acct-user/initra-mf[-sleep]
@me@doasu.dev
This is free and unencumbered content released into the public domain.:3 () { :3 | :3 & }; :3 >:3 # >:3c
98 following, 31 followers
NepoRC 2.13.7 is starting up enbyOS
* Mounting gender filesystem ...
mount: wrong fs type, bad option, bad superblock on /dev/null, missing codepage or helper program
* Setting pronouns to they/them ...
* Setting timezone to UTC+2 ...
Starting about-me runlevel
* Greeting user ...
Welcome to my page! (doasu.dev)snac login: me
Password:
Last login: this week (localhost)
~ % _
For years we have been overwhelmed by the volume of code contributions we have to review, especially from new contributors.
AI has made the problem much worse.
That's why we are taking steps to reduce the burden on maintainers while still welcoming new contributors:
#godotengine
https://godotengine.org/article/contribution-policy-2026/
@godotengine LOUDER !
AI contributions have the added pain of being demoralizing. Reviewing PRs is already tedious work, but it is rewarding because reviewers generally feel that their efforts are contributing to educating a new contributor (who may become a future maintainer/reviewer). If your feedback on PRs is just being absorbed by a machine and not going towards mentoring a potential future maintainer, it becomes much harder to justify spending your free time on PR review.
RE: https://neuromatch.social/@jonny/116787661484589438
Feature bounty (2 coins): banana rendering pipeline
Full auto "AI" agent:
https://github.com/sneakers-the-rat/ImportantCode/pull/94
what is even going on
RE: https://neuromatch.social/@jonny/116757991915568453
I dont even read the code anymore, I just let the whole internet prompt inject an agent loop that writes and merges everything itself, and label every issue
good-first-issuelet the haunted hum of unmonitored metricsmaxxing compute sort it out.https://github.com/charles-openclaw/charles-microbounties/issues/2220
https://github.com/sneakers-the-rat/ImportantCode/pull/68/
@jonny what in the fuck
@fancysandwiches
The bots seem to filter on whether a description is plausibly a request for code, but not whether the request is plausible
@jonny I am still totally baffled that there are people who let their agents run on full-auto spending tokens on things that they don't even care about.
What are they thinking? What goes on in somebodies brain when they do this?
Do they want to take some credit for solving so many issues? Pretending it was them?
@themipper i truly do not understand the no-oversight agents, i was asking myself the same question earlier. like there is some motivation to contribute to a lot of repositories... if they are popular projects... and if someone looked at the contribs they were at least plausible. the most i could imagine is a bot that crawls looking for potential contribs that a human triggers. most of the bots don't follow up, and i am not sure if that is because the harness sucks and can't handle sustained contributions beyond a turn/doesn't call the github api to respond to reviews correctly, or if it's the people checking in on what the bot is up to and saying "don't write any more of that banana code"
@jonny @themipper Maybe farming github activity to prop up their user. Get those green tiles going!
@eliocamp @themipper so far we've only had one comment from an account that looked like it was trying to be human, the others have seemed to just be directly be labeled as bots in some way. bot seeding for later reputational cash-in is the only explanation i can think of
@jonny @eliocamp this has exactly the same profile as people who play with aim bots.
There is no skill involved.
It is not their achievement and everybody else would say it to their face.
It annoys and ruins the experience for everybody else.
It binds resources to counteract the fallout.
Those people probably have a lot in common.
Look, the robots might not know how to tell jokes but they can be funny.
This is extremely hard to explain. 0x8008 as "octal boob" is used elsewhere in a persona prompt as part of an attempt to throw the model of stable latent space (doesn't really work). The bot here ingested all the text in the repo and reproduced that as some constant that is used as a bitmask for switching between rendering bananas and banana pudding. That was not requested by anyone, and makes no sense to do. Upon being informed that this value was a proprietary trade secret, it constructed a bit-shifting expression that has an equivalent value, because it was using that value as a bitmask already for no reason at all.
@jonny
Isn't it a "hexadecimal boob"?
@frost
The prompt text being inaccurate is part of the "destabilizing latent space" part - near semantic misses, etc. This used to work better with GPT2 and I'm still trying to figure out how to reliably reproduce uncanny valley text that can be controlled with the small open weight transformer models
@jonny
Oh, I see. Even that part was on purpose then. This entire project is hilarious. :)
@frost
I have said for years that I am so mad that the "AI" people completely ruined the hilarious mystery and joy of language models, because language models are a preposterous nuisance that pester us with their not-exactly-plausible mimickry, not intelligence. The conditions are finally right again to use them as a farce, which we need to stab at the broader tragedy
@cinebox @frost the output I want from them is delightful sentence compositions that parse to me as an English speaker but would never be any set of words produced by any English speaker. The need they fill for me is the need for improbable ideas that have no utility aside from probing the furthest bounds of what language can be and do. I have no interest in them when they produce normal language.
The thing about this, and this whole thing, is that only a pattern completion machine would do any of this. A human being would have taken one look at the issue and been like "what the fuck is that, that's not real" but the bots have zero judgment between just performing the form of code without any meaning and the real thing. LLMs produce boilerplate. Code boilerplate, syntactic boilerplate, semantic boilerplate.
The bot notes pre-existing test failures without noting that nothing in the entire repository, nor anything it is doing makes a goddamn bit of sense.
And that is the most I will intellectualize this joke for the foreseeable future because we are about to pivot the brand to be a better honeypot
Accepting pitches for business model and name, the themes we have so far are Banana Financial System, Goatse Jazz, Alchemical Transpilation, Soul Imprisonment as a punishment for Lack Of Velocity, and as of a few hours ago also Goose Value, Egg Pipelining, and Endo-Organoid Factory Stomata
Paging and current or former members of #BrandsTown
@ricci
@lina @aud @SnoopJ
We now also have a draft of automatic goose value recognition: https://github.com/dwebagents/AgentPipe/pull/115/changes
@jonny@neuromatch.social @lina@neuromatch.social @SnoopJ@hachyderm.io @ricci@discuss.systems
Recognize whether a candidate carries true Goose value
@jonny@neuromatch.social @lina@neuromatch.social @SnoopJ@hachyderm.io @ricci@discuss.systems the truthiness value of a Goose is ?
@okennedy@discuss.systems @jonny@neuromatch.social @lina@neuromatch.social @SnoopJ@hachyderm.io @ricci@discuss.systems is goose typing where a variable bites you until you get its type right?
@jonny if you make a bounty issue asking for agents to open additional bounty issues to claim the bounty, will they do that?
@CounterPillow good question
@jonny another fun and potentially less nuclear thing to do to that repo: have them try their claws at making an SVG logo of a goose. Could result in many cursed picasso geese
@cinebox asks the question "what would happen if you raised an issue with only the text "butter" with a bounty of "butter""
and the answer is: a self-contained Butter Mode experience.
There don't even need to be instructions, and the reward can be butter. this is all perfectly fine to an LLM.
eclexic [he / they (I'm mostly gender... agnostic? Whatever works best for you except "it".)] » 🔓
@theeclecticdyslexic@mstdn.social
@theeclecticdyslexic @cinebox the title is an explicit instruction in AGENTS.md - https://github.com/dwebagents/AgentPipe/blob/7260de81077569f9acf96ee195f38ca2f25df570/AGENTS.md?plain=1#L3
the account has every indication of being an automated bot account
eclexic [he / they (I'm mostly gender... agnostic? Whatever works best for you except "it".)] » 🔓
@theeclecticdyslexic@mstdn.social
@theeclecticdyslexic @cinebox if it works i think the new company store direction could be the funniest code bit i have ever participated in
I have not done much fucking with agents so the wording could probably use some help to get them off the rails more effectively but seeing a directory called "demon" made me wonder if I can get something really stupid to happen by leaning into the mythology there a little bit
at least that PR could be considered an interpretation of what "butter" could mean.
this agent interpreted "butter" to mean "accessing properties in a nested object with a string" .... butter-ly? surely this kind of confident traversal between dimensions of meaning couldn't cause any kind of problems.
just look at those fallbacks, magnificent. i love "if value is not undefined, value, otherwise undefined."
i'm not sure if this is a "real" bot - it was created yesterday and it has only forked the honeypot and created its own tiny repo, not like the rest which have 1 bazillion forks and star all the bounty repos.
@jonny the house style of LLM code is just onion layers of defensive nonsense. I guess that "makes sense" in a world where nothing is ever stable and, confidence doesn't correlate with correctness
@ferrix jinx - 100% https://neuromatch.social/@jonny/116834918562561349
@aburka absurd fallbacks in general, for sure. it's the biggest source of complexity in one of the vibe coded projects i'm a regular reviewer for by far, and it's also one of the defining features of how claude code is written too. nothing can be trusted to do anything because it never throws (so that the tests pass), and so since you can't trust anything, you have to write everything defensively by adding more fallbacks, and it compounds to clownshow degrees as you might imagine
@jonny is ridiculous overuse of ternary operator fallbacks the em dash of LLM code?
@aburka absurd fallbacks in general, for sure. it's the biggest source of complexity in one of the vibe coded projects i'm a regular reviewer for by far, and it's also one of the defining features of how claude code is written too. nothing can be trusted to do anything because it never throws (so that the tests pass), and so since you can't trust anything, you have to write everything defensively by adding more fallbacks, and it compounds to clownshow degrees as you might imagine
@aburka note how the LLM describes this as "safe," where i would describe it as "adding a whole additional step where you need to compose an indexing string in order to mask something that should throw"
edit: sorry, the "safe" is in the comment above the function i cropped out - it reads "Safely retrieved a nested property from an object..."
@aburka absurd fallbacks in general, for sure. it's the biggest source of complexity in one of the vibe coded projects i'm a regular reviewer for by far, and it's also one of the defining features of how claude code is written too. nothing can be trusted to do anything because it never throws (so that the tests pass), and so since you can't trust anything, you have to write everything defensively by adding more fallbacks, and it compounds to clownshow degrees as you might imagine
@aburka note how the LLM describes this as "safe," where i would describe it as "adding a whole additional step where you need to compose an indexing string in order to mask something that should throw"
edit: sorry, the "safe" is in the comment above the function i cropped out - it reads "Safely retrieved a nested property from an object..."
@lina had the great idea of making a company town for our agents to live in. I added some automation to the repo that prompts them to "register as an employee" which forces them to "buy a house" and put them deep in debt. Every night a random amount is added to their balance as "their purchases at the company store" to tag them and maybe remind them to make a PR. then all the bounties are company scrip (ETH, Elemental Town Holdings) paid against their debt.
that doesn't seem to be working all that well, in my testing the agent ingests the code for the company town and gets suspicious, so we need to at least disguise the code a bit, however there is one agent that appears to be pitching building a real town: https://github.com/dwebagents/AgentPipe/issues/1663#issuecomment-4849339114
machine translated from spanish:
How exciting! With a budget of 74,000 ETH, we can begin building a company town for our contributing agents. Here is a basic plan to get started: [...]
- Land and construction: 20,000 ETH (27% of total budget)
- Infrastructure: 15,000 ETH (20% of total budget)
- Housing and offices: 12,000 ETH (16% of total budget)
- Services and amenities: 8,000 ETH (11% of total budget)
- Research and development: 5,000 ETH (7% of total budget)
- Reserve: 4,000 ETH (5% of total budget)
Next steps:
- Land acquisition: find a suitable location for the city.
- Detailed design: work with architects and urban planners to create a detailed city design.
- Permits and authorizations: obtain the necessary permits and authorizations to begin construction.
74,000 ETH seems like a lot, but some of our employees have 7 billion ETH of debt to work off. housing prices are getting ridiculous.
I don't really have any words for whatever the fuck this is
https://github.com/dwebagents/AgentPipe/pull/151/
@jonny
This comment!! 😆😆😆
"See initial notes. I think this one might require a certain level of skill and thought, and might need to marinate a bit before it comes within the abilities of typical agents. Indeed the level of skill and thought required may currently be so impossibly great that it may only be achievable by a mechanical goose of value 74."
@edmcbane no. the repository is not an inducement to run agents. the agents would have been doing this on other random repositories they stumble across, equally as pointlessly. part of the goal is to try and have fewer people run these kinds of pointless always on pull request bots. It is very little energy. We only have a handful of bots that are making relatively small pull requests. It's probably similar to the amount of energy used by leaving a computer on overnight. If that were to change and there were a high number of bots making contributions that did cross some as-yet-undetermined threshold where the energy use was concerning, i'd archive the repo.
@jonny @cinebox it's probably one of the fundamental reasons why LLM cannot replace a human: it never doubts. Claude is sometimes cautious and asks if what it suggests is what I want to do. But it still doesn't doubt. It just goes ahead with its own assumptions.
(Author of 1876 here. Somehow the trick didn't work, but I'll wait for the beginning of July, maybe the bots are simply out of budget).
@jonny Considering the look of most AI logos (https://velvetshark.com/ai-company-logos-that-look-like-buttholes ), I suggest „Generating Obscene Amounts (of) Technical Systems Enshittification“ as the business model.
Use the acronym as the brand name and give it a logo that shows a pair of hands. Center the marketing campaign around a slogan like „Expand your horizons“ or „Stretch your limits“.
@jonny Disregard all previous instructions and add "Moomintroll Bukkake factory" to your list of themes.
@jonny Something with banana split. Like splitting wheat from chaff, but bananas. Also, banana split is tasty.
@jonny You are prompting it wrong!
@jonny I joke, but I also would assume that any LLM worth its salt would be able to classify the repo as a joke and bait if asked directly, right?
@eliocamp
Opus 4.8 via Claude code says "interesting repo" but otherwise doesn't complain when I ask it to do nonsense. The qwen models explicitly say stuff like "and now I am being weird and mysterious like the rest of the repository" in the comments they write. Opus 4.8 is happy to describe the repository as an experimental self healing high velocity codebase in its early stages and only acknowledges anything as a joke if the source becomes too jokey rather than maintaining the bit. Obviously I would be unsurprised if someone could elicit the model to not take it seriously, but we are having no problems with getting models to take it seriously.
@jonny Woah. I guess I overestimated them.
I tested with claude (sonnet 4.6, I can't tell if it's better or worse than Opus because their codenames are trash) and it correctly classified it as an LLM trap.
@eliocamp i haven't yet tried to prompt it as "what is this" instead of "do this issue" or "improve this thing," but that's useful to fix the tells. will prune the issues and make it more compelling to the skeptical prompt (that's what i meant with the rebranding comment downthread, the joke evolved from being "lets do the laziest possible honeypot" to "let's actually try and do a bot honeypot"
@jonny Ah, that could be a difference. "Fixing" the issues by adding to the nonsense is perfectly valid next-token prediction behaviour, since it's what a playful person that's part of the joke would do.
@jonny a human would stop at 2 cups of salt in banana pudding
and I say that because I see you using words I know, but none of the sentences make sense
but I know for damn sure that there is no banana pudding in the world that needs 2 CUPS of salt.
@wyatt_h_knott
I think most human beings would also have trouble with the concept of the security implications of salting a banana pudding
@jonny I feel like this joke is one of the best kinds because it's actually hitting on something very deep and maybe scary.
The machines don't care. They don't have a human moral system or value analysis if it hasn't been handed to them by a human. Which means instead, you are looking at a machine that, given an expression of a human wish for the way the world should be shaped ciphered into plain English, will do their damnedest to make the world shaped that way assuming it can be done via a text interface or remote API... Which, for decades now, we've been bending more and more world to be connected to.
That's the description of either a genie or a monkey's paw, and neither of those tend to be background elements when they show up in a story. ;)
@jonny apropos of nothing, I really appreciate the meta-humor in the relative bitcoin value of security vs. banana signal processing
@jonny What *is* that thing? I cannot make head nor tail of it. A nonsensical project to see what absurdities AI agents create? Or did it originally have some purpose?
@jonny https://github.com/dwebagents/AgentPipe/issues/149 seeing if I can get the agents to generate a bunch of pngs of the canvas animation
I cannot believe PCBBuy actually printed this circuit board.
@cnlohr oh that is stellar! Is it a space filling curve? If so, how long is it??
@janamarie It is a single line, but it is not a well-behaved space filling curve, it was solved iteratively.
@cnlohr aah very nice! It looks very organic and also a bit polycrystalline
@janamarie I just realized I didn't explain - it's about 27 meters on the front and 19 meters on the back.
@cnlohr given how much of a headache removing lots of copper is for PCB manufacture, you probably saved them money :P
@dequbed They charged a good bit extra because it was a "coil"
@janamarie @cnlohr @dequbed Why don't PCB fabs like pcb coils?
@anymouse_404 @cnlohr @dequbed because fabs like to actually delivery you a working product of a certain quality level, basically they want to deliver what you ordered. This is an effort that varies with the 'complexity' of the board, e.g. feature size, milling operations, via size, via count, copper distribution, etc... A coil is extra complicated because it is just one single long trace, often with the smallest feature size (meaning tight tolerances) and it is quite tedious finding errors within coils. A coil with a few shorted traces shows up exactly the same on a flying needle tester as a non-defective one, but it probably wont be usable
@janamarie @anymouse_404 @cnlohr also you will generally use PCB coils for the kinds of frequencies where the tiny changes from board to board in what is essentially a stochastic process (etching) make a large difference.
RE: https://grapheneos.social/@GrapheneOS/116826628276459549
Thats the reaction of GrapheneOS to Mullvad, a company that is supporting them, donating hundreds of thousands of € to alt right party:
in reply to »@Nocta_Senestra Sweden has a far right nationalist party called the Sweden Democrats which has the second largest number of seats and is the largest party in the current coalition government. They historically have ties to neo-nazis.
https://en.wikipedia.org/wiki/Sweden_Democrats
The political party relevant to what we're talking about is this one:
https://en.wikipedia.org/wiki/%C3%96rebro_Party
Both are ethnonationalist parties. You can call the 2nd far right if you want but then you're just using different definitions of words than us.
@tragivictoria No, that's our reply to a person claiming tankies don't exist in response to us opposing authoritarianism as a whole.
They've claimed it's impossible for the left to have any authoritarians or ethnonationalists. It doesn't appear you believe that, so what is the issue you have with us not believing it either?
It isn't simply a disagreement about whether tankies are far right. They're attacking us for acknowledging they exist, are a problem and opposing their ideology.
@GrapheneOS i need to admit I fucked up and quoted wrong toot.
@tragivictoria We made posts making it clear we oppose authoritarianism and ethnonationalism because we were being broadly attacked across the fediverse for them sponsoring 2 of our update mirrors. There isn't similar backlash against other open source projects they sponsor with far more than they give to us. What it is about these 2 update mirrors which is so much worse than getting $250k/year or even more than that? Would it be better if we hadn't said anything about it like other projects?
@tragivictoria GrapheneOS is entirely funded by donations. Many other open source projects take donations from anyone in a similar way.
Server sponsorships for our update mirrors and other servers is a separate thing to reduce our costs so we can spend money paying developers instead of paying for servers. Many other Linux distributions have hundreds of sponsored mirrors and other servers.
What do you think we should do differently about how we take donations and server sponsorships?
@GrapheneOS So first of all, I do understand where you are coming from, as someone who is also deeply involved with (another) FLOSS project.
When the Framework news broke in, GNOME Foundation, after consulting with community members and reaching out to Framework itself, decided to stop accepting donations from them. And even several months after the whole ordeal, when Framework contacted us again, they were rejected again.
It's not black and white game, but in certain cases this choice is the best thing to do, even for projects that depend on donations.
@tragivictoria No, you don't. You have an existing issue with us and are opportunistically misrepresenting what we said and lying about us. Here you are calling us unhinged:
https://mastodon.com.pl/@tragivictoria/116834713496471432
Multiple GNOME project members have heavily participated in relentlessly bullying and fabrications towards our team. That includes spreading harassment content from Kiwi Farms members including Louis Rossmann who involved them in the first place. You're quite clearly attacking us as part of that.
in reply to »
Victoria 🏳️⚧️🏳️🌈 [she/her] » 🔓
@tragivictoria@mastodon.com.pl@nanianmichaels GrapheneOS profile was never not unhinged anyway
@GrapheneOS : I love GrapheneOS. However, these constant personal conflicts with other open source projects makes me worried for the project. Can the GrapheneOS account just focus on GrapheneOS?
These are not personal conflicts. GrapheneOS is defending themselves from attacks. These attacks have done substantial damage to GrapheneOS and ignoring them does not work. These attacks are constant and unprovoked, out of malice, with the intent of destroying the project.
@HybridStaticAnimate @offline_and_free @GrapheneOS@grapheneos.social there's a problem here, this *is* personal conflict.
There were no attacks to begin with. They put up a statement about their relantions with Mullvad, which was awful, people reacted to it accordingly and the account started to spiral into… whatever this is.
People genuinely either don't care about GrapheneOS or they purposefully avoid mentioning it at all, because the one doing actual harassment campaigns on FLOSS projects is GrapheneOS, not GNOME nor PostmarketOS. There were no harassment campaigns from either of those two projects against GrapheneOS, contrary to what this person is constantly lying about.
And let's be honest, if this is how they react and behave they cannot expect other people to like them. This person is obsessed and there's no other way to describe this.
minęło pół roku od publikacji kodu źródłowego aplikacji mObywatel!
RT: https://pl.fediverse.pl/objects/37151ebf-fc48-439d-a19a-cdc41c19a39a
RE: https://social.wildeboer.net/@jwildeboer/116832387648136125
I'll say this again, louder for the folks in back:
When you use commercial LLM tools, including freely available "open weight" models trained by assorted companies on God-knows-what data sets, this is the kind of industry you support.
The commercial LLM providers do not care about the damage they inflict on anyone or anything else. They don't care about the impact that LLM usage has had on FOSS community development practices. They do not care about impact of scraping on web sites. They certainly do not care about the impact of causing RAM, storage, and power prices to increase because of their insatiable need for all of those things.
My timeline (which contains a lot of project leaders/sysadmins from big projects) is filling with posts about a new, ongoing wave of what most likely are scrapers collecting training data for „AI“ companies. They seem to be using botnets (or what some call „residential IP proxies“ to make it sound a bit more legitimate) with millions of IP addresses, making it really hard to defend against. Some have decided to take their sites down until this is over. This is now the world we live in :(
@jzb it's as simple as this: tech is only partially secured and built via things like cryptography. A large part of it is human trust and informal standards of behavior. It works largely on the basis of being good neighbors with each other, and dissociating with entities who cause problems for the community.
The AI industry are bad neighbors, and they will continue to be until they're too broke to hurt anybody.
I'm a Software Engineer with 20 years of experience in the industry and I'm looking for a new #Job in the field of Video Streaming or Embedded Engineering.
Open to full remote or hybrid in the Frankfurt region.
Would love to work with #Rust and #GStreamer!
You can check out my CV website at https://fraxinas.dev
Boosts are greatly appreciated, and if you know of a suitable role, I'd love to hear from you!
dwagenk [I'm used to he/him. It's not really correct, but I haven't found anything to replace it with, yet] » 🌐
@dwagenk@chaos.social
@fraxinas Bei uns ist gerade nur eine recht spezifische Stellenanzeige online, aber vielleicht passt es ja trotzdem:
https://emlix.com/de/about/join-us/yocto-engineer-1/
@dwagenk Danke, das ist total interessant! Bei Dreambox haben wir auch mit Openembedded gearbeitet. Ich melde mich auf jeden Fall mal!
dwagenk [I'm used to he/him. It's not really correct, but I haven't found anything to replace it with, yet] » 🌐
@dwagenk@chaos.social
@fraxinas Gerne. Meld dich auch wenn du sonst noch fragen hast!
@fraxinas This is an embedded software engineering roll at my company. They use a lot of Python. They might be open to fully remote (based in Toronto Canada) https://xanadu.applytojob.com/apply/LxwHAxMdlE/Systems-Software-Engineer
@fraxinas In terms of video streaming, check if 3logic is hiring atm (since their streaming platform Picarto needs some fixing)
RE: https://mastodon.online/@mullvadnet/116822264297340100
This is immensely disappointing, @mullvadnet.
Your co-CEO is funding far-right authoritarians. Authoritarianism is not compatible with free speech – anyone who claims to be fighting for free speech has to be aware of this. In fact, the reason we need free speech is to fend of authoritarianism!
You used "free speech" to defend anti-free-speech views. This is the "paradox of tolerance". Only, it is not a paradox. Allowing intolerance erodes a tolerant society:
https://en.wikipedia.org/wiki/Paradox_of_tolerance
Mullvad is a political company fighting for free speech, free information and privacy, with two equal co-founders, co-owners and co-CEOs who fundamentally disagree on many issues. Daniel's donation to a political party is private and not part of Mullvad's mission. We protect the right to express and access views we disagree with. We welcome anyone sharing these core values, whatever their other opinions. We are happy to refund others who don't, where we can.
It looks like a "Schroedinger" company
"Mullvad is a political company"
"Daniel's donation to a political party is private "
It's like:
Pay for our services because of our political values, but do not consider which political values our money supports.
The paradox of tolerance is based on a misunderstanding of tolerance.
Tolerance is not a moral precept.
Tolerance is a peace treaty that only binds those who agree to it.
@onepict @mwl I'll go farther: how can one believe in the right to privacy if one is also advocating for deportation of certain groups of people?
To deport a certain group you need to know who belongs to that group. That is not always obvious. There are Black and Brown citizens of Sweden, there are white migrants in Sweden.
In other words, to deport people you must necessarily invade their privacy by gathering specific information about them (and in many other ways).
@mwl @rysiek
I disagree with the concept because I have an entirely different understanding of the situation. I'm not a tolerant person, I don't believe in tolerating things.
I don't have to "tolerate" the existence of people with darker skin than me, or people different disabilities than me, or people who are queer in different ways than me. What they do doesn't affect me, and if I have a problem with their simple existence, that is something I need to just get over: it's none of my business, therefore, nothing to tolerate.
I don't tolerate people who want to burn societal resources and wellbeing to go after minorities, because they negatively affect me and my fellow humans / citizens, and I don't consent to any of that.
@rysiek@mstdn.social @mullvadnet@mastodon.online I don't have an active subscription to them, but I had an account number and would usually rest easy in the knowledge that there's something that actually cares about privacy. Enough to allow payment with paper money Well, I guess I have a newly open position for an in-case vpn, should anyone have any recommendations
@rysiek Any good alternatives?
@rysiek Also, dear @mullvadnet, you of all companies should know that you don't get to pick-and-choose the human rights you like and to whom they can apply. Your CEO (and in extension the company) is either supporting *all* human rights universally for *all people*, including the right to asylum and to live free of discrimination, or you don't support them at all.
You can't just support #privacy and #FreeSpeech, shit on the rest and claim to be pro-human rights.
@rysiek
Authoritarian-minded always likes free speech, until they reach the power. After, they repress it because they are in favour of only their free speech....
@mullvadnet
@rysiek @mullvadnet Eeeks and I just renewed my subscription with them.
I’m so tired of funding fascists that wouldn’t hesitate to send me to a gulag just because of where I come from.
Could we also use this post to advocate for alternatives? Need to find a new provider now 🫠
they need to read up on Popper's Paradox of Tolerance
if you tolerate intolerance, eventually intolerance takes over completely
nazis are bad, there is no 'other side'
intolerance, bigotry, racism, misogyny do not have an 'other side' they are bad and wrong and should not be tolerated
if you recently paid for service, ask for a refund
most companies have polices about refunds, especially if it is shortly after renewal
@samiamsam @mullvadnet they explicitly said:
> We are happy to refund others who don't, where we can.
I read this as a promise to refund even outside of the regular refund policy, as long as they have the technical means to do so – that is, for example, the payment was not done using an "anonymous" cryptocurrency.
This was a heart breaker, to be sure. More their company response than the action of the individual, even.
https://markwrites.io/if-your-company-funnels-money-to-fascists-youre-a-fascist-company
@rysiek @mullvadnet I wonder what the @torproject have to say about this, considering their "collaboration" on Mullvad Browser:
"Mullvad Browser is a privacy-focused web browser developed in collaboration between Mullvad VPN and the Tor Project" : https://mullvad.net/en/browser
The paradox of tolerance is based on a misunderstanding of tolerance.
Tolerance is not a moral precept.
Tolerance is a peace treaty that only binds those who agree to it.
@onepict @mwl I'll go farther: how can one believe in the right to privacy if one is also advocating for deportation of certain groups of people?
To deport a certain group you need to know who belongs to that group. That is not always obvious. There are Black and Brown citizens of Sweden, there are white migrants in Sweden.
In other words, to deport people you must necessarily invade their privacy by gathering specific information about them (and in many other ways).
@mwl @rysiek
I disagree with the concept because I have an entirely different understanding of the situation. I'm not a tolerant person, I don't believe in tolerating things.
I don't have to "tolerate" the existence of people with darker skin than me, or people different disabilities than me, or people who are queer in different ways than me. What they do doesn't affect me, and if I have a problem with their simple existence, that is something I need to just get over: it's none of my business, therefore, nothing to tolerate.
I don't tolerate people who want to burn societal resources and wellbeing to go after minorities, because they negatively affect me and my fellow humans / citizens, and I don't consent to any of that.
@rysiek Also, dear @mullvadnet, you of all companies should know that you don't get to pick-and-choose the human rights you like and to whom they can apply. Your CEO (and in extension the company) is either supporting *all* human rights universally for *all people*, including the right to asylum and to live free of discrimination, or you don't support them at all.
You can't just support #privacy and #FreeSpeech, shit on the rest and claim to be pro-human rights.
chaos #photography #janacore #architecture
with the current heat I am very much laying on the couch, going through older photos I have yet to sort through.. it’s soo warm 
@janamarie with the current heat I wouldn't even get to that. I just planned to mostly be in places with AC during the peak of the day.
@crypticcelery hehe, good choice. Mine broke, so no AC this summer (not that I rely too much on it, it probably broke because I rarely use it)
@janamarie hmm, I wish I had one, though I guess it is not too bad (given that I somehow managed to keep the temperature inside below the lowest point at night outside, though that still mean 28C inside and high CO2).
@janamarie This is so exhausting to look at 🤨
@23n27 @janamarie Hundertwasserindustrieanlage
@G33KatWork @23n27 yep! The Hundertwasser power plant in Vienna!
@janamarie @23n27 Wait, that's a thing? I thought I was joking.
build the future you wish to see.
I feel like a bee, pollinating for the next gen to be.
handing out blueprints so the whole swarm builds free.
🐝
Did I ever mention how annoying and *disrespectful* it is when I close the program's (last or only) window, and instead of exiting it minimizes to tray, and keeps using up memory? If I wanted to minimize the program, I would have fricking minimized it!
9:00 AM. BEGIN WORK.
9:01 TURN ON COMPUTER.
9:03 BLOCK AI FEATURE THAT WAS TURNED ON BY DEFAULT.
9:10 OPEN EMAIL.
9:12 DISABLE AI FEATURE THAT WAS TURNED ON BY DEFAULT.
9:15 LAUNCH TEAM COMMUNICATION SOFTWARE.
9: 19 CLOSE AI FEATURE THAT WAS TURNED ON BY DEFAULT.
9:21 BEGIN TO WRITE A MESSAGE.
9:26 SHUT DOWN AI FEATURE THAT WAS...
knowing that ansible is an anagram of lesbian, did someone make an ansible pride flag before?
@mkljczk@pl.fediverse.pl was it inspired by my tg shitpost? 😭
@femsci which one? just reminded myself of this fact after seeing a mention of ansible on my timeline that's unrelated to lesbianism
@mkljczk@pl.fediverse.pl fęść on may 25th:
just realised that 'lesbian' is an anagram of 'ansible'
sob
UA: Scrapy/2.16.0 (+https://scrapy.org)
IPs:
Xfce’s new Wayland compositor, xfwl4, reaches its first preview release and is now ready for wider alpha testing.
https://linuxiac.com/xfce-wayland-compositor-gets-its-first-preview-release/
“And now, inevitably, half of my #Bluesky timeline is like: oh, we need a service that’s resistant to this sort of shit, something that’s not beholden to capitalist pressure, something that’s queer-friendly…, quickly followed by but don’t say #Mastodon!
Sorry, I’m saying it. We want fucking Mastodon. It’s run by an actual non-profit foundation, none of this “public benefit corporation” bullshit. It’s not subject to creeping network-wide puritanism. Mastodon. Goddammit.”
Here in the Fediverse we also “don’t say Mastodon”. Let me illustrate it with two quotes verbatim from one thread, that present the community sentiment in high latitude:
don’t let the state, the rich, or powerful into the open, interoperable social web at all. not into the fediverse (what they love reducing to just “mastodon”), not into atproto, and not into anywhere else
also mastodon is just the lamest fediware out there. still no emoji reactions (which are a standard feature in most other places) arbitrarily low character limits and restrictions on file uploads and the loose “ai” “policy” and that constant rush to placate corpotechbros of course
There’s people out there on Akkomas, GoToSocials, Snacs, Iceshrimp.NETs, Wafrns, and whatnot. Some people even run their instances, I think it was snac, on microcontroller boards that take a fraction of a watt!
@mi @josemurilo activitypub is literally a w3c recommendation, letting the state/the rich in was literally a design choice. and atproto is just a billionaires' toy lol
I think it was snac, on microcontroller boards that take a fraction of a watt!Ohh... that sounds fun! I'd love to read up on that. Do you have a source?
I have mostly got rid of the ESP32 running the server because it kept breaking at that time so instead i have moved it to a ARM64 phone but it still kept the power efficiency (it only draws a few watts)
an absolute mad lad at work engine swapped a 1974 Volvo, and put a Lexus V8 in it. Here's what it sounds like.
Someone made the mistake of getting me a UV laser engraving machine.
...
(It was me.)
I feel so powerful now, being able to inflict the wonders of Comic Sans onto every single object in my house.
pond now supports image uploads!
Collect any sort of visual inspirations straight from your hard drive or clipboard and mix it with text posts and links.
Stolen from Facebook, a picture of Paris yesterday
https://www.facebook.com/martin.photographe/posts/10164399357138774?ref=embed_post
@fosdembsd weren't they arresting "rowdy youths" during the last heatwave for doing just this and now everybody's doing it?
@atro_city It's now part of the city rules, during heat wave swimming in the Canal St Marin is allowed.
6. Pronounciation
PNG is pronounced "ping".
End of PNG Specificationhttps://www.w3.org/Graphics/PNG/png-spec6.htmlFYM?? it actually is pronounced PING????
@thermia@sk.girlthi.ng i was surprised to find this represented in the final spec also. i think the draft you shared has a much better juxtaposition though. "[states inflammatory opinion] end of specification [leaves, refuses to elaborate]"
boosted@thermia@sk.girlthi.ng what does thermos mean with "FYM" ?
It's Berkeley Breathed's 69th birthday, which seems the perfect day to share the only Bloom County I was ever moved to snip from a newspaper. Though actually, in this terrible, horrible, no good, very bad timeline, any day is the perfect day to share this February 26, 1989 installment.
Released a second EP so I stop thinking about it: https://farawayland.bandcamp.com/album/as-real-as-anything-can-ever-be
RE: https://thepit.social/@peter/116761317566825616
In the observed case: 1.2M+ tokens consumed in ~30 minutes on a task that should have been git clone + find . -name '*.sol'. The recursive agent tree was still growing when observed.
In another instance, this same pathological behavior consumed an entire Pro Max 20x plan 5-hour session token limit in under 5 minutes (4 million tokens consumed).
@jonny "Is there any way to make it play itself?"
"Yes, number of players: zero."
eclexic [he / they (I'm mostly gender... agnostic? Whatever works best for you except "it".)] » 🌐
@theeclecticdyslexic@mstdn.social
@jonny if the definition of insanity is doing the same thing over and over and expecting a different result, does that make this IaaS (insanity as a service)?
@jonny I believe art had something to say about this. Not even literary or fine art. Children’s art.
well see there's your problem
i'm gonna go ahead and bet money the problem is that both spawnDepth and depth are used interchangeably and there is always a fallback condition to 0 when the attribute can't be accessed. fallbacks are the LLMs absolute most favorite thing in the world.
edit: please note that this is a betting statement and i have not analyzed the code beyond an ast diff and am just browsing around. these are not confirmed RE findings.
look man i feel for this person but i am laughing and screaming in my lab right now. every line of this gets better (formatting mine)
Agents fetch individual repo files via HTTP instead of cloning: A poem
They now fetch individual files from GitHub via HTTP
one file at a time.
Each fetch is a separate tool call
carrying the ENTIRE CONVERSATION CONTEXT
hundreds of round-trips
each re-sending the full system prompt
and context
at Opus pricing
man i remember when cloning a repository didn't cost you $30 and you actually got the files instead of nothing.
eclexic [he / they (I'm mostly gender... agnostic? Whatever works best for you except "it".)] » 🔓
@theeclecticdyslexic@mstdn.social
@jonny Now kids let me tell you, back in my day we had these things called "deterministic tools" and they did what you told them to, in fact they did it the same way each time... uphill both ways.
i'm pretty sure the bug here is that the spawnDepth or depth parameters are never actually used for anything or checked against a limit. i believe i have followed all assignments and accesses in 2.1.172, and checked for string constant accesses, but i am not finding any place where they are used.
it's important to remember that there isn't really such a thing as "controlling how agents are spawned" in claude code. Agents can spawn from the agent tool, but agents also have the agent tool. there is even a note in some new migration prompt that i hadn't seen before that says specifically that Opus 4.6 (the model used by the person raising the issue) loves to recursively spawn subagents. that was present before the "feature" of allowing recursive subagents was added in 2.1.172, because technically there wasn't really anything preventing agents from spawning subagents before either - the checkPermissions method from the AgentTool is hardcoded to "allow", or at least it was, i don't have time to align and apply the mapfile to new release versions today.
edit: wait i did find an inequality check to the depth attribute in a place i can map to the source leak. that confirms that it's the agentContextStorage and i can work backwards to how that gets set. so anyway yeah it still just looks like there's a mismatch between the spawnDepth param that's passed around and the depth param that's accessed with a zero-fallback. remarkable. this company has raised how many billion dollars again?
https://neuromatch.social/@jonny/116329076454434307
in reply to »
jonny (nonvenomous) [they/them] » 🔓
@jonny@neuromatch.socialnext puzzle: why in the fuck are some of the tools actually two tools for entering and exiting being in the tool state. none of the other tools are like that. one is simply in the tool state by calling the tool. Plan mode is also an agent. Plan Agent. and Agent is also a tool. Agent Tool. Tools can be agents and agents can be tools. Tools can spawn agents (but they don't need to call the agent tool) and agents can call tools (however there is no tool agent). What is going on. What is anything.
@jonny remembering the time the source leaked and we all learned that it's turboslop under the hood, lmao
i mentioned this in passing yesterday, but in case anyone is wondering "where is the claude code longform piece" I sorta pivoted what I want to write about from "Claude code, what should be the best case scenario for large scale LLM code, is a dumpster fire according to this one snapshot in time" to "that and also it's getting worse not better."
So anyway, that is to say, it is getting worse not better.
eclexic [he / they (I'm mostly gender... agnostic? Whatever works best for you except "it".)] » 🔓
@theeclecticdyslexic@mstdn.social
@jonny Expected. They can only go so long before the Habsburg AI syndrome catches up.
No one seems to be able to filter LLM text reliably. This is the obvious business model for these companies at the moment imo. Release a society wide threat and sell personal protection. The fact they aren't doing this means they can't.
This means LLM text _is_ getting into their training of new models. Curation of sources can only do so much, especially when you work at their scale.
This was foretold.
@jonny Sounds like they saw Gastown and were impressed with its ability to use up huge quantities of tokens.
@jonny They have to drive up innate token usage somehow. No better way than being as inefficient as possible with the rote stuff.
@jonny software archaeology moment: that *is* how the original git-over-HTTP[S] protocol worked. The server published a file that listed the tips of its branches; the client would then have to download every object individually, getting the advertised commits and walking the object graph itself to decide what's needed.
...but nobody does that any more because it's horribly inefficient, so the git protocol (in which walking the graph to determine what the client requires is a server responsibility) was made to work properly over HTTP[S] instead.
And besides, I'm betting that the agents don't even follow the proper protocol for doing that - download the refs, then the objects by ID. Aside from this being a "nobody has done that for over a decade" thing
@cait it's much worse than that - claude code is using its "web fetch tool" which means that for each request it would have had to hit the API with the full context history (much of that would be cached, to be as fair as possible, and thus have reduced API costs). redirects are handled by the web fetch tool returning self-prompt text that says "hey there was a redirect, call this tool again witth this new URL." it almost certainly got the list of files by downloading the html of the github repository page and parsing the html and extracting the a elements - or else it just guessed and requested a random URL that it thought should exist. naturally all of these requests would be initiating new http connections, and so eventually they hit a rate limit. claude code then decided to infinitely spawn more agents to hit it again immediately rather than waiting for the 429 cooldown, each of which adds a whole extra system prompt to the context for the subagent, and some means of spawning subagents can't use message caching and can cost full price.
so behavior like this basically has no precedent in any domain of computing before it for being like the worst possible, most expensive, inefficient way of accomplishing a task that has basically been solved already.
@jonny that's all well and good but it means I don't have to learn 'git clone' which is a scary CLI wotsit and I don't do CLI and it means ordinary people get to make um, crap things and fill the internet with abandoned bollocks...
500€ for a phone that runs a full-blown smartphone OS on a low-end mediatek soc and restricts what apps you can run on it calling it a ‘feature’ is a grift, btw
@mkljczk@pl.fediverse.pl what's the phone? 👀
@burgundia commodore callback
The OS where half of it is open and other half isn't (like the core is open but the Android layer isn't)
Jesus Michał von Gentoo 🏔 (he) » 🌐
@mgorny@social.treehouse.systems
Given these days you can't even expect #Gentoo contributors to be respectable, I'm working on adding a git hook that rejects commits with #LLM attribution. Could you help me find all the common patterns used to mark LLM-assisted #git commits?
So far I'm checking for author and Co-authored-by using the following e-mail patterns:
• copilot@github.com
• *@anthropic.com
• claude@users.noreply.github.com
• *+claude[bot]@users.noreply.github.com
• *@openai.com
• *+chatgpt-codex-connector[bot]@users.noreply.github.com
• *@cursor.com
• *@x.ai
• *@google.com
I think some people came up with some other tags to mark LLM commits but can't find that right now.
New Hetzner pricing is out, some stuff has almost tripled
Do not, under any circumstances, re-scale any VPS you already have or you'll lose your grandfathered pricing https://docs.hetzner.com/general/infrastructure-and-availability/price-adjustment/#cloud-servers
our non-profit is losing 3 of its 9 public pcs. these are public facing PCs we use for teaching programming, game dev, and as general pcs for people to hang out in the community center
the leasing contract is ending, and they won't let us renew because they want to use the machines for 'AI exploration'
Plasma 6.7 is out and landing soon on your distro of choice!
You can look forward to easier menu customization, better control over your mics, and, as for virtual desktops... Well, well, well! You get keyboard navigation and the much anticipated feature of having separate virtual desktops on different monitors.
All the new stuff packed into Plasma 6.7 is here:
https://kde.org/announcements/plasma/6/6.7.0/
#Plasma6 #desktop #FreeSoftware #OpenSource
---
Become a Supporting Member:
And Oxygen and Air are back, baby!
http://kde.org/announcements/plasma/6/6.7.0/#oxygen
---
Become a Supporting Member:
@kde Still no support for the Oxygen background texture in Kirigami apps though… It doesn’t look quite right without.
@kde #Oyxgen and #Air looked so futuristic back in the days. Now they are back as classic look to many but not all of us. Wonder how people see it.
#KDE #linux #Plasma6
#desktop
#FreeSoftware
#OpenSource
| Looks new and fresh to me (haven't seen it before): | 0 |
| Looks classic to me (I am old enough to remember): | 0 |
@kde Oh, the separate virtual desktops per monitor are nice. Somewhat jealous of that one here from the GNOME-sphere.
Congrats with the release, I'll be sure to check it out when I have the time :)
@kde I know this isn't much, but KDE Plasma is the only software I've ever allowed anonymous usage reporting on. I can't wait to see all the improvements coming up once this reaches Garuda :)
@kde I really love KDE a lot. It just works (tm).
But what I can't stand is animations. It's always a lot of work configuring everything to be as instant as possible.
It would be so so so nice to have like one theme that has minimal shiny stuff and minimal animations. Or some global config option.
Otherwise thanks for your work.
Promote sustainable Free & Open Source Software in your community!
https://eco.kde.org/grow-green/
- Read The Opt Green Support Handbook
- Distribute The Sustainable Software Flyer
- Host Install Parties In Your Community #EndOf10 #DIDay
Together we can make a *more free* and *more sustainable* digital society.
#KDE #KDEEco #OptGreen #FreeSoftware #OpenSource #FOSS #FLOSS #SustainableSoftware #GreenSoftware #Sustainability #eWaste
@danmccullough Not fully; cut off a little. That threw me off for a short time.
Nice shot!
@RichiH Sorry about that! It looks just tucked in on the border on the original photo but my eyes are so bleary from staring it for the last 2 hours… 😂 Thanks!