:38 I looked at Google AIAssist’s answer to “joint theory planned behavior” against my better instincts, and it has a helpful explanation I cannot see in the hits below, showing I need to use “a” rather than “the”. Aaargh. I need to develop more self control!

2025-10

10-27 16:32 Interesting analogizing of Adam Smith’s invisible hand in commerce and the Tao’s 無為 (=take no action, let nature take its course, adopt a laissez-faire approach, accomplish nothing, peaceful, detachment, https://en.wiktionary.org/wiki/%E7%84%A1%E7%82%BA ) by https://www.nakedcapitalism.com/2025/10/how-the-tao-explains-our-economies.html

10-27 7:59 Invention suborning intention.

Intention subverted by invention.

A beautiful intention slain by an ugly invention.

• https://quoteinvestigator.com/2020/12/26/ugly-fact/

Allowing invention to subordinate intention

• https://larrycuban.wordpress.com/2025/10/25/what-the-history-of-supermarkets-teaches-us-about-ai-in-schools-andrew-cantarutti/#comment-95761

10-25 14:19 But what is this fast and slow mind of which Daniel Kahneman speaks?

10-25 13:38 The opaqueness of a dog’s texts on the Internet is mirrored and matched by the opaqueness in real life of the thoughts and actions of people we appear to understand.

10-22 8:56 Two observations from https://philipball86.substack.com/p/what-does-the-turing-test-test : 1. Alan Kay’s metaphor for the surprise of AI is Daniel Kahneman’s fast and slow thinking: The slow mind sees the marionette/puppet as blocks of wood. The fast mind is moved by its gestures. 2. ChatGPT is life-like, but even in 1996 some humans mistook ELIZA for a human.

10-22 8:39 Then https://philipball86.substack.com/p/what-does-the-turing-test-test suggests we may have gotten the Turing test wrong, based on English prof, Sarah Dillon’s observation Turing was kidding a lot of the time.

10-22 8:21 https://philipball86.substack.com/p/can-we-find-a-good-definition-of in opinionated take-down of AGI paper, linked by bsky.app/profile/melaniemitchell.bsky.social/post/3m3plzv2bsc2a notes that AI developers’ aim was to mimic human cognition works against their claim that AI does more than (checks notes) mimic human cognition. But it also puts psychometric testing in a hard place because inferences we make about humans from test results are not justified when the testee .. might not be human.

10-20 15:29 Mmmh? A file opened in ‘cmd’ shows only tofu glyphs for the Chinese/Japanese (漢字) characters and Korean 한글. Opened in ‘powershell’ it shows the 漢字 but nothing is shown for the 한글. 2/2

10-20 15:10 From Microsoft: An account from 10,000 meters and coming in slowly of the development of UTF-8 support for the Windows Console, https://devblogs.microsoft.com/commandline/windows-command-line-unicode-and-utf-8-output-text-buffer/ But did I leave it too late to update ‘cmd’ for it now with the ending of support for Windows 10? 1/2

10-19 17:27 More strange ways to become a father on wikipedia

• https://en.wikipedia.org/wiki/Nathan_Ross_Margold

A Felix Frankfurter ‘Happy Hotdogger’ and 1930’s contributor to NAACP’s legal strategy “married Gertrude Weiner (the couple would go on to have a son)”, “Margold is [also] remembered as the father of adult film pioneer William Margold.” Is it just me, or is “go on to have children” an idiomatic way of describe becoming a father consequent on marriage?

10-19 7:41 I wonder if I should try linking perlbrew virtual environment’s perl to system perl, ie /usr/bin/perl via alternatives. I guess it should be a link to pl.

10-18 22:23 w32 neovim not finding modules installed by cygwin python3 (actually python3.9 as set by alternatives command). To find them, it requires python to be invoked as python3.9, not python3.

10-13 8:46 git apparently breaks hard links. https://unix.stackexchange.com/questions/231955/why-are-hard-links-are-not-updated-when-modified-with-an-editor

I think now I remember reading that somewhere, but it didn’t register trying to understand why stow wasn’t updating linked files after I edited and committed them

10-13 8:42 git apparently breaks hard links. https://unix.stackexchange.com/questions/231955/why-are-hard-links-are-not-updated-when-modified-with-an-editor

10-11 16:29 Typing “balkin.blogspot.com” into the address bar in qutebrowser to read current posts there, I see down toward the bottom of the alternatives. “The Big Picture”, so I guess that’s the keyword or title I was looking for

10-09 17:33 As @OrvarLog@mathstodon.xyz said: When novelty is king, we stand not on the “shoulders of giants”, but on a pile of strawmen

10-08 7:53 I’m trying to remember Jack Balkin’s 4 causes for the US’s problems. They were inequality, polarization, something and what was the fourth: a breakdown in trust, as laid out at https://balkin.blogspot.com sometime at the start of 2025. Findable with keyword, which was? Was response to NYT request.

10-06 21:15 I dreamed someone asked me to talk about my beliefs. I took the opportunity to talk about Laozi (老子), who probably did not exist, Zhuangzi (莊子), who may not have existed, and someone who did exist, but whom I may not have identifed by name or what they did.

10-06 20:54 No, they use it as a joke. It means that they think it could be that sex is a means to an end, but that they don’t think that it is. Were we also supposed to laugh when Socrates said, “Others live to eat, but I eat to live.”?

10-06 17:37 Does men’s use of “tool” to refer to their genitals suggest that they think sex is agentic?

10-03 15:41 https://rss-parrot.net and its rss bridge to mastodon posts are offline. So no rss posts showing up in timelines. This seems like a service mastodon should provide. @FediTips@social.growyourown.services

10-03 9:35 @OrvarLog@mathstodon.xyz said:

Putting too much emphasis on ‘novelty’ researchers are pushed into characterising work prior to theirs as lacking or flawed.

Doing so we are not building our science on the shoulders of giants, but on a pile of strawmen.

That is, “When novelty is king, the ‘shoulders of giants’ are threshed into a pile of strawmen”

2025-09

10-03 7:52 OrvarLog@mathstodon.xyz said:

Putting too much emphasis on ‘novelty’ researchers are pushed into characterising work prior to theirs as lacking or flawed.

Doing so we are not building our science on the shoulders of giants, but on a pile of strawmen.

That is, “When novelty is king, the ‘shoulders of giants’ collapse into a pile of strawmen”

09-27 14:41 Alerted by https://languagelog.ldc.upenn.edu/nll/?p=71138 I need to read https://www.cambridge.org/core/services/aop-cambridge-core/content/view/2395C24EB472B60B2514F7D5F93EB9A8/S0140525X2510112Xa.pdf/how_linguistics_learned_to_stop_worrying_and_love_the_language_models.pdf about the implications for linguistics of LLM language user verisimilitude.

It starts with an allegory of disagreement between Norm (ie Chomsky) and Claudette (Claude, I guess), and is easy to read.

09-26 10:39 『실용 한국어 문법 상급』 안진명, 선은희 에서 11

09-26 9:58 『실용 한국어 문법 상급』 안진명, 선은희 에서 399

09-25 9:52 “dir” and “mkdir” typed with the left hand one key to the right spell “fit” and “mkfit”

09-21 16:46 Google book hits for ‘in a broader sense’: https://books.google.com/ngrams/graph?content=in+a+broader+sense&year_start=1800&year_end=2022&corpus=en&smoothing=3 impressively monotonically increasing.

09-21 16:23 How did I do? The phrase, in a broader sense is used to indicate some form of hypernym/hyponym relationship and synecdoche (?), I think. It is not used to introduce a different thing which is also an exemplar of the category being discussed. There are probably animal and plant terms used this way, Eg, “cats”, “big cats” and, in a broader sense, “the cat family”?

09-21 9:59 My attempt to use “in a broader sense”:

A "car" is a sedan or similar-styled vehicle, but the term may be used *_in a broader sense_* to refer to a wider range of vehicles, eg buses and other vehicles with more than 4 wheels, but probably not motor cycles or bicycles, as in the statement, "Many cars on the road this morning."

09-21 9:04 Checking google AI’s response to “living breathing doll” again to see if it had a different Japanese reference this time. (It did a second time), it said:

It sounds like you're interested in the idea of a "living breathing doll." To give you a more specific answer, I can help you find products, people, or media related to this topic. Would you like me to find out more about interactive plush toys, lifelike reborn dolls, or something else entirely?

That’s unsettling. I am being stalked.

09-21 8:10 『暮らしの韓国語表現6001』、今井久美雄 에서 399

09-20 23:07 I’m going to try to use “in a broader sense” properly and show that AI Assist was not using it properly. I may have misunderstood the point of the sentence in which it occurs. I took it to be a statement about Japanese 生人形, but actually it was probably saying life-like store window mannequins might be called “living breathing dolls”.

09-19 8:52 Agent definitions collected at https://simonwillison.net/tags/agent-definitions/ I’m unhappy about term, ‘agents’. http://drbean.sdf.org/AgainstAgency.html They never discuss the double meaning, http://drbean.sdf.org/DuplicitousAgent.html I’m afraid of the technical AI meaning tail wagging the established, real-world meaning dog.

09-17 9:47 Scene: A car at the bottom of a cliff, with blood splattered everywhere

09-17 8:55

language as a general cognitive function

3 keywords:

• memory
• imagination
• analogy

09-17 8:22 I need to learn how to use readline better. It has almost 800 lines in ‘man bash’, around a sixth of the total 4452 lines

09-16 21:05

alias arrow="echo -n → "

$ arrow hey howdy ha
→ hey howdy ha

drbean@ESPRIMO ~
$ echo -n hey ; arrow howdy ha
hey→ howdy ha

drbean@ESPRIMO ~
$ echo -n $(arrow howdy hey) ha
→ howdy hey ha

drbean@ESPRIMO ~
$

I guess this is marginally useful for something

09-16 20:20 I might make a bash alias for a Unicode right arrow. I can find the code point in vim digraphs, though there are a lot to scroll through. It’s probably just ‘->’.

09-16 17:59 “bash unicode typing” wasn’t much use, but “mintty cygwin terminal unicode” led me to the mintty manual, where I found I didn’t have the Ctrl+Shift+letter shortcut option enabled

09-16 17:16 Googling “bash digraphs like vim” the first line of the AI Assist answer says something about creating aliases. I want to resist reading further but the hits are unhelpful, being all about vim, not bash. I will change the search string to “bash unicode”. ‘man bash’ only offers ‘\uHHHH’. Ugh

09-16 8:01 I don’t think Melanie Mitchell’s magical thinking (https://aiguide.substack.com/p/magical-thinking-on-ai) and Murray Shanahan’s roleplaying (https://www.nature.com/articles/s41586-023-06647-8) bring much to the table, explaining the reaction to or the nature of LLM, but Alison Gopnik’s stone soup story (https://simons.berkeley.edu/news/stone-soup-ai) is illuminating, inspirational because it takes LLM apart, showing users are being fooled, and still shows respect.

09-15 23:07 The reference to iki-ningyo (生人形) or ikiki-ningyo (行き来人形) is completely gratuitous, I assume included because the literal translation is “living doll”. (Though, I’m wondering if the Japan reference is because google knows I’m in Japan from my IP address.). They were exhibited waxwork figures and are not mentioned in the Japanese wikipedia doll page, but are at https://ja.wikipedia.org/wiki/%E7%94%9F%E4%BA%BA%E5%BD%A2 The English wikipedia page has more gory details than the Japanese.

09-15 ??:?? There’s much wrong with this unstable AI Assist. The 2 sentences are 2 bites at the apple, failing because they are saying essentially the same thing twice. The verbs in the 2 sentences, “describe”, “seem”, “refer,” “possess”, “be”, “have”, “seem”, “come to life”, “refer” again, “characterize” are weak verbs. Without meaningful verbs, the effect is of having a series of noun phrases hanging together grammatically, but failing to communicate semantically anything more than they would have as a list.

09-15 13:?? Ha. gmail is putting notifications from me as an administrator of my google blogger.com blog to me (under another id) inviting me to be a poster in my spam folder.

09-15 10:?? On 09-15 10:16, google AI overview returned, “Living breathing doll” is a figurative phrase or idiom that describes a person who is exceptionally beautiful, charming, or endearing, often to the point of seeming unreal or like a perfectly crafted doll come to life. It can also refer to the “Living Doll” aesthetic, a beauty trend characterized by a hyper-feminine, doll-like appearance, and in a broader sense, can refer to life-sized, lifelike mannequins or dolls, such as the historical Japan iki-ningyō.

09-15 9:?? “Living breathing doll” is a song title, a later rendering of the earlier song title, “Living doll”. ddg returns “breathing doll” doll hits, but google doesn’t on the first page. The first seems to be a title for a NY Post article about a Japanese robot in 2005, a second a quote from a book in 2017, and the third the song title in 2019. Given this lack of depth for the phrase, I suspect the google AI overview of the phrase is “extrapolating”, or playing fast and loose with the facts.

09-15 8:50 My analogy for this AI hair on fire moment was Clever Hans, because people were amazed at the horse’s ability to do arithmetic in German. Ny new analogy is #ventriloquism and people’s amazement a dummy could be a living breathing doll.

09-15 8:14 An LLM is designed to generate statistically likely responses to the question “What would an answer to this query sound like?” This is not the same thing as answering the question. It might produce what you are looking for, or it might not. This is one reason why output from an LLM will sound authoritative even when it is wrong, and apologetic when mistakes are pointed out. It is not authoritative or apologetic, and it is not “thinking” about the question, says @mcnees@mastodon.social

09-15 8:01 I was distressed that version control was not seeing changes I made to my blog. Then I couldn’t understand why the changes were appearing in the repo. Then I remembered my blogging script commits changes I make to the blog to the repo. The feeling was like wandering around, lost, for a time, and then coming across a place that seems familiar, because I’ve already been here. LLM hallucinations, anyone?

09-14 21:36 Painting yourself into a corner:

Walking on air, anyone?

Walking on water, anyone?

Walking on paint, anyone?

Watching paint dry, anyone?

09-11 22:36 sponge(1) is counterintuitive. The man page says, “sponge reads standard input and writes it out to the specified file. Unlike a shell redirect, sponge soaks up all its input before writing the output file. This allows constructing pipelines that read from and write to the same file.”

But this code,

< file.txt sponge | tac > file.txt

truncates the file. This code,

< file.txt tac | sponge file.txt

works. I don’t understand. At the end doesn’t seem like the right place for it. By then the file’s already truncated.

In thread:

joe wyohjoe@mastodon.sdf.org

@drbean In the first one, you’re using shell redirection (the > file.txt part) to write into file.txt. All shell redirections are opened by the shell before any of the commands are invoked, which is why file.txt gets truncated.

In your second example, you’re passing file.txt as an argument to the sponge command, and the sponge command opens the file itself, which occurs later in processing which is why that doesn’t truncate it.

09-11 9:13 Instead of saying, “was to have 3 children,” and “would go on to have 3 sons,” they should just say, “and they had 3 children”

09-11 8:42 Strange ways of becoming a father on wikipedia

• https://en.wikipedia.org/wiki/Anthony_R._Michaelis After World War II Michaelis .. married Ann Aikman, with whom he was to have three children.

• https://en.wikipedia.org/wiki/Pierre_Bourdieu In 1962, Bourdieu married Marie-Claire Brizard, and the couple would go on to have three sons ..

09-08 10:36 Halliday’s ‘had’ + past participle as ‘past in past’, as opposed to ‘have’ + past participle, which is ‘past in present’.

09-08 10:28 I had had my breakfast when she woke up. She became angry that I had had my breakfast.

vs

I had my breakfast when she woke up. She became mad that I had my breakfast.

In the first she is mad because he already ate while she was still sleeping (she now has to eat by herself). In the second she is mad because he is now eating (maybe they agreed on a diet the night before).

‘had had’ by https://www.reddit.com/user/Causative/

09-08 10:22 she, where I had had ‘had’, had had ‘had had’. ‘Had had’ had been advised, and ‘had had’ had been chosen.

09-08 10:00 qutebrowser history to the rescue! I read the characterization of language as compositional, systematic, productive in https://aiguide.substack.com/p/an-ai-breakthrough-on-systematic that I probly found in https://aiguide.substack.com/p/do-half-of-ai-researchers-believe. But where I had ‘subsitutability’, https://melaniemitchell.me/ had ‘systematicity’. I stand by my choice.

09-08 8:26 賈惠京(가혜경) 의 『한일 유사 속담 관용구사전』 527

09-07 17:18 I just saw this, but can’t recall where, Distinguishing characteristics of language:

• compositionality, Basically, the whole is the sum of the parts,
• substitutability, ‘Dog bites man and man bites dog’ ‘我愛她和她愛我’ https://chineseedge.com/how-to-say-and-in-chinese/,
• productivity, You can say anything with language.

They forgot grammaticality? Or did they.

09-06 17:53 김현근의 『クイズで学ぶ韓国語』 16

1. 몇 번도
2. 몇 번이어도
3. 몇 번으로도
4. 몇 번이라도

09-05 9:22 The real question is not whether machines think but whether man does–https://quoteinvestigator.com/2025/09/04/machines-think/

09-05 8:54 Man should think. Machines should work–https://quoteinvestigator.com/2025/09/01/people-think/

09-04 21:56 間章のきかけでは1975年にレイシが最初来日した。今度、1986年レイシーは富樫雅彦の音楽生活三十周年の記念コンサートのために、また来日したと言ったが、富樫雅彦に関してその他のもの何も言わない。

09-04 21:28 ジャズ．アヴァンギャルドは日本のフリージャズに触れることの一つが広島リアル．ジャズ集団の出したLunatic Records 001 と002 が話題になったステイーヴ．レイシーの章だ。Lunatic 001 はただ「日本を代表する四人の若手ピアニスト」のものだけですむが、 レイシーの Lunatic 002 にある「Blues for Aida」が清水を間章に関する7（七つ）行を書かせるようになった。

09-04 10:11 前にジャズ．アヴァンギャルドに日本のフリージャズが出ない理由は清水俊彦が日本人のミュージシャンを蔑視すると思った。今はかれはそういう風にした理由は「触らぬ神に祟りなし」がましだとかんがえたと思う。

09-03 8:02 Look at this Google Books comparison of frequency of “time-consuming” vs all other permutations of “time” and “consume_INF”, https://books.google.com/ngrams/graph?content=time%3D%3Econsume_INF%2Cconsume_INF%3D%3Etime&year_start=1800&year_end=2022&corpus=en&smoothing=3

09-01 9:47 @jensorensen@mastodon.social envisages reading AI summaries resulting in the death of the Internet. I’m not guilty of this, though out of the corner of my eye I can see google’s AI overview has the answer to “input string sed” that the hits below only tease with intros like “Read ..”, “Learn ..”, The lesser sin I’m guilty of is scrolling through google hits to find the answer to my question, “Use a <<< here string” without having to visit sites, which may or may not have that answer.

09-01 9:06 Toot posted: https://mastodon-japan.net/@drbean/115126121722591558 ジャズ・アヴァンギャルド クロニクル1967-1989著者 清水 俊彦 のタイトルが気にならない。なぜならジャズはカウンタカルチャの一つだ。アヴァンギャルドはカルチャの現象だ。カウンタカルチャの現象ではない。

2025-08

08-30 22:51 https://simonwillison.net/2025/Aug/29/lossy-encyclopedia/: An encyclopedia metaphor for LLM. Mine is a bag of statements, cut up into phrases, nouns, verbs, all linked in one big tree, with all the google search keywords ever used to find the statements. We’re used to google finding things on the Internet with the right search terms. This is just a step-up from that. LLM are mixing and matching statements & pushing one out. The surprising thing is that the statement is native-speaker-like text.

08-30 17:17 https://www.baeldung.com/linux/sed-with-string

was helpful explaining how to pipe string to sed without using ‘echo’ and how pipes and process substitution play roles.

title="$(sed -nE "1s/^(([[:graph:]]+[[:blank:]]+){4}).*$/\1/p" <<< $story )"
html="$(pandoc -f markdown -t html <<< $story )"

08-28 22:44 The Weizenbaum wikipedia article, https://en.wikipedia.org/wiki/Joseph_Weizenbaum, explains his ambivalence on AI with his distinction between deciding and choosing, a distinction I had not recognized. Deciding, being calculating, could be calculated, but choosing, could not be. Being based on values, ie, non-rationality, choice, on the basis of tasks and criteria, can only be creative human acts with consequences. Deciding and choosing, a distinction without a difference?

08-28 17:29 That said, @melaniemitchell‘s bringing in Weizenbaum, a chatbot creator’s misgivings, like Frankenstein’s horror, about what they’ve done, is refreshing, the other AI bros going, ’La-la-la I can’t hear you.’

08-28 16:28 That is, we don’t know why she asked him to leave the room. It may have had nothing to do with her interaction with the chatbot. We never hear the other side’s story.

08-28 12:11 Alerted by https://bsky.app/profile/melaniemitchell.bsky.social/post/3lxcsdxrv5c26 to Eliza chatbot creator, Weizenbaum’s mixed feelings about AI, I challenge his disingenousness in being disturbed by his secretary’s being sucked in (?) by the bot, and asking him to leave the room. That is what he was trying to achieve with it! Perhaps she was only using it to say something about him, or to adjust her dress!

08-28 11:32 Going to try to commonplace book my mastodon posts first #POSSE

Older mastodon posts rePOSSEssed at MastodonPostRePOSSEssion

Read it again at