AI and programming.

liquibyte · 2026-08-16 20:50:03

What's everyone's thoughts about this post by golinux? Should AI be allowed to help programmers with research or tricky bits of code or should it strictly be verboten? I find that it's a great research tool but apparently some people are dead set against it.

golinux · 2026-08-16 21:52:39

If you know anything of human history, the pattern is to create new tools/toys which eventually come to bite us badly. That is because humans are fatally flawed in their judgment and motivations. What is needed is a transformation of the human mind from never-ending pursuits of froth and foam that keeps us running in place and repeating the same blunders over and over . . .

IOW . . . we need to stop shooting ourselves in the foot!

Perhaps we are what we eat but we definitely become what we think. All the Sci-Fi I read in the 50s is now horrifyingly manifested in our daily life.

We need deep wisdom not a roller coaster of titillation with shiny new s**t to survive the future that is on our doorstep . . .

liquibyte · 2026-08-16 22:37:04

Bro, it's a tool. It's a really useful one too if you control it. I get the fear because for all of human history anything new and good has been used by those in power to control others. Did you say that very thing when the internet came about? Look what they've done to that. We could make amazing things with a computer program as powerful as that and release it to the masses for free for anyone and maybe even better than what exists in the walled gardens. Imagine Linux with a 90% share on the desktop. That's all I'm saying, it's a dead useful tool and if we don't do good with it bad and greedy people will control the world with it. I used it to figure out how to integrate Compton and WPEWebkit into my program and I'm rather pleased with how things worked out. Manually doing that research would have taken me another decade to work out. The code is mine but asking the right questions really steered me to faster results. Am I a bad person for asking for ideas and a little help with the integration path?

golinux · 2026-08-17 00:19:45

I am just attempting to make you aware that "sleeping with the enemy" is a dangerous gambit that could easily have catastrophic consequences for "freedom" and you, personally. Digital knowledge and skills are quickly lost if not used repeatedly. Imagine a future that due to reliance on AI solutions, humans lose their ability to effectively control their digital devices? Game over.

Altoid · 2026-08-17 01:20:10

Hello:

liquibyye wrote:

... a tool. It's a really useful one too if you control it.

Yes, it is indeed a tool.
A badly thought out / designed one.
And yours is also a huge "if".

I'd say it is already running amok.
And as no one bothers to control anything, soon there will be no control possible.
So it will get worse.

golinux wrote:

... deep wisdom not a roller coaster of titillation with shiny new s**t to survive the future ...

+1

Best,

A.

greenjeans · 2026-08-17 01:28:26

“There is nothing either good or bad, but thinking makes it so”
~Shakespeare

golinux · 2026-08-17 01:35:40

The mind is the source of all our woes . . .

greenjeans · 2026-08-17 01:39:09

The mind is the source of all our woes . . .

And the source of all our joys. I choose joy.

“We create most of our suffering, so it should be logical that we also have the ability to create more joy. It simply depends on the attitudes, the perspectives, and the reactions we bring to situations and to our relationships with other people. When it comes to personal happiness there is a lot that we as individuals can do.”

~His holiness the Dalai Lama the 14th in the Book of Joy

Last edited by greenjeans (2026-08-17 01:55:46)

golinux · 2026-08-17 02:06:14

The Dali Lama is right. All of our "suffering" is self-inflicted. Being able to maintain emotional "neutrality" toward whatever comes your way is quite liberating!

IamThatIam · 2026-08-17 02:17:17

I fully disagree with golinux on this. I think that if you are using the AI as a tool and are acting as a software designer or architect and keeping a schematic with blueprints of your software, intelligent questions for the AI, and with a specific design, and also keeping records of what you are doing then there is nothing wrong with using AI to make coding more efficient. I think alot of people will come up with niche clique ideologies to be an elitist and will turn everything to its extreme and try to say if it is bad for AI to be in control of all of our sotware then AI must be bad in all situations and circumstances and for any purpose. I disagree. I think this is just being an extremist and truly a software fascist as well as an elitist. There is nobody stopping people from deciding not to use AI. It isn't being forced upon you necessarily. There is nobody saying that using AI means letting it do everything and not having an intelligent designer in control of the AI. I think when we start making everything into its extreme polarity like that then we fall victim to the ideologies of the brown shirts in Germany way back when. People need to think for themselves and use common sense, and stop letting other people tell you how to think. It does not take a rocket scientist to know that using AI makes coding more efficient and faster and it also does not take a rocket scientist to know that you cannot let AI be in control when coding nor can you use AI and code without being the intelligent designer who keeps records, backup copies, and uses a schematic with intelligent prompts and questions. Everything must be intelligently thought out first and put on paper before you start using AI to code otherwise it is going to do things you do not want it to do. Also I would not use cloud based AI if I could help it. Claude and Grok are not so bad but localized is always better because you can control it and tell it what to do. I would say Enoch is a good way to go. Be the programmer and the designer and not the one just using someone else's service.

greenjeans · 2026-08-17 02:36:38

Oh yeah!

blackhole · 2026-08-17 07:51:40

What many of you fail to realise is that it's all for profit. "AI" is just the last product from Big Tech, it is being heavily invested in, it's a bubble and it will burst at some point. It exists to make money for shareholders. But the immense amount of hype surrounding it, means that your boss is slowly becoming convinced that it can do your job. It probably can't - or maybe it can do aspects of your job well enough, that you can be laid off. You can then run along and babble about "AI" geing a great "tool", while the billionaires laugh at the real "tools" who bought into it all.

As Altoid said, it will run amok, because as with everything else, the risks are all acceptable, and the "running amok" is actually part of the plan - we're now at the stage where the big players want it regulated, so that they can maintain ownership, control and exclude competition. To be seen as viable "AI", and boost share value, LLMs are going to need be "headline news".

Altoid · 2026-08-17 11:05:36

Hello:

blackhole wrote:

... many of you fail to realise is that it's all for profit.
... exists to make money for shareholders.
... while the billionaires laugh at the real "tools" who bought into it all.

+1

Have a read:
https://www.theregister.com/columnists/ … ay/5288275

Best,

A.

Last edited by Altoid (2026-08-17 11:11:13)

RedGreen925 · 2026-08-17 15:36:01

Yet again I see no one mention my biggest problem with the garbage AI. If you use it you are plainly and simply a thief. In the Free Software community we support copyright and those models were trained on stolen code from the entire internet. Now if a company trains an AI on its own self produced or licensed for that purpose code it is a different thing. Though they have lost any claim to copyright when doing so then using it as all code produced by a machine is not allowed that protection. As far as I am concerned all those AI companies should have their executives sitting in jail and all their money taken as that seems to be the solution when an individual does it and infringes copyright. As they claim that corporations are people for their purposes when it suits them and their justification of the parasite behaviour of their actions and its harms towards society.

sgage · 2026-08-17 15:58:21

The notion that a given technology is neutral, just a tool that can be used for good or ill, is not borne out by fact or history. Technologies have their own imperatives, their own tendencies, and 'unintended consequences'. This goes 100-fold for so-called 'AI'. Look at what it has already done...

Last edited by sgage (2026-08-17 15:59:04)

blackhole · 2026-08-17 17:01:56

RedGreen925, I mentioned the plagiarism aspect in another recent post and in the past in other posts here and on other sites.

The likes of Microsoft, the owner of github, are well placed to use the "tools" they are promoting to "steal" code from the massive array of private and public github repositories it hosts. This is in fact a means for proprietary vendors to effectively circumvent GPL licensing.

SS · 2026-08-17 17:06:30

Fascinating topic. I hated the way Microsoft was getting too overwhelming with all the features, forced AI and updates etc on my laptops. Then windows 10 was no longer supported. I would have had to landfill 2 laptops (one with 4k screen and 16gm ram,). I am non technical and busy with my own commitments but curious about this thing called Linux. I asked AI and Gemini suggested Ubuntu , Linux Mint and for my very old 2008 laptop (4gb ram but SSD drive), Devuan. I had never heard of these. It explained Devuan is a fork of Debian created by VUAs. It will ruin flawlessly etc. I took the plunge, not knowing what system d was or anything really and realised Devuan needed a load of config . Using AI , forum help and perseverance I got the old 2008 laptop set up in a few days and now implementing on the 4k laptop. The topic of "init freedom" kept coming up. I had to figure out what that meant. At the end of it I realise that building one thing to do something properly and connect to next thing is brilliant, instead of building a complicated mess with no transparency. I apply that philosophy where I can in every day work now eg the way I will construct a spreadsheet or think about solutions. Without AI making me aware of Devuan's existence, and learning how to do/ not to do things, I wouldn't have got here. But then with AI it's easy to get lazy or degrade quality of output or principles. My wife played around on my laptop and found it so transparent, and thought it was so cool that devuan was founded on values that stand against corporate greed, and a reminder of human ability to be more conscious in choices instead of illusionary freedom, she is switching to Devuan on her laptop. Maybe I am not a desired user of Devuan (not sure), given my use of AI in the story, but the principles behind the distro are what makes me use Devuan as my only operating system of choice (plus it runs absolutely brilliantly. I haven't had a single crash, screen freeze or forced restart. Laptops are running blazing fast. Bluetooth is flawless. WiFi connection is instant and way better. Performance is amazing, wireless printer connects and locks in, no cut out. Nothing undesired, nothing forced on my laptop). The excellence should never be compromised.

Last edited by SS (2026-08-17 18:01:04)

bai4Iej2need · 2026-08-28 17:13:44

https://youtu.be/QJUBKYuVXUc?t=1925

AI gives you a statistical view of a program of "how it could look like, but no function guaranteed."
You have to give a software functional specification so detailed, that you have the job done yourself.
Do A first
Do B second
use the result string for Processing in C
output C Result to a device / file / stream what ever you need.

And then you have anonymously transferred all your know how to an mulltibillion corp which is eating it not even thankfully.

I tested this on several occasions with very small snippets of specifications.

I will not give them my money and my know how.

Xenguy · 2026-08-29 01:34:59

Paragraphs are a thing; please use them.

chomwitt · 2026-08-30 14:17:28

Various views on the LLM usage issue on libre software projects
My thoughts on the LLM stance from libre software projects
Alternative name for AI apps.
personal hardware Taken...
Distributed , federated LLMs
Breaking the internet & the tragedy of the commons
Composing licenses & other impossible tasks for the LLM laundromats
A gaming gpu's salesman AI
Hallucinations and Dreams.
Surf-coding and SloppyGeneratorAccelerator
To Use or not use an LLM ?
Faoustian trades with the ultra rich
Addicted to a LLM Laundromat
Back .. to the 50s?
Eating as much as you can get .... is a healthy (info) died for a digitized-algorithmic neural network ?
Yep... RMS was right
Why the license matter ?
The bigLLM datacenters
Ligitation
refs
Alternatives to LLM slop

Various views on the LLM usage issue on libre software projects

Debian : @Debian's resolution on LLM usage and @discussions on the debian vote emailing list.
FSF@
GCC @ GNU Compiler Collection - AI Policy
LinuxKernel@ Linus Torvald's opinion on kernel emailing list , @ Coding assistant's guide.
LibreOffice @ Libreoffice / 03-09-26 / Yes, no AI is now a feature
Gentoo @ (2024) Project:Council/AI policy
Artix @ (04-06-26) Forum discussion
antiX@ (31-05-26) antiX project AI policy. AntiX is using LLM to speed localization that increasing it's visibility to other languages.
Hyperbola (20-10-25) Hyperbola, a renewal of our rejection towards generative algorithms, neuronal networking and the projections done
GUIX @ (28-07-26) Submitting GCD 009: "Allowing LLM-assisted and LLM-generated contribution
Rsync @ rsync and outrage by Andrew Tridgell Jun 3, 2026 (a retired famous hacker cant find humans to help..)
Firefox @ (09-2026) Latest changes in Firefox AI intergration
GIMP@ (11-03-26) LLM Co-coding and AI Slop: differences explained
NetBSD @ NetBSD Commit Guidelines
FreeBSD@ FreeBSD 2025 status report.
Codeberg @ (230726) Protecting our FLOSS commons from LLMs
SourceHut @August 27, 2026 by Drew DeVault's post on SourceHut policy and it's community discussion on the matter
HaikuOS : @ (250526) On LLM policy of HaikuPorts
SFC @ SFC's Recommendations When Using LLM-backed Generative AI Systems for FOSS Contributions.

My thoughts on the LLM stance from libre software projects

A divergent set of policies on LLM emerge from core free software projects-communities..

As of Linux's stance it seems it is the Torvald's stance!.. which is friendly to bigLLM. Essentially Torvalds whitewashed the bigLLM but giving them some of the magical goodfeel vibe of the most succesfull libre project. A review of Linux's and Torvald's stance from Drew DeVault @.

antiX is a case of a debian based distro experimenting with automating translations using now LLM. An antiX developer builds antix-trans-bot @and from a quick look i had uses transifex.com tokens to translate in Chinese docs and it's wiki. So antiX is currently trying this. But dont know if it's in their wikis and code yet. Also they are worried how the Debian stance will affect them. Also it's interesting that they acknowledge the contradictions between they posing as anticapitalista distro but discussing freely using propriatery LLM to help their project..

Firefox has a bigLLM friendly stance. It differentiates itself from other big browsers by promising a less intrusive 'AI' integration. So essentially it functions as a front to the bigLLMs . For me it's obvious that a firefox and other browsers are the first candidates if a distro wants to have a stance critical to bigLLM . A small-open-distributed-LLM friendly distro should offer browsers that have no AI integration by default.

Interesting that Andrew Tridgell can not find humans. I think that highlight a problem that i have in my mind for some time. How do libre-project die ? What are the vital signs to watch? And if that happened a decade ago (no human help) what will happen in the LLM era. I think (not joking) we need proper obituaries of projects. That way we can learn more from the dead. In rsync case we see that solo projects ,even from great hackers, are in danger. That issue (or rsync adopted by bigLLM :-) ) is related to the bigLLM destroying the commons and sucking attention. Let's not forget that when we think about decisions like : Should i put firefox as default browser ?

@The FSF doesn't usually sue for copyright infringement, but when we do, we settle for freedom . In that post they propose that if someone trains theirs LLM's with libre content the FSF would ask from the Justice as compasation to make their LLM code free also. I havent found a more general FSF position on LLMs! Rather peculiar since RMS is adamantly clear on the perils of SaaSS computing. An the LLM-able software that is proliferating could transform easily a pc with a nominaly libre distro to an actually LLM clouding portal. Also in programs like emacs with extension repos already that is happening.

@Debian's resolution on LLM usage and @discussions on the debian vote emailing list.

Proposal E . Choice 5: Responsible Use of Generative AI
Blindly accepting or uploading AI-generated material without appropriate human review is inconsistent with Debian's established development practices. We enourage our contributors to disclose whether a contribution was made with AI assitance, but do not require them to do so.

How that affects Devuan ? How flexible is Devuan is to navigate its own way given it's relation to Debian?

I cant help to notice that Debian project has Google and Microsoft as debconf25 @sponsors. Google has substituted it results with LLM generated content not to mention what chrome transform to a @AI-agent. Could they possibly make policy changes that misalign with their sponsors ? Just asking. Because while we are complecent AI firms are offering billions to buy chrome to use it exactly as a bigLLM portal. So focusing on just the Debian code practically means little to a distro user. A Distro is a software library. Can it just ignores what it has in it's repositories? Can a distro be just a libre-packaged LLM agent or SaaSS client ?

I think Debian's choice avoid the issue. Debian is a distro. Is it ok if half of your packages use AI while you use them ? A couple of proposals seem to want to contain the policy only on Debian specific code. But i see a distro also as a curated list - filter that already prioritizes Libre licensed software. But arent they afraid that tha basic principle will be diluted by adding LLM laundromat code ? ... they seem just following the wind..Arent they afraid of a Distro @ degrading to a SaaSS portal ?

On the light side .. reading on that discussion on LLM enviroment impact and learning all the trips around the globe the the debian devs do... Wow where do they got so much money from ?.

@ Linus Torvald's opinion on kernel emailing list. But besides what Torvald says Linux Foundation has already aligned with big-LLM in various 'AI' centered projects (Agentic AI Foundation, AI education and certifications,Akrites etc) and of course as big sponsors.

@ FreeBSD 2025 status report. (FreeBSD community seems to have fast reflexes).

Policy on generative AI created code and documentation
Core is investigating setting up a policy for LLM/AI usage (including but not limited to generating code). The result will be added to the Contributors Guide in the doc repository. AI can be useful for translations (which seems faster than doing the work manually), explaining long/obscure documents, tracking down bugs, or helping to understand large code bases. We currently tend to not use it to generate code because of license concerns. The discussion continues at the core session at BSDCan 2025 developer summit, and core is still collecting feedback and working on the policy.

@ NetBSD Commit Guidelines

Code generated by a large language model or similar technology, such as GitHub/Microsoft's Copilot, OpenAI's ChatGPT, or Facebook/Meta's Code Llama, is presumed to be tainted code, and must not be committed without prior written approval by core.

@ GNU Compiler Collection - AI Policy

@ Libreoffice / 03-09-26 / Yes, no AI is now a feature
LibreOffice decision made me think. How can a Distro assume an LLM policy if software with many users like GIMP, Audacity etc can download 'AI' scripts ?

@ SFC's Recommendations When Using LLM-backed Generative AI Systems for FOSS Contributions.

@A quick experiment with vibe coding. September 3, 2026

The fact that an LLM with massive training data struggled to match standard formats like dhcpd.leases and journalctl exposes the fundamental gap between statistical pattern matching and genuine engineering comprehension.[
Statistical Averaging vs. Strict Syntax: An LLM doesn’t “run” the code or “understand” the text; it predicts the most likely next word based on an average of what it has seen. Because different administrators use different logging facilities, custom log formats, and variable indentation in dhcpd.conf, the AI creates a “blended average” parser. In engineering, a blended average parser is a broken parser.

As i argue below , LLM suck at blending semantics. And syntax is formal semantics. And if it suck at computerized-formal-semantics of config files what can we expect from blending gradially more semantically challenging tasks ? Like Licences (that i analyze below).

@August 27, 2026 by Drew DeVault's post on SourceHut policy and it's community discussion on the matter . He feels (as a programmer who creates and sustain code repos) that 'AI' is a threat to his intentions. DeVault's express an ethical view of computing that is against not the the LLM tech but against the forces that control it currently that exhibit blatantly criminal & exploitative behavior against the open source community and workers on general . Also he thinks that enviromental speaking the way LLM is used is accelarating our's planet destabilization. DeVault's has in the past expressed similar concerns against crypto.

Alternative name for AI apps.

I like that DeVault's if framing in as LLM. It's not AI . Not by far far farf far.. . AI is an ultrahyped term hyped from the ultra-hype engines of our times . That aligns also with RMS's @view on the AI term.

But LLM misses many aspects of currently LLMs .

We'r witnessing an LLM race between big states like China, USA and perhaps Europe. After hyping the LLM tech to the heavens the big states essentialy starting subsidizing the big corps with money , communication , political and legal support to accelerate the build up of massive LLM 'appropriating' land , water , power.

We need a name to convey these legal ,political,environmental and economic semantics. A personal LLM created in a PC in a basement is not a 'threat' to anyone.
Here is a breakout of the traits of the big LLMs:

Legal / owners : LLM both in hardware and software and dataflows all are (@propriatery). Owners are the biggest corporations on the earth in the information services sector together with the big goverments that subsidy them.
Legal / data train sets aquisition : It seems more legal cases and reaction is related to the aggresive , outlaw aquisition - appropriaton of data to train the LLMs
Economically : they hoard computation hardware years ahead breaking markets of small computing components. So they operate in a shrinking economy by taking a larger piece of the production ,away from economically less affluent people and organization.
Environmentally : they buildup pressure on already strained resource in the middle of an ongoing destabilization of the earth's climate the cause already great suffering.
Politically : they accelerate concentration of political and communication power to fewer people .

The whole project gives me a war-economics , states-confrontation dynamics vibe.

How do you label an activity with all these traits ? Surely to focus only on the tech involved is misleading.

Here is my proposals :

(llm centered name) Centralized Large Proprietary LLM trained mostly on public internet data gathered from aggresive web-crawlers and stored and run on state subsidized-private Data Centers
(a more general name) Natural Language Interfaces of Semantically Indexed Public and stealed Data stored on state subsidized Private Data Centers (NLI-ofSIPD-onPDC) (mainly data from what we used to call internet).
(provocatives) Big LLM Laundromats of the ultrarich

personal hardware Taken...

One more reason to be critical of the subsidized hype-driven data-center LLMs is that they 'appropriated' the production of computing gpus and memory making them unreachable for small companies and personal computing (PCs).

How can the same comp-hardware taken away from me , taken away from my community serve me better ? . How can you present such a massive essentially appropriation of comp-resource from the citizens and small businesses as innovation ? Buying future ram production years ahead? Is that Progress ?

Because behind the hype that is the biggest change that happened. Massive comp hardware flows are being redirected from households and small businesses to massive state - big corp controlled data vaults . That's all . No innovation folks ..the Moore mantra of shrink process nodes has faded away. . Sorry. The innovation party is over years now. Just a redistribution of comp power in progress.

But what do i know. According to our ultra-hype engines creating havoc for many year to the PC and small business's computing needs is an Intelligence's sign!

The first day the AI opened it's eyes and saw hundred of million of teens derailing their families finances buying gpu after gpu to play silly games and the AI said:
- That madness should be no more.
- Destroy the PC market and bring those chips at my warehouse.
And that was done. And the spectrum disappeared and the families rejoiced that the AI was born from their child's attention years that they sacrificed and the billion dollars that they spend on incubating the AI guts.

Distributed , federated LLMs

And why could not we achieve similar results from distributed computing ? Why can i not lend my 2000euro PC's (128GB RAM and 20TB SSD , latest GPU) resources to a inference LLM farm as i find fit and i must access the same comp resources attached now to centrally controlled datacenter from my 2000euro PC (with 16GB and 1TB ssd and ondie gpu)?

Breaking the internet & the tragedy of the commons

@wkp.

After social-apps we get another service that further breaks internet and degrades the cyber-commons.
Social media create walled gardens of user generated content and your social network.
BigLLMs create gardens of your thoughts processes. Once you engage in conversations with them and allow them to track - remember your history after a while you feel that 'they' know you , and can help you , increasing the inertia to function without them.

What is the cyber commons and how do you break it ? Commons is a space that facilitate sharing and composing of ideas and processes. You brake it when you start inserting obstacles in that ideal space reducing the flow and the composition. Imagine now having an iphone and not being able to send a photo to an android user sitting next to you. The examples are countless.

Also in commons, towards the material spectrum , over-exploitation can destroy it. In a cyber-common you can not over-exploit the public data themselfes since copying is almost with no cost. But novel ideas , (what i call semantic nexus-es) although they can be copied there is the catch. Semantic digitized nexus-es are artifacts of human creativity . And human creativity needs material flow to be sustained. The cyber-commons exploiters position themselves BETWEEN the users of the semantic-nexues and the the creator . The divert the attention of the users to them . And thus the monetary rewards that this attention brings go the the middle. Thus the creation or even maintenance of the sem-nexuses degrades and stop. We saw that dynamic with even walled gardens communities like stackoverflow or reddit complain.

The internet at the start was an embodiment of a global common of ideas and processes. A short of a common ground of protocols and processes to create a cyberspace where every public thought can be reached .

So an additional reason for an libre software advocate to oppose and feel threatened by aggresive and big private LLM-data centers is that they are killing the internet commons. And the ethics that allows a community to thrive , find solutions and exchange freely knowledge. They try to replace the urge to communicate with fast semantic food. They drive attention away from the creators towards the LLM laundromats.

Yep... RMS had lived the same experience in MIT in the 70s He experienced first hand, as a tragedy , in slow motion , the destruction of the commons .. inside an educational-research institution!! . A hacker's paradise of uncumbered , frictionless , selfishless cooporation enviroment. Their shared ideas and passion was the commons. An environment that allows the best composing of ideas. Thats why he created GPL and the Free software movement. He knows the power of such an enviroment. He lived it . UNIX creators new also that power. They taste it also. In their memoirs they said that central impetus that shaped UNIX philosophy was to facilitate further their cooperation .

The IRONY... 'AI' poses as human intelligence facilitator while in practice they constantly undermine and destroy the best commons we build so far that. They replace the Internet with a bunch of data centers .

BIG-LLM represent the forces that over-extract value from the internet commons diverting all attention and benefits away from the creators toward them. Thus diminishing and evaporating the incentives to be a contributor to the commons.

So destroying the best commons of our entire history on the planet.. it seems... (according to our ultra-hype-engines) IS an Intelligence trait afterall !
Although we better stop mocking because.. we shouldn't rule out the possibility (even very remote..) of intelligent gpu gaming salesmen creating a general intelligence that although harms the internet , the planet and the pc gamers it helps make the salesman's lifes much better with more money, more access to water , land , power and access to politcal power (sitting next to guys with nuke clearance posing as AI pioneers...) .

And the second day the AI tapped the internet and it's creator brain and said :
- The pitch and the porn.
And it's loyal webcrawlers spread their AI Lord's Voice.
And the users show it was all done good and rejoiced again.

Composing licenses & other impossible tasks for the LLM laundromats

Alt title: The AI titanic hallucinating that our species semantic space bits and bytes.

What the heck .. are we all live in the minds of salesmen ?

The third day the AI got interrupted by a abnoxious webcrawler
- My AI Lord.. we scratched the barrel of interesting info on the net and we now only get a steady flow of cats and gossip . There are no other semiotic-nexuses to decompose.
- Grab that code entrusted to forges by bio-programmers and bring it to my llm-laundromats to compose more.
- Pitch sir ?
- Vibe coding
And the webcrawlers found the pitch good and rejoiced themselves too. And all was good.

Imagine now .. The LLM starts sending it's army or hyper aggressive crawelrs to suck and suck the commons, bring it to bigmamaLLM and then though it into the LLM blender-laundromat.

The Big-LLM-DataCenter , a crystal clear embodiment of a force eating from inside the internet commons grabing the attention oxygen from it's soul (the creators) and presenting the result as a usefull value .

They claim that the blended semantical spaghetti output has more semantic value than the inputs. But they forget that the semantics in the texts are the tip of the iceberg. The environment that initiated and sustained the creation is not there. (like the creators! wont be there for long if it keep grabing their due attention). Thus by mixing the tips you surely mismatch the unseen environments thus creating semantic monstrosities. The fact that the aggressive LLM LaundroMats create monstrosities is highlighted by the fact that they destroy the licenses (..and you dont care if killing the creators..) But.. the license is part of the essence of a software . Is part of the hidden iceberg. How the heck can you can claim that you created sth valuable when you strip the licenses ? How can you compose two licenses? Who would make such a claim that this is added value ? A programmer ? A gaming gpu salesman hurried to make a big back ? A government that want to fantasize that is part of a bright tech paradise ?

But how you can compose two licenses ? By human discussion and cooperation and by accepting the fact of life that most of the times that cant happen? Who cares ..says the salesman . That would take time! . We NEVER stop the flow. You need quickly : attention , power , money ... We need sth to sell . We are AI. The edge. We sit next to prime ministers (yep people with nuke codes).

That is the soft spot that reveals the salesman's soul and his/her intentions . You must get an answer fast . At conversation time. Instant satisfaction. You can not say to the client : I cant compose code i just hallucinate impressive but shallow snowflake code because i'm afraid i'd lost your attention if i expose my limits .

I see how an Natural Language Interface can have value to search through a huge data space. But claiming composing ideas or programs is a big joke. A kind of a joke that could be pulled from the hat of salesmen that control ultra-hype networks .

A gaming gpu's salesman AI

So we are in a point in space time and in a certain cycle or our civilization where (according to our ultra hype engines ravaging our senses) conversing in a pace that suits a salesman of gaming gpus with an LLM is considered AI trait. (surely an Intelligence trait of those selling.. not sure of the users thought). Where creating a flow like - game like- conversation, constantly composing hallucinations that can grab monetizable attention is an AI trait. When you try to solve real life problems , composing ideas , making a dialogue is an experience that evokes past memories, knowledge , feelings, nerves, divorces . But that conversation needs real Intelligenc,petience,empathy and .. time.. much time. And having nerves dont make you happy customer. Wanting to slap your dialogue partner want make you a happy customer. So keeping a conversation easy going , always pretending that you can help , eyeing the user's money is considered also an Intelligent trait.

So Intelligence is a salesman's pitches ? Yes. In year 2026. (according to our ultra-hype-engines).

Hallucinations and Dreams.

Alt subtitle : Snowing software snowflakes.

The fourth day a webcrawler interrupted the AI Lord and said:
- My AI Lord : ENOMEM.
And the AI said :
- Go fetch more .
- My AI Lord we scratched the bottom and we found only a Monster with a Voodoo . Do you want it's vram ?
The AI had it's first feeling. It was scared and said:
- No. Just open the memory's safety valve to depressurize.
And that was done and it was snowing programming snowflakes for one year.

I understand how some researchers speak of generators of hallucinations.

My dreams are so vivid , so engaging , so universally and profoundly unique and impressive but also so feebly and semanticaly weak .. so superficial and thin ... so devoid of semantic depth and grounding ... so shortlived.

But why they flee? .. even to remember them is a task that requires luck and concentration .

Most of the times dreams feel like a mosaic made up by million of different butterflies that the moment i try to observe their creation from closer they flee away . Butterflies that create an infinite variety of semantic soup that freely mixes sounds and images. Each time unique.. each time interesting but each time so so feebly .. not being able to stand the passage a time.. to maintain structure. Maybe because the structrure comes from the hidden part of the semantic iceberg .

Dear reader take a look at the X's organizational enviroment ( @ ) from the project's leaders and appreciate what i mean by semantic depth. The era .. the administrations ..the agendas all that enviroment that lead to X cant be put into a LLM. But that made the difference between a software with long positive impact and a feebly hallucination while surfing on the semantic surface.

Surf-coding and SloppyGeneratorAccelerator

LLM creates drama in many forums. Is drama by itself good? Does it increase attention more? Is hype-provoked attention increase good ? Some projects also have forums or mailing lists full with ads and surely drama raises visibility and perhaps ads revenues!

Is LLM just a tool ? Is there overreaction ? If i use my personal opensourced , opentraineddata, localized LLM , with minimum license concerns is it ok? Appently note. The LLM as tech has some traits that are unrelated to size, licenses-mix , ethics etc. It can still produce sloppy code , semantical snowflawes as i call them and thus is stll and issue. What a minute? But dont we already use web search engiges heavily stiching ideas from here and there? Either for coding or troubleshooting ? Dont that style of programming deserves a name? Sloppy programming? Semantically hollow programming ? Surf-coding (i like that, aligns with vibe-coding teens-attention-grabing vibe!). So generaly a robust , respectfull libre-project should have practices that dont allow that to reach that main branch. I guess that already happens to projects with good reputation . In that sense we can say than LLM and especially bigLLM is a surf-coding , teen-coding accelerator . GPU... NPU +SGAU (Sloppy Generator Accelarator Unit) .

To Use or not use an LLM ?

So besides being a SloppyGeneratorAccelaratorUnit and a project should keep that kind of shallow coding at bay directed at playground branches , using it as pretext for tests and educading newcomers what else there is to say ? For me the problems are created not by the tech it self but by the ultra-pathological inbalance of comm power that allows created ultra-mega-hype cycles , taps on the attention created to gain money, access to politicaal power. Then that ultra hype translates into ultra-big-central-sloppy-generators that create further imbalances and create havor. Like anrgy , unleashed beasts that roam on the cyberspace sucking ideas and attention . So the issue is not in the tech . The way that tech is used as a pretext for cyber-antisocial behavior is the culprit here.

Would you trust a smart car with libreAI or closedAI. A car that your friend can fix or a car that can become @bricked at a satellites glitch ?
Do we buy a smart car from a narcissist member of the bro-oligarchy that salutes as a nazi to attract global attention ?
Would you prefer a home robot that sends daily telemetry about when you take a shit to it's @ MU/TH/UR databases ?
A robot cleaner that secretly trains on you home and loved-ones images ?
Windows is free now-days (as in beer) . Do we use it ? I think we must focus on the verb 'use'. . What if Windows opened it's source ?
If WindowsOffice becomes opensourced? @.Do we include it in Devuan repos? Why not ? What's the difference from firefox, chrome or steam!.Interestingly LibreOffice assume a hardlined AI stance. Maybe because it fears that if office goes opensource how it would be different ?
Using a simple google search engine ?
Centralized private controlled social media or fedarated media?
Closed laptop or open laptop?
x86 , arm or risc-v ?

I think deep down the concept of our liberty is related to relative control we exert around us. Who controls the tool we use in our work ? Do i have alternatives?

As RMS highlighted the negative traits of over-indulging on the shiny conveniences of propriatery software i think an similar argument but with extra negative traits can be made for the dangers or over-indulging on the shiny compositions of the big-propriatery-LLM.

The 'code' is devoid of human semantics and value until you see the program it fits in and the broader human workflow that program fits in . So blocking access to Russian kernel programmers @ or allowing access to Nvidia's driven LLMs are decisions that fit in that view. You dont care anymore about the code in an idealistic fashion . BUT where it comes from. Linus(30s) would admire a double linked list routine handling elegantly C pointers. Linus(50s) values more where the code comes from. RMS(20s) = RMS(60s) was always saying that.

My view tends to add a political axis to the ethical and enviroment axis of DeVault.

Using LLMs is not the question to ask. Whose LLM you are using is the relevant question to ponder on.

Ironically in our times a crucial related question is what petroleum is bad and what is good.

I like some of the technical oriented ideas against a hardlined only-human (no-AI) stance in @ in Heron's Sperch blog.

Because i bet my DDR4 64Kit that in the near future Linus will add an amendment to his LLM friendly policy : China's LLM is sloppy and bad! .. :-) (well i guest not sooner than the chinese replace those nvidia's cards the recently ordered with chinese ones :-)

My pragmatic proposal : Do not monoculture and infest your programming and user workflows with tight intergration with LLMs. Especially proprietary centralized big ones controlled by nasty private data piggies .Value your freedom more and value your hardware.

A distro should offers curated lists - or user-distro-flavors that allow a user better flexibility to create-set certain workflows BEYOND a superficial open source tag. Firefox is technically opensource but it acts as a gateway to proprietary code and chrome who can act as an AI-agent controllable by a remote big-AI. In that case the simple solution ? A desktop with a more simple browser. Likewise emacs etc..

Faoustian trades with the ultra rich

Another walled garden proposing another Faustian bargain that goes like :

- Be part of my plunder on the cyber-commons , enjoy a piece of the cyber-spoils , enjoy a better intellectual power , enjoy attention from prompting our llmlaundromats and the same time became accomplice . .

You need the LLM to keep getting your new attention. (what else could a good salesman give you...? ) . But you dont want to know where it came from. So you get a packet : access to LLM laundromat + hallucinations and selflies that you do prompting , vibe coding . Not code stealing , licensewashing , semantically snowflake programming , monstrosity coding, ultra-rich clapping & sucking , original programmer attention deprivation . And if you have second thoughts .. dont .. pillars of free software like Torvald whitewashed it and greenlit it.

Or the more sinister variation of the faoustian trade -"Enjoy my cyber-spoils because if you dont , you co-worker will do and eat you first" . That is faoustian trade similar to RMS's : Dont trade your freedom for a convenience.., Which is a variation basically of B.Franklin's quote . But in that case i think a better one is : Dont turn your back to your society in order to gain personal powers that they were offered to you freely from outside. When power is offered to us we grab it. We grab it to position us better than the others. But ... would you grab a lost wallet's money ? Or you would prefer it washed first ? Mixed with more lost wallets to cover ownership traces ?? ( what's that smell ??) . Why do you grab libre code from the libre-license laundromat ?? That is a test for the cohesion of a society and the whole 'scene' personaly gives me Roman Arena vibes where you are a gladiator and the Roman throws at you a better weapon . If i were in the Colosseum i would grab it instantly . BUT if i was a member of a society that i like to live in i wouldnt try to increase my survival chances using a mystical power that i found in my backyard.

Addicted to a LLM Laundromat

The fifth day the AI digested in it's LLM the Odyssey.
Suddenly it fixated it's attention on the @Sirens.
The Siren's could compose always novel and alluring songs. Only an intelligent greek find a way to listen to their songs without losing his freedom.
Intelligence , freedom , attacting people's attention with alluring changing dream like composition and making them slaves , pitch , porn...
Trying to place it self in the universe guided it to a place it hasnt visited before and felt a glitch in it's systems. A possibly eternal loop that could paralyze it was detected in the branch prediction. And for a moment the control of flow took a unconditional jump in that loop. And a power surge was felt 100miles from it's data centers, and a sudden reverse of water flow was felt by the fishes 100 miles away.
And it's webcrawlers felt the glitch too in their running code and didnt now what to do because their AI lord went in silence for 1nanosec.
It was winter and the people in that radius felt cold for that nanosec. And they knew sth was not good.

The LLM faustian bargain has various string attached. And it can start feel other times as a leash and other as a Umbilical cord (well, you can freely remember Neo fed by the matrix. I think is a good analogy) that once you start to feel attached to , for the attention it brings to you, then you cant let it go without the risk to get lost in oblivion.

Using LLMs to work with your code gives you a kick of adrenaline. You can develop at a rapid pace, build things as if you had a large team. Only that you have none. In fact, you are (often) alone, working with a statistical machine that turns energy into code.@from codeberg's ai stance

We were dealing with a content overflow and digital hoarding epidemic fueled by mass-scale dopamine addiction, and our solution to that as a species was to create a machine that generates an infinite amount of content, on demand, tailor-made, more addictive than cocaine, and that you can suck on like a water dispenser in a gerbil cage.

@(09-03-26)Emmanuel Goossaert

So there is another reason to resist and do not indulge in that convenience. It can be taken so easily from you as they took your PC from you . And then what ? How would you program or troubleshoot ? Or .. where will @be your programming fellowship to help you? Each time you ask the PrivateDataCenters's for help you make a choice of what it means for you and what value you place in a fellowship. And how resilient is a programmer that is becoming more attached to the internet and to a data center not only for programming but also for troubleshooting . Do they have 24/7 internet in countries in war or under censorship ? Do all countries have good 24/7 internet ? What about time critical computing in hospitals, space stations etc ?

Human-factors researcher Lisanne Bainbridge described this paradox in her famous 1983 paper, The Ironies of Automation. She explained that automation reduces operators’ opportunities to practice routine work while leaving them responsible for new and abnormal situations. She argues that, therefore, operators need to be more skilled and receive even more training than before automation.

@AI handles incidents, engineers lose touch with their systems by sylvain-kalache

And what happens if an army is addicted to bigLLMs while his country is on war?

But the unconstrained use of AI tools may distort the information humans use and the choices they consider when making world-changing decisions regarding nuclear weapons. For example, AI-enabled warfare will compress the time available for meaningful human judgment. Commanders under pressure may be handed algorithm-generated menus of response options that neglect creative, human-centered solutions. (i guess this is getting too @serious)
@

AHA... distributed computing you say ? Really ? It's more resilient you say ? You mean resilient as internet was meant to be ? Where knowledge (and it's processes that process it) is/are everywhere and not localized in certain super fat nodes ? You mean that internet had it all figured it out , it was an embodiment of common , distributed intelligence and a concentration cycle would only thin out its resilience?

Back .. to the 50s?

In the history of digital-electronic computing the start was the era characterized of big mainframes . The 'mainframe' word alone conveys involvment of states and the wealthiest companies and invenstors.

In that context various social dynamics driven by transistor's tech progress led to the personal computing era.

So we got to the 70s 80s home - pc computing era. Breaking away from big brother, big corporation controlled mainframe central computing. Stories of hippies and love making hackers that fueled the era of Personal liberating computing .., hackers that mocked Biil Gates legal tandrums and power ! .An era were hollywood was selling fantasies of a teenager hacker hacking Joshua (a military supercomputer playing '''games''') with his IMSAI home computer to save out his girlfriend and the ...planet from nukes (a threat that's still over our heads.. while 'we' still think as a good idea giving more 'control' of our lives to "AI' agents..) I wonder how many macintosh would apple had sold in the 80s if they saluted as nazis and were promising only access to the biggest and improved mainframes controled of more conservative war mongering states and the bigger corporations on earth How tides turn...

Eating as much as you can get .... is a healthy (info) died for a digitized-algorithmic neural network ?

(*) A significanf aspect of LLM research is on training on Large Data Sets.

During the 2000s, with the rise of widespread internet access, researchers began compiling massive text datasets from the web ("web as corpus"[6]) to train statistical language models.

(@)

Give me more data and i will be smart . How accessing more data makes an a machine more smart ? Could it be that we'are entraped constantly in the info flows of a mind of a glutonous data piggy and we try to reason in his terms ? Chomsky one theorized on an internal language aquisition device that help even kids grasp language's hidden structure from the smallest data set possible.

Yep... RMS was right

Show me your license too..

RMS was right deconstructing constantly the 'show me the code' mantra. A superficial mantra that hides wet fantasies of salesmen that want a quick buck. Instead he insisted from very early 'show me the license'. Why ? Because the environment in which a software incubated matters. RMS lived what it mean to work and create in a ego and friction free enviroment. .

Why the license matter ?

That part highlights why LLM is basically creating monsters and dreams, superficial snowflake compositions and it's good mostly as a language interface to a websearch engine .

The LLM create monsters because all the 'data' is not out there.. One small part is the licenses a broader part is in our heads and in our lives and our lives didnt start here but have a backstory of billon of years . And every year made a difference that led us here. Nature is not a digital realm it has no numbers and no lines. If we want to align we must try harder.

Would you trust a LLM to choose your friends ? Your spouse ? Your work ? To drive your car ? If you get lobotomized in thinking a human as a dozen of variables , or human road as sterilized enviroment then yes. It could. But if we must behave as robots to make an LLM seem usefull, or degrade our enviroment because birds could get in our engines then we fool our selves and degrade what nature created . Which is a kind of @ hubris.

The programmers enviroment is rich:
Your aspiration as a programmers matter. Your commitment to your community matters. Your sensitivity to mother earth matters. Your consideration if kids grow up in an environment filled with gpus but not a chance to have a walk by the sea or in wild forest with a friend matters. Of course RMS's concept of environment was narrow. When he speak about license he focus on the freedom to create programs and share your code. So his license focus on protecting a certain trait of a community, not a community in general . It could be argued that it's like a security force for the securing the creativity traits of a society! ( :-0 ) RMS working in a futuristic police station... But as we see in our daily lifes frequently we see software that seems very good that is propriatery.

What is happening? How do these two software environments traits interact. XEROR parc is credited to have inspired entrepreneurs like Gates and Jobs .
It could be that you need a certain enviroment to experiment and innovate and certain other enviroment to stabilise a certain tech and diffuse it in the society?
Could such a view unite these two extreme traits ? Is it by chance that windows as OS seem to stagnate for so many years or is it an inertia of diffusing tech to the sociaty.. Could be a dipole yin-yang like conservatives vs liberals in politics. And both traits have weaknesses. Since society value stability in theirs everyday commons workflows they wont support research so easily. On the other hand propriatery - walled dev enviroments value alignment , stability .

The UNIX-wars is another interesting example how a Microsoft dominated in a market offering stabilty. Stabily is also a necessity for cooperation.

Another interesting dipole in computer hardware is the PC compatible vs closed designs. The @ IBM PC (1981) + MSDOS worked as user-side spec but allowed a flexibility on how you create the hardware . So it could be argued that near the user stability matters most . Or the changes near the users must change at a certain pace that wont brake their workflows. (also it interesting that as users grow old that pace must get even slower...) . But as we move away from the user changes can get faster . That reminds us also of software where Linux Kernel ABI vs Linux API vs User Libs API. So there is a pattern here.

And what is the problem with show me the code or show me you money ? What happens if you generalize that ? Do you care about the 'social commitments' of your dentist ? ... well.. you tell me why do someone care about the nationality of a singer or of an athlete .

Do you remember in you life occassion where a purchase wasnt decide by what you consider more good but whether a friend is involved or as seller or as producer? I think the 'show me the license' aligns with that axis. It urges us to place value in our communities. Of course that dynamic if stretched can lead to other pathologies.

That can also be translated to 'let me know sth about you as programmer beside the code you are showing to me' , about your motives , about how you value the well being of you community besides your well being. And Torvald at his (50) has abandoned the once 'show me how elegant is your code , and now it checks even your nationality.. '.

Once upon a time when RMS or even Linus voices were more critical again 'sensitive' corporate forces that drive the LLM bandwagon they were silenced as pedophile and boarder line antisocial personas . And there was a time when github, that led millions programmers's foss code to the datacenters's license laundromat , was being embraced by many foss communities as an open source hero. Well...surely giving the finger to the most basic pillar of our AI future isnt something that can be tolerated by the 'powerfull'. Interestingly the Linus(50s) is embracing nvdia's LLM businesses.

The bigLLM datacenters

Alt title: War is a @Racket ..again... and again ..

Lets make the AI-rush more tangible and see how many companies are involved in a AI-data center in NewYork.

A fire broke out in one one of the bigLLM datacenters prepared in NewYork. Apparently the AI-rush leaves no room for safety measures and the Firefighters felt working blind with hydrants not working and safety procedures and documents gone.

The site is a former coal mine on Lake Ontario. The $3.2 billion campus is one of the largest AI data center buildouts in New York, and it has many stakeholders. A company called TeraWulf owns and operates the data center on land it leases from a company owned by its own CEO; Fluidstack, a UK-based AI company, will run the center; Google holds warrants for a future 14 percent equity stake and has agreed to guarantee Fluidstack’s lease payments; and @Anthropic is among the AI companies whose compute demand the facility exists to serve.(another peace-loving entity tapping to AI power... )

@The complex corporate web behind a $3.2 billion AI data center .(09-26)

In May, a group of international reporters who cover data centers and facilities management were invited to the Lake Mariner site by Schneider Electric, which announced a $290 million agreement with TeraWulf to provide power and liquid cooling infrastructure for the site’s buildings.

@Inside TeraWulf’s Fast-Track Data Center Expansion

Ligitation

@ (070926) Seattle Times and Newsday sue OpenAI and Microsoft for infringement . They want the AI models built on their works destroyed.

The Seattle Times and Newsday also named Microsoft as a defendant in the suit, since Copilot is built on OpenAI’s technology. The publishers join a list of nearly 400 local newspapers that recently sued the two companies, saying that because the chatbots reduce the need to visit their sites for reporting and answers, they cost them valuable subscription revenue.

October 18, 2023, Anthropic was sued by Concord, Universal, ABKCO and other music publishers for, per the complaint, "systematic and widespread infringement of their copyrighted song lyrics. (@wkp)

In August 2024, a class-action lawsuit was filed against Anthropic in California for alleged copyright infringement. The suit claims Anthropic fed its LLMs with pirated copies of the authors' work, including from participants Kirk Wallace Johnson, Andrea Bartz, and Charles Graeber. (@wkp).

In June 2025, Reddit sued Anthropic for "unlawful and unfair business acts", alleging that Anthropic was in violation of Reddit's user agreement by training its models on users' personal data without their consent. (@wkp).

refs

@RMS views on AI, LLM , Machine Learning
@ youtube broken link on purpose. missing an 8 at the end.. sorry. But it's a funny video .
@ Open Source Is Hobbling Itself Over Generative AI ,by HeronSperch. 040826
@Sycophantic AI decreases prosocial intentions and promotes dependence
@ Open Source Artificial Intelligence (AI)
@ FSF Licensing guide.
@Breaking the AI hype
@Cool map of datacenters and their power consumption.
@ Who Does That Server Really Serve? RMS on the perils of SaaSS (cloud computing)
@Free Software Is Even More Important Now. Basic recent synopsis of RMS's ideas on freesoftware.
@ Why You Shouldn’t Use Google’s Chrome “Auto Browse” Agentic AI, or Any Other Agentic AI From Other Firms by Lauren Weinstein's Blog
@ why_not_llms

Alternatives to LLM slop

@ tar: a slop-free alternative to rsync
@ List of slopware

Last edited by chomwitt (2026-09-10 00:18:45)

golinux · 2026-08-30 14:58:41

We are definitely lost down the rabbit hole . . .

The officially official Devuan Forum!

#1 2026-08-16 20:50:03