proposal

I was being facetious -- we don't have grass, here (unless you are a glutton for punishment and plan on reseeding twice yearly)

Yes. A grouping of values. The order of the values is significant like (x,y) for Cartesian coordinates or (Mother, Mary, 37) for a dictionary that stores (<type_of_relative>, <name>, <age>) tuples.

A C struct.

Then your cost varies with the length of the list.

If, for example, you knew the "key" was a small, zero-based integer and the "value" was a 6 character identifier, then you could elect to implement the dictionary as a simple array where the key was the index into the array. Insert, remove, lookup all become constant time operations.

OTOH, if the key was an arbitrary length string and the value was yet another string, each tuple consumes a different amount of space so such an optimization isn't possible. If you dynamically allocate space for each tuple and link them together (list, AVL tree, etc.), then you have to move through the structure to examine, insert, delete, etc. tuples.

And, have to run a GC to deal with the inevitable memory fragmentation when a 5 character entity is deleted and a 45 character entity replaces it (what happens to the space those 5 characters consumed?)

With an AI coder, you can have it implement multiple different solutions, each constrained differently, and dynamically chose which to use, based on "hints" regarding the types of keys/values, the number of tuples supported, the relative frequencies of insert/remove/lookup, etc.

Otherwise, you are writing each of these implementations by hand.

AI search is great.

Fix capitalism? How?

You don't like the faucsts and microwaves at Home Depot? The VNAs and oscilloscopes on Amazon?

Right, it's unfair for people to invent things or start companies.

What do you do?

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

What I have come to conclude, is that people who are against copyright protections never actually /create/ anything, and find it unfair they have [1] to pay for their internetworked entertainment.

Regular people who want entertainment and don't have budget for anything on the other hand have no problem going to the park, play a bit of soli- tary [2], or just visit friends and family.

These people [from the first paragraph] will always find excuses for their leeching behaviour [sic], and A.I. is just the latest fad for them.

Best wishes, and happy forming of new companies!

Example, from NewsFleX: struct urldata { char *scheme; char *user; char *password; char *host; char *url_path; int port; char *localpathfilename; char *localpath; char *localfilename; }; Add whatever you like

char *strsave(char *s) /* save char array s somewhere */ { char *p; p = malloc(strlen(s) + 1); if(p) strcpy(p, s); return(p); }/* end function strsave */

extern char *strsave(char *s); struct object (as many char fields as you like). free(object -> field_1); field_1 = strsave (new_data);

etc etc

Look at the NewsFleX source.

You always have to know what you do and what it is you are coding for, and what the restrictions are.

AI likely has no clue even about Usenet, and you cannot easily verify it knows all.

I wrote NewsFleX using the rfc977 . Later rfcs added new things, have not bothered much with that, html postings are sort of gone from Usenet these days too.

postings.c: /* from rfc1036.html */ url.c: /* parse URL (from rfc1738.txt */

Linked lists, structures, so flexible, what more can you want?

I wouldn't say that. I asked it about john larkin & his company & it brought up Usenet and sci.electronics.design, not me.

In a dictionary, you would select which field(s) to use as keys (indexes) to access individual records. The keys are typically unique within the set of tuples -- so *a* key produces *a* record. Much like looking up a word in a dictionary produces the definition of that word.

Now, what happens when you want to delete the array and replace it with another that is larger or smaller? The free()-d memory may not support the new allocation. So, you end up with allocated arrays interspersed with unusable space.

If the arrays are always referenced indirectly, you can relocate them to allow the interspersed space to coalesce into a larger section. But, now you have to stop while you run this collection process; the next attempt at allocation (that failed because there wasn't a large enough chunk of free memory) suddenly appears to take considerably longer than it should (due to this overhead).

If you move this activity into another thread, then the collection can happen earlier AFTER the most recent activity, so the next action happens in the time expected.

If you move this activity into another process (on another processor), then the action can happen in parallel with your normal activities.

If you leave the implementation "in-line" in your code, then your code (i.e., executing thread) has to incur the penalty.

It's considerably more involved than that! "Programmers" tend not to understand the hosting hardware. So, they set up algorithms and data structures that don't always fit, nicely, with the optimizations the hardware has put in place. E.g., an algorithm that hops from one cache line to another instead of designing data structures that can coexist in a *single* cache line.

Because they typically won't see the impact of this (because every run of their algorithm is equally brain-dead), they don't realize the performance improvements that they are missing; the processor just looks slow.

At least an "AI programmer/coder" can be smart enough to understand these issues and code to exploit them -- much easier than trying to teach your run-of-the-mill (human) "programmer" about them -- and HOPE he keeps them in mind each time he designs an algorithm!

And willfully misunderstand simple economics. They'd rather live in a workers' paradise like North Korea or Cuba.

you can be pretty sure that almost every woman on youtube/tiktok/facebook/instagram is using "filters" that make them almost unrecognizably "perfect", that came well before AI got big

Op 21-9-2026 om 18:26 schreef john larkin:

It's not that great. It's marginally better than google, but that's not so hard, since google is crap.

Fix capitalism by optimizing a framework that ensures fair competition and a level playing field, so there are actual market forces driving down prices and ensuring optimal product quality. If you just leave it up to corporations they will pretend to be competing while they are actually just eliminating the competition. Like microsoft when they buy skype. They didn't buy it to improve it, they bought it to eliminate it, so they no longer have to compete with it. If you have corporations that are so large they can buy their way into the government or they operate internationally, so they can evade local laws, there is no longer any real capitalism and they just go through the motions while having a monopoly or having so much power that they can avoid actually competing in a fair way. There is only real capitalism if there is a large number of competitors that are in actual competition with each other.

You think climate change is a hoax? You think we should increase productivity to ensure we completely screw up the environment as soon as possible?

You can invent things as you please, but numbers belong to the public domain and that simply means that copyright is nonsense as far as it applies to digital content. We need a solid framework of human rights in order for people to share the wealth and to ensure that people who become wealthy are only able to do that in proportion to their contributions to society. We don't need copyright for that, but we do need human rights and it would simply be unacceptable if a company becomes very rich with activities that clearly violate human rights (like slavery or the copyright scam). Given the current state of technology, we should rethink economic principles and how to ensure that all of society benefits from the shared cultural wealth we have at our disposal. Keeping information artificially scarce by means of outdated copyright nonsense is contributing to mass stupidity and is a way bigger existential threat than AI could possibly be. I'm not saying I have all the answers, but I think AI can potentially help humans restructure society in a more sensible way, because we have evidence from a historic perspective that humans have not always come up with the most sensible economic practices. Science has eliminated a lot of idiotic human behavior, like burning witches at the stake and similar nonsense. But we have to remember that humans are not more intelligent than other primates. Like if they grow up on a deserted island brought up by apes, you can't expect them to independently come up with something like calculus. So human intelligence is a matter of having huge populations of humans screwing around for many generations and collectively accumulating scraps of knowledge. Humans can then appear reasonably intelligent if they manage to absorb a substantial part of that shared collective intelligence.

What do you do?

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

I collect numbers.

The usual patern is that pople who don't produce things are fans of redistributionist philosophies which would extract their "fair share" from people who do make things.

I think that is called social justice.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Sharing is caring.

North Korea isn't any kind of worker's paradise. It's just one more family run autocracy (like the the one that Donald Trump is trying to impose on the US).

Cuba is slightly more democratic, but since the US has been trying to subvert it and crush it economically since Castro took over, the standard of living there isn't great either.

Question. Was it OK for the K.G.B. to subvert and crush economically, if not outright use terrorist tactics to take over otherwise healthy economies during the reign of the Soviet Union?

A less usual pattern is that people notice that societies that have very unequal income distributions don't do as well as countries that have more nearly unequal income distributions.

Unsurprisingly, the US touts statistics that show it as doing remarkably well.

Disinterested observers are less impressed.

Social security is a device to getting the population as a whole to give enough money to the poor to make sure that their kids grow up into healthy and educated contributors to their country.

The US prefers to out-source manufacturing and bribe well-educated citizens from other countries to immigrate to the US to solve US problems. It's still a migration magnet for the less well-off, but it has become less attractive to the highly educated.

For most of human history there were a bunch of different human species. The Neanderathals and the Denisovians are now extinct. H*mo sapiens is now busy killing off itself.

The human population may not have peaked quite yet but it's likely to start declining quite soon (and it has to - the planet can't sustain the current human population).

The question is whether we can change our habits enough to create a sustainable population. John Larkin doesn't look like a good choice for inclusion in that population.

Obviously not. Their aim was to impose a particular political system on the countries involved. The US inspired coup that put the Shah in charge of Iran was equally foolish.

Indeed, that is the point

Well we live in Giga bytes memory days here,

Depends, that is a compiler task as well as the operating system's task to make efficient use of the memory space. There are many different operating systems and hardware systems some a bit more efficient with memory use than others. This all runs great on a 386, a 486 and on ARM Never had an 'out of memory' situation. If that happens malloc() returns 0, strsave() returns 0, and NewsFleX displays an error message.

How you write it (by hand) depends on the understanding of the application. Keep it simple.

In the Usenet example postings and headers normally have a finite and well know structure. It is not impossible to crash NewsFleX, I have used some fixed array sizes at some point. But it only happens if AI tries to poke into it with its crap postings.

Basically we agree.

Largely a time saver, using gcc, I did some - and still do some PIC - asm programming I remember at works the day the boss said: "Now every body should learn C". Easy for me as I was running C80 on my Z80 system at home already.. And I also had written a CP/M clone in Z80 asm..

It all depends DO NOT OVER-ESTIMATE AI and do not under-estimate it either. Hacking I have done too.. Do not under-estimate humans !