[FRIAM] What is FRIAM and is it still alive, also Cognitive Offloading

Wed Aug 12 17:52:02 EDT 2026

too many good things to respond to, and I keep putting off saying anything until I get my head into order, which of course I never will

also too many good things to read, especially when I am stuck and looking for a time-waste

so in no particular order

Marcus had a couple of posts that shook me out of my normal unimaginative mindset and reminded me of ways I should see this, from arguments I have already been making for a long time but put out of mind:

— there was the reply that got to me on Aug 2nd, which began with “I use both Claude Code and Codex heavily, ….” in reply to my original post about junk papers being written by people who are effectively imprisoned workers in academic junior positions, quite close to the call-center prisoners in SE Asia, and Marcus’s argument that the solution to problems with LLMs is more LLMs: that paper writing and reading is all, going forward, to be machine-first and human-(when?)…  But crucially with examples of how, for code architecture, the better tools are doing a very good job of seeking and achieving “grounding on factual things”.  And then another post on Aug 4th in reply to Glen’s endorsement of micromanaging, to which Marcus’s answer (as I read it) was: diminishing returns (or already diminished).

the thing I wanted to send back in reply to these posts is the argument for opportunity in all this (so, signing off on Marcus’s objection).  I have been trying to sell people on the argument, for many years now, that “complex systems” or “complexity science” isn’t characterized by any particular domain or phenomena of inquiry (nobody would argue against that), and just as much, that it isn’t characterized to an important degree by any specific attributes or aspects: being non-decomposable, or adaptive, or evolutionary, or whatever (I have many people in mind who would disagree with me here, and say those are the definitive things).  What I argued it was, rather, was whatever is over-the-fence from the most that can fit into human heads.  Used to be that we advanced knowledge by cramming as much of what existed as possible into a mind, and hoping the mind would see an invention.  But von Neumann was about the last of that species that was really a type-specimen.  Most of the low-hanging fruit from that way of doing things has been eaten.  Now we need to be responsible for best-level understanding of more learned-areas that bear on a problem than any person can learn.  But ideas still happen, it seems, within individuals.  So what do we do?  What kind of community can use — properly — more than any of its members can understand?  That was, by my argument, the “new” problem to which “complex systems” was supposed to contribute some new insight or practice.

But in my post that prompted Marcus’s original objection, I was still (by metaphor) only willing to sign off on the use of LLMs as appliances.  The Lee Sedol or Terry Tao approach, where we know what the problem is (playing more advanced Go, creating new math), and the LLMs expose regions of practice that the people-community had never opened, because people herd and miss things.  There, I was being dumbly too narrow.  If the new Big Data is a net over the world, capable of filtering and rendering, it can in some way “cover” everything that is available to be covered.  Marcus’s code-creation cases show that it can do quite a good job of abstracting precepts of architecture.  So the right question for complex systems people is: how much of a model is the already-demonstrated performance, for whatever counts as “understanding”?  Up to now that has been one of these cultural hand-me-down words that we can use with each other, and we don’t need to be able to articulate “what it (understanding) means”, since by exchanging it was can put each other in the intended frames of mind.  But to make good use of it with this non-people-thing, maybe we should do a better job of articulating what “understanding” something means, as it bears on how we want to create.  Or more generally, what kind of life going forward we would find tolerable?

It’s also cicada season, and I am aware, while they sing themselves to depletion and fairly early death all over the streets and parks and buildings, that it is useful to try to deal with the question Why do cicadas do anything?  The problem-word being “why”.  If we can get that right for cicadas, we probably have most of what we need for people.  Trying to do it for cicadas is helpful because it deprives us of the crutch of attaching too much significance to the speaker’s own personhood (my rendering of another of Marcus’s expressions).

Then there was Jon’s reply on Aug 3rd to my poke at Bourbak (which I sort of expected and hoped for).  Apart from many fond memories of RLM, a building that I was told the physics department had designed, and that the lowest of the Soviet era would have been proud of: energetically rejected any kind of elegance or grace and managed still not to achieve functionality; elevators that cluster on all floors, while the Otis employee stands in the lobby and watches them and declares “our elevators do not cluster”, or the escalators on the bottom 3 floors that, when the building opened, had been coordinated raster-wise like those in department stores, but the kid-traffic got so high that the people-flow jammed like gravel in a hopper, and kids were jumping off the sides to the floors below in the Kaaba-crush, or the main lecture hall where big concrete posts obscure the view of the speaker and slides for about 15% of the audience — 

Apart from all that, there were two actual things to say.  Diaconis wasn’t objecting to the insights and content of the mathematicians of the Bourbak school and its descendents.  As Jon says, a ton of that was creative and consequential.  Diaconis was objecting to the aesthetic that mangled (in his view) the effort to communicate, losing faithfulness even to how the creators thought, and putting a needless obstacle in front of others to understand them.  Since Diaconis’s lectures have been, for me, so uncommonly beautiful and self-propelling to follow, I took him (as he wanted to be taken) as an example that it didn’t always have to be a slog, in everything.

But the other point upon reading Jon was that, in a way, the current Techlords’ aspiration to upload themselves into an engineered thing isn’t new at all.  It’s what mathematicians have always wanted to do, and that David Hilbert articulated most coherently:  Can you create a formalism that will carry _all_ of what you mean, so that _nothing_ is left that depends on you and that has not been exported to the formalism.  Can the formalism on its own have metaphysical substance?  Of course, I have no ability to appreciate the Hilbert program the way a mathematician can, so I am out of my lane, and I know it (Glen’s ultracrepidarian).  But suddenly it seems that everything new is old again, and this theme has been there for a long time.

And then finally today, getting only part-way through Celine Nguyen’s piece on taste that Glen sent (before needing to stop fooling around and work), I was reminded of Darwin and Mendel.  Darwin hadn’t thought through the fact that the heredity model as he understood it would be mixing paint and would bleach out the preservation of variation.  It was the Modern Synthesis that brought in Mendel, and said No; it’s like shuffling un-mixable cards so variation can be preserved.  But there is this lovely paper
https://www.pnas.org/doi/10.1073/pnas.0603780103
Collective evolution and the genetic code
pnas.org
(seems not to be paywalled)
that says, even with genes, if there is no bundling into genomes, you go into a “thermodynamic” regime, where there are few attractors but they are very fast and robust, whereas if you block lateral transfer, you go into the “Darwinian” regime where variation is preserved because mixing is inhibited: 

I never liked the reference to integrated big-data/AI/LLM-etc systems as “conscious” or whatever, because the word is bad enough that the main content I can see in it is human projection, and relatively little of understanding, since we didn’t understand it in the first place.  But it does seem right to say it is a net over the world, and that that will be a new thing for us to deal with, unlike anything at this scale and in this area that we have ever lived with before.  The economy, and fossil-fuel-powered global transport, were two much narrower nets over the world, and already we see how they have changed life, and when they get into snags after we have committed all our dependence to them and re-orgznized society and infrastructure and demography and population around them, we have not been good at getting ourselves untangled.  

So the argument that people preserved variety because they were incapable of learning enough to blend it all out (returning to Nguyen), and that now that barrier is removed and we stand to lose a lot, becomes the distillation of a lot of people’s varied worries about commodity use of LLMs to replace everything they find laborious or tiresome (meaning, all of living).  Of course Marcus can rebut that, too, saying that LLMs can be set up to prioritize maintenance of variety as easily as they can be set up to do anything else, an argument we have seen demonstrated on much older technology (which shows whatever it shows): 
https://picbreeder.net/7338

Anyway.  As I said; anyway…

Eric

> On Aug 13, 2026, at 1:12, glen <gepr at ropella.name> wrote:
> 
> It likely matters *how* you present the posts to the machine. Did you simply point it to the archives? Both sets? Or did you download it all and curate it? A good thing to do might be to remove the footer that gets repeated a billion times in an active thread because people refuse to clean up their posts. [grrr]
> 
> The threading indices might be more interesting than the by-date indices:
> 
> https://linkprotect.cudasvc.com/url?a=https%3a%2f%2fredfish.com%2fpipermail%2ffriam_redfish.com%2f2026-August%2fthread.html%23start&c=E,1,c9mtOl3J2x_2DdPKBRWBt624pJITZPSnp2Lejvp5fsGACtEO8OpvtM6guXyt0BeylOd7azscbOjvgrsMp9FDgoUMI9ZOsPMv7Y_WBAGDTf_I&typo=1
> 
> versus
> 
> https://linkprotect.cudasvc.com/url?a=https%3a%2f%2fredfish.com%2fpipermail%2ffriam_redfish.com%2f2026-August%2fdate.html&c=E,1,sQtH77sWTBaqht4ez2QXH5JvWv9UikFIPqxIm5ToU6Nj3gQchweI7bhHFD9elzQvJfwVcsvJwmCbdP17_3Ld_CnwLA6Hry35dX39CXQMXT0gvG_Pfg,,&typo=1
> 
> Of course, Codex could help you reorganize the corpus for each given inference type. The best thing to do might be to gather the corpus and build a queryable database. Then for any given type of inference, the LLM queries the database.
> 
> Garbage in → garbage out ... and it's clearly garbage if the machine thinks I'm funny. cf https://linkprotect.cudasvc.com/url?a=https%3a%2f%2fasteriskmag.com%2fissues%2f15%2fwhy-we-like-things&c=E,1,Hs5H2eHOnTwk3pn-PW0zUBn5mcf-M0hO3PLIY_QFeTMkTBnhJS9ycyTBkbh2lBmUE9V1hdoSkj6J7Y9PeQ6X6fDUR5xHJVimU79t3t5coe4-YCY,&typo=1
> 
> 
> On 8/12/26 8:31 AM, jofr wrote:
>> A simple model is based on timezones. When I post something at 9 AM CET (or CEST) Central European Time, then I usually get no response before 5 PM when it it is 8 AM in California and 9 AM in Santa Fe. The posting times reflect where a person is physically located.
>> The rest is more or less unpredictable. Posts could be correlated to even more posts. Some kind of self-reinforcing bubble. It also reminds me of the famous El-Farol Bar Problem which predicts visitors for a bar. Does this bar still exist in Santa Fe?
>> https://en.wikipedia.org/wiki/El_Farol_Bar_problem
>> In general it seems to be unpredictable. I posted for example less in recent weeks because my mail server provider from New York changed the settings and I had difficulties to access this email account I am using now.
>> -J.
>> -------- Original message --------
>> From: Marcus Daniels <marcus at snoutfarm.com>
>> Date: 8/12/26 5:05 PM (GMT+01:00)
>> To: The Friday Morning Applied Complexity Coffee Group <friam at redfish.com>
>> Subject: Re: [FRIAM] What is FRIAM and is it still alive
>> I was looking for a (null space) model of when and why a post occurs.  Like why a person would reach for the remote and change the TV station.   It may have nothing to do with the coherence or silliness of any particular programming.   Like going out for a cup of coffee or a smoke.
>> *From: *Friam <friam-bounces at redfish.com> on behalf of jofr <jofr at cas-group.net>
>> *Date: *Wednesday, August 12, 2026 at 7:58 AM
>> *To: *The Friday Morning Applied Complexity Coffee Group <friam at redfish.com>
>> *Subject: *Re: [FRIAM] What is FRIAM and is it still alive
>> You mean a "credit model" depending on correspondence length and authorship time? GitHub has a statistics page for repositories which shows who has committed the most lines in what timeframe. And some open source frameworks like Ruby on Rails list their most active contributors https://contributors.rubyonrails.org
>> ChatGPT says about you Marcus that you are the most active poster by a substantial margin. You win clearly in this category.
>> Nick and Glen are funnier according to ChatGPT though. ChatGPT is particularly fond of the Nick's post "All those who think Steve was wrong raise your hands" and Glen's post "Say it again … Say it again in a different way … Again … Again, damnit!"
>> -J.
>> -------- Original message --------
>> From: Marcus Daniels <marcus at snoutfarm.com>
>> Date: 8/12/26 4:28 PM (GMT+01:00)
>> To: The Friday Morning Applied Complexity Coffee Group <friam at redfish.com>
>> Subject: Re: [FRIAM] What is FRIAM and is it still alive
>> Correspondence length and apparent authorship time that follows the distribution of a compile/link cycle or system integration test suite execution duration?
>> *From: *Friam <friam-bounces at redfish.com> on behalf of jofr at cas-group.net <jofr at cas-group.net>
>> *Date: *Wednesday, August 12, 2026 at 4:50 AM
>> *To: *The Friday Morning Applied Complexity Coffee Group <friam at redfish.com>
>> *Subject: *Re: [FRIAM] What is FRIAM and is it still alive
>> ChatGPT proposed these two laws:
>> FRIAM's First Law: We take difficult philosophical problems and worry
>> them to death until someone produces either a profound insight, a
>> completely insane hypothesis, or a phrase like "quantum woo".
>> FRIAM's Second Law: If a theory is suspiciously elegant, somebody on the
>> list will throw a shoe at it.
>> The reason I have asked ChatGPT and Gemini is I am writing a new
>> self-published book, and the first chapter is about FRIAM. Their
>> summaries are surprisingly good. I asked them to find funny anecdotes in
>> the FRIAM archives and they found a post from Nick that said "OK
>> friammers, All those who think Steve was wrong raise your hands" :-)
>> If you do NOT want to be mentioned in the chapter because it is too much
>> quantum woo for you then please let me know or raise your hand. Will
>> share a draft in a few days.
>> -J.
> 
> 
> -- 
> 8647 ⊥ ɐןןǝdoɹ ǝ uǝןƃ
> ὅτε oi μὲν ἄλλοι κύνες τοὺς ἐχϑροὺς δάκνουσιν, ἐγὰ δὲ τοὺς φίλους, ἵνα σώσω.
> 
> 
> .- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-..
> FRIAM Applied Complexity Group listserv
> Fridays 9a-12p Friday St. Johns Cafe   /   Thursdays 9a-12p Zoom https://linkprotect.cudasvc.com/url?a=https%3a%2f%2fbit.ly%2fvirtualfriam&c=E,1,-AHg9AEnF1yI9Ni6veUuQ8iLv4BlNcRsDChEr9C5ej7R-JvakqEzAqlyqN3ZjR7vspGzBitQQEeO2fAx4c25BmD31CtTub0rKtRjU_qt-oT_gkmWGVAEqdXy&typo=1
> to (un)subscribe https://linkprotect.cudasvc.com/url?a=http%3a%2f%2fredfish.com%2fmailman%2flistinfo%2ffriam_redfish.com&c=E,1,J1kjmo0OITj21RfEaiAu2GkGlc4Pos2FIoziOtAla-A5C9MvIEgo18JYiOE_HpT0D7McBpgTrXJadyVL0V0sBPAOv-ygitJKheKWZCL3UeoCpKg3vaQBr1wF-NDK&typo=1
> FRIAM-COMIC https://linkprotect.cudasvc.com/url?a=http%3a%2f%2ffriam-comic.blogspot.com%2f&c=E,1,MPLIXP7YZufyF0pFEzDNQkNQNpO2lT51BrCvhp_O7wcgx4iAPySj6Cs0xmy_zQy6XJKFAfHmNOi1OuwYP7NU5UsLhFOhJ5Q2oSboLPh44SBz7g,,&typo=1
> archives:  5/2017 thru present https://linkprotect.cudasvc.com/url?a=https%3a%2f%2fredfish.com%2fpipermail%2ffriam_redfish.com%2f&c=E,1,I9Bxe2HfwJFv94Kf74g0szY9LcAHbv-3g3rdvXxA1TS0DzcR5_5_SzQks6GzWs42pfAnmqdi5Od9PKpU7bocgd0nN1-zFxC3uggMx-gKyFQEozxC&typo=1
>  1/2003 thru 6/2021  http://friam.383.s1.nabble.com/

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20260813/ba9d47fe/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pnas.2006.103.issue-28.cover.gif
Type: image/gif
Size: 185493 bytes
Desc: not available
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20260813/ba9d47fe/attachment-0002.gif>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: icon.gif
Type: image/gif
Size: 1488 bytes
Desc: not available
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20260813/ba9d47fe/attachment-0003.gif>