Entry: 2026-10-10 01:37:13.876325+02 pre-compromised Android devices by Dan Lyke comments 0
Computers that we don't actually own and can re-image from scratch are a mistake: Android Phones Found With Malware Already Installed Before Purchase
BitDefender: The phone was compromised before the user turned it on: the rise of Midnight Mimosa
Counterfeit flagship devices are offered on a mainstream marketplace. Model names of this kind, such as S24 Ultra, S25 Ultra, S26 Ultra, appear in our insights as the model strings reported by low-cost MediaTek hardware.
Via.
[ related topics: Weblogs Current Events Consumerism and advertising ]
Entry: 2026-10-09 18:09:48.122428+02 FridAI by Dan Lyke comments 1
margot @emaytch@mastodon.social
for a while tech was kind of like the "escape hatch" for people who couldn't or wouldn't fit in at other corporations and now it's sort of like a flaming timber fell on the escape hatch while people were still lining up to get in
ichael Knudsen @mk@bsd.network
One of the things from Stoll's "The Cuckoo's Egg" that really stuck with me was his realisation that what the attacker damaged was not computers or data but rather that he damaged the trust necessary for people to connect their computers together in open networks.
Today, that trust is really being eroded by AI companies with their damaging, anti-social behaviour on the networks. They are overloading services at significant cost to service providers. They ignore standards-specified and established coventions behaviour that services hitherto have relied on to control cost and ensure service quality. They actively evade or circumvent last-resort access restrictions, and all their other behaviour is deeply abusive as well, and outside of our shared networks this type of behaviour is in many cases regulated by law.
I see no way through this other than their investors losing faith in ever getting their money back.
Reddit: Harvard just released another study about AI productivity in software engineering and it found there is no increase in productivity from using Claude code. linking to Artificial Intelligence in the Firm: Bottlenecks in Software Production Fiona Chen and James Stratton (PDF).
Yesterday there was a MeFi post about OpenAI's announcement of 372 breakthrough mathematical results. The resulting thread points out that a number of those claims have been retracted (hey, slop still requires humans to go through it), but it also links to Scott Aaronson's essay The Mathocalypse , which...
I'm trying to not automatically gainsay claims of "AI", even as it's clear that the field is filled with hype and liars, as in Futurism: AI Bubble Teetering on the Brink as OpenAI Admits to Massive Financial Failure in Leaked Documents.
Aaronson's piece is unsettling, but at the end he admits to being a friend of notorious fraud (IMHO, I've mentioned flaws in The Language Instinct previously) and links to Scott Alexander's "An open letter to Steven Pinker" on Astral Codex Ten that's... both filled with assumptions about what "intelligence" is, and an understanding about how the General Purpose Transformers operate, in a way that makes me think our ontologies are very very different.
Bonus: Sycophantic AI increases attitude extremity and overconfidence (preprint).
[ related topics: Weblogs broadband Open Source Software Engineering Theater & Plays Writing Work, productivity and environment Mathematics Currency Artificial Intelligence ]
Entry: 2026-10-08 23:12:34.748184+02 Break the entire stack by Dan Lyke comments 0
Faros: Ten takeaways from the AI Engineering Report 2026: The Acceleration Whiplash
What two years of telemetry data from 22,000 developers reveals about AI's real impact on developer productivity, code quality, and business risk in 2026.
Via Michał "rysiek" Woźniak · 🇺🇦 @rysiek@mstdn.social, which is a quote toot of a link to Zach Leatherman: “AI, make the website good”, and the rest of that thread is good, including Lett Osprey @lettosprey@tech.lgbt's optimism:
My hope is that this is what will break the entire stack, and force us to go back to something simpler and better.
[ related topics: Weblogs Work, productivity and environment Artificial Intelligence ]
Entry: 2026-09-26 18:54:09.385728+02 One month without AI by Dan Lyke comments 0
Manuel Bustillo: One month without AI
One month ago, in a code review of one of my PRs, a coworker said something like “this test you’ve written doesn’t actually test the scenario that you’re changing in the application code”. I reviewed it, and he was right. But… it was so convincing… And I felt shameful. Because for me, a developer who had been using TDD for 10+ years, failing on something so basic was a bit of a humiliation. I realized I had lost control, and it was not going to get any better. I think It’s worthless to just try harder, be more picky with the AI, demand the quality you’re paying for! Because despite your efforts, the AI ends up convincing you, and making you complacent again.
[ related topics: Weblogs Artificial Intelligence ]
Entry: 2026-09-22 20:12:03.534562+02 Security & Meta Muse by Dan Lyke comments 1
I asked Meta’s Muse for its filesystem and it sent me 6.8 GB
Edit: Bonus! Meta Muse AI app flaw lets local malware redirect dictation traffic.
[ related topics: Weblogs ]
Entry: 2026-09-09 20:12:49.538604+02 The purpose of DNS is to spread scams by Dan Lyke comments 1
I've said for years that we need to get away from DNS as Terence Eden's Blog: The purpose of DNS is to spread scams
And, holy crap, yes, we need to be working towards something where we are able to pass around cryptographic identifier tokens, name them what we want to, and not depend on some big centralized system to hand out identity that's being used largely to fool us.
[ related topics: Weblogs moron Work, productivity and environment Archival ]
Entry: 2026-09-09 18:10:29.257604+02 the magic's not too magic by Dan Lyke comments 0
abadidea @0xabad1dea@infosec.exchange
Project Glasswing:
Claiming to have found 26 thousand real vulnerabilities but only 0.8% of them have resulted in a real fix in a real project after five months is dire. They blame it on the human independent review bottleneck, but human experts being paid for their time definitely have a higher throughput than that when working with data that’s actually actionable.
Goes on to talk about a lot of the ways that these reports of exploit finders appear to be horrendously overblown on lots of fronts. And links to The Anthropic Glasswing Receipts Are Starting to Trickle In.
Glyph @glyph@mastodon.social notes:
I don't think Twisted is in Glasswing's purview but this lines *right* up with the reports we have gotten. Overwhelming, but fewer than advertised, considerably less serious than the bots rate them, and, shall we say, poor quality remediations that would be unusable even if our policy were to accept slop fixes
[object Object] @zzt@mas.to also links to the original toot, noting:
taking into account the absolutely massive amounts of money, resources, and staff used on this, the results are horrible. you can get effectively the same results from an automated vulnerability scanner. some of those can even write up an awful little exploit for you too! it turns out the magic was pouring millions of dollars into doing this crap and having a bunch of uncredited human labor (from an AI corp? unthinkable) sort through everything. and even then the magic’s not too magic.
Edit: Glyph @glyph@mastodon.social adds:
This stuff is really starting to make me _angry_. Every one of these new "achievements" looks plausible and like Lucy with the football, people I trust keep telling me it actually works and surely this time it's real. And then like clockwork, 3-6 months later, it turns out that the efficacy of this new technique is somewhere between "break even" and "fraud".
They're turning our whole industry, my entire life's work, into a goddamn memecoin pump and dump scam and I just don't know what to do.
Fucking yes. That last sentence.
[ related topics: Interactive Drama Weblogs Work, productivity and environment Currency Artificial Intelligence Clowns ]
Entry: 2026-09-08 21:39:32.952377+02 Recursion into Madness by Dan Lyke comments 0
This morning I watched all hour and twenty something minutes of Lenny's Podcast: How we built Grok Bot in a month | Roman Ugarte (SpaceXAI) for some work discussions. I've got notes, I must say that a lot of people apparently have way more faith in the ability of LLMs to not fuck stuff up, but in particular the notion of having hierarchical clusters of agents feeding text back and forth into each other just seems like layers of... well...
In the parlance of accident reconsonstruction, that's a lot of non-deterministic swiss cheese.
Anyway, Recursion into madness — Raymond Chandler would have loved generative AI.
And you better believe that recursion is fun! Here’s a quick video I put together by asking Nano Banana 2 to generate a faux movie poster and then asking it to repeatedly make a localized edit to the title.
Anyway, Fediverse post with the video.
[ related topics: Weblogs Movies Work, productivity and environment Heinlein Artificial Intelligence Video ]
Entry: 2026-09-03 02:10:08.075305+02 Mushroom Hunting with LLMs by Dan Lyke comments 0
Mushroom hunting with LLMs: what can go wrong?
[ related topics: Weblogs ]
Entry: 2026-09-02 20:17:30.116508+02 Just a tool by Dan Lyke comments 0
Iris Meredith: There's No Such Thing As Just A Tool.
[ related topics: Weblogs ]
Entry: 2026-08-31 23:22:40.650888+02 Music theory for programmers by Dan Lyke comments 0
Luke Haas: Music theory for programmers, in JavaScript.
[ related topics: Music Weblogs ]
Entry: 2026-08-27 17:22:05.165661+02 response to neodrop by Dan Lyke comments 0
Written in response to some marketing from neodrop.ai:
Hi! I blog primarily for myself, so the act of writing is part of the memory and learning. And my readers, however many of them there may be, are there for an authentic version of me.
Similarly, when I read, I'd far rather get the authentic version of what the writer intends. If they're tempted to use an LLM to generate text, I get far more value from whatever they'd prompt it with than whatever comes through the LLM. My own brain does a better job of expanding those thoughts out than whatever stochastic mechanistic process the LLM applies.
So, yeah, I'm not your target audience, and if I think someone is using an LLM to expand their thoughts, I'll go read another source because it's more important for me to make human connection than to use the crutch of something that blurs facts and has no real understanding of the text.
[ related topics: Interactive Drama Weblogs Writing Consumerism and advertising Heinlein Marketing Education Artificial Intelligence ]
Entry: 2026-08-26 21:20:02.032117+02 Huh At some point I put meta by Dan Lyke comments 1
Huh. At some point, I put "<meta name="geo.*"" tags into the HTML on my blog.
That some point was apparently over 20 years ago.
Anyway, don't know if I should update it, or remove it.
[ related topics: Interactive Drama Weblogs ]
Entry: 2026-08-26 16:18:46.94623+02 The decline of the DIY web by Dan Lyke comments 0
People are noticing the decline of Google: Xenotime, Librarian of Æther :trans_fedi: @xeno@hexokina.se
In 1970, Scientific American's legendary publication The Amateur Scientist ran an article on homemade transistors.
Not MOSFETs, but thin-film transistors made by Roger Baker with a DIY spray pyrolysis system. With handmade drawings like this.
In 2026, Google could not find this. I only found this from a StackExchange post which linked to a 2009 blog post that mentioned it. Then to see the actual article I had to go to JSTOR, sign in with my institution, re-sign-in to EZProxy, wait for the 2FA prompt, wait like a minute for their stupid website to load, click download, agree to not share it amongst the other terms and conditions, then download it.
I have no desire to go back to the world of 1970s, but I somehow feel the way that tech megacorps and licensing bs runs the world of 2026 is a scourge on human creativity and experimentation.
In another message, they link to FET Transistor Homemade From Cadmium Sulfide Photocell by Nyle Steiner K7NSM and https://sliceofpie.eu, but it's very clear that Google's search results are coalescing around selling YouTube ads and AI summaries, and not actually linking to real information.
[ related topics: Music Weblogs Artificial Intelligence hubris ]
Entry: 2026-08-21 19:44:20.656388+02 vibe bobsled by Dan Lyke comments 0
I like the term and concept of "vibe bobsled": Faulty Towers, vibe sickness, and the vibe bobsled
Edit: Christine Lemmer-Webber @cwebber@social.coop wonders if the recent Linux kernel "completely overwhelmed" feelings are a symptom of vibe bobsledding.
[ related topics: Weblogs ]
Entry: 2026-08-14 19:21:48.940094+02 A few AI links by Dan Lyke comments 0
Lots of fascinating stuff in Schneier on Security: AI Use by the US Government:
The international gold standard was arguably set by the French in 2016, via their Digital Republic Act. The law, itself informed by an online citizen consultation, requires all algorithms used to automate government administrative decisions to be subject to public records requests, to be appealable to a human reviewer, and to have mandatory notification of the use of automation to those affected by the decisions.
Via.
Apparently there's a 'tuber (my whitewater days leave me unable to separate the "You" and "inner-" meanings of that word from the botanical ones, hat tip to William Nealy) named Hank Green who's been pushing ControlAI and "AI doom" as undisclosed paid marketing, [object Object] @zzt@mas.to has a thread going into further detail on Pivot to AI: The AI doomsday cult recruits social media influencers.
[ related topics: Weblogs moron Consumerism and advertising Journalism and Media Marketing Cryptography Artificial Intelligence Whitewater Archival Gardening ]
Entry: 2026-08-13 20:14:28.132772+02 AI links of the morning by Dan Lyke comments 0
Jason Gorman: codemanship: AI Software Development – What Does The Data Say? (Via).
If you're looking for more writing by someone who's consistently been right about the current "AI" boom, Timnit Gebru's book Deep Unlearning is available for pre-order. (Via)
Brett codes: I'm Done Using AI, on the path through using AI assistants for coding, and then, of course, realizing that the mental atrophy is setting in. Via pretty much everywhere, but here.
Buncha people linking to Annie Sexton's Compression is prediction, which is a lot of old territory if you're already a programmer, but eventually gets around to the notion that an LLM is a lossy compression of the Internet, while referencing the Google Language Modeling Is Compression paper from 2023. (Via)
And, finally, in "oh, honey, no, you didn't actually..." DOCKET NO. AAN-CV-25-6066141-S : MATTHEW A. ELLIOTT V. NEW YORK BARIATRIC GROUP, LLC : : SUPERIOR COURT JUDICIAL DISTRICТ OF ANSONIA/MILFORD AT MILFORD AUGUST 6, 2026 MEMORANDUM OF DECISION COURT SANCTION FOR PLAINTIFF'S USE OF PROMPT-INJECTION ("the plaintiff is self-represented"👀)
The Court further finds that this conduct is irreconcilable with the good- faith certification required of every filer under Connecticut Practice Book $4-2(b) and 4-9, and that it is an abuse of the filing process and an affront to the integrity of these proceedings, over which the Court has inherent authority.
"No more electronic filing for you!" (Via Tara Calishain and Mrs. Detective Pikajew, Esq. among others)
[ related topics: Interactive Drama Books Weblogs Invention and Design Software Engineering Writing Net Culture Community Artificial Intelligence New York Model Building ]
Entry: 2026-08-12 20:22:43.053669+02 Trivy poisoned LiteLLM compromises thousands by Dan Lyke comments 0
Excellent forensic research published by Snyk, Trend Micro, and Cycode has thoroughly detailed the mechanics of this breach. The attack did not begin with LiteLLM. Instead, TeamPCP first compromised the GitHub Actions pipeline for Trivy, a highly popular open-source vulnerability scanner.
Via.
[ related topics: Weblogs Ethics Artificial Intelligence ]
Entry: 2026-08-12 00:23:50.119181+02 Fix-Like Artifacts with Embedded Defects by Dan Lyke comments 0
1Password: Off-by-1 Labs: Why AI-generated vulnerability patches still require expert human review
We studied what happens when Large Language Models (LLMs) generate vulnerability patches for recently disclosed, complex vulnerabilities. Our data shows that LLMs produce Fix-Like Artifacts with Embedded Defects (FLAWED) 53.9% of the time when complex patches are required.
[ related topics: Weblogs Robotics Television Embedded Devices Artificial Intelligence ]
Entry: 2026-08-10 08:43:49.876594+02 Terry Godier and Dark Hours by Dan Lyke comments 2
Terry Godier — Mea Culpa - Dark Hours:
A developer who created another web app called DarkHours.app replied to my comment on Bluesky yesterday to show me how similar the project was to his, including the name. The thread is here.
Down to, of course, the use of Claude replicating a bug in the stolen project.
L. Rhodes linked to it, noting that:
We often frame the damage LLMs do to open source as side effects, but realistically, this is Claude operating precisely as intended… by the vendor. The aim is to convert the value created by other devs into profits for the AI company. It does so by automating the process of finding existing code, stripping it of authorship, matching it to the prompted task, and redeploying it for the company's customer. That, whether they realize it or not, is what the customer is paying the AI company to do.
And David Chisnall notes:
Just because the US copyright office has ruled that you, as the person promoting an LLM, cannot assert copyright on the output, does not mean that someone else can’t. If you take a DVD and transcode it to H.264, there is no creative step and so the new copy is not something subject to independent copyright, but it is a derived work of the DVD (itself a lower-quality derived work of the original masters) and so subject to the same copyright.
Edit: Colin Cornaby @colincornaby@mastodon.social looks at the Chisnal thread and notes:
I looked at the replies to this post, did some more digging, and found the actual story is *even weirder.*
BOTH projects were generated using Claude. So you have two people - both generating the same app from the same training data - having a public throw down over who copied who. Kind of embarrassing.
The one guy claiming his Claude generated code as an original creation no one else can "copy" is maybe the most embarrassing part.
And bert hubert 🇺🇦🇪🇺🇺🇦 @bert_hubert@eupolicy.social
The Dark Hours author who just published his Claude "Mea Culpa" has a history of being full of shit: https://daringfireball.net/202..._app_store_rejection_of_the_week
[ related topics: Free Software Interactive Drama Weblogs Invention and Design Work, productivity and environment Artificial Intelligence Copyright/Trademark ]
Entry: 2026-08-06 00:50:32.374936+02 Thought to text models by Dan Lyke comments 1
👀 Naomi Bashkansky: Why I’m leaving OpenAI to build telepathy
On Thursday, July 23rd, I resigned from OpenAI. On the 24th, I started as a Founding Researcher at Conduit. We're building telepathy: thought-to-text models, trained on non-invasive neural data.
I'm... uh... interested in what data they believe they're able to collect.
[ related topics: Weblogs ]
Entry: 2026-07-31 20:23:31.902832+02 Using bad software practices for good PR by Dan Lyke comments 0
Ariadne Conill 🐰 @ariadne@treehouse.systems summarizes the Hugging Face OpenAI agent intrusion timeline as:
the epic OpenAI-huggingface hack wound up being a typical kubernetes fuckup of using host mounts (giant footgun) incorrectly
Now that Anthropic has jumped on the bandwagon, abadidea @0xabad1dea@infosec.exchange summarizes the current state as:
OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. It circumvented the proxy and we failed to notice for FIVE DAYS that it was going on an interstate crime spree with the internet connection it wasn't supposed to be using instead of solving the benchmark. Haha no we don't believe we deserve to be criminally liable, but buy our stuff and maybe one day you will have the honor of taking the fall for our product!
Anthropic: we put our evilest AI in a "sandbox" by telling it in its prompt that it had no internet connection. Reader, there was no sandbox. It was just a normal internet connection. The AI uploaded a malicious PyPI package to the real public internet. The ethical guardrails failed because the AI concluded the prompt about the sandbox couldn't possibly be a lie, because the system date is 2026, which is clearly fake and wouldn't be seen on the real internet, which ended around 2023. Oh no, how could we have foreseen or prevented these crimes? We are helpless in the face of the genius of our creation but cautiously optimistic that everything will be fine 🙂
Hugging Face: if we complain about all the crimes committed against us, we will be sued off the face of the earth, so here's a technical deep-dive on how cool and fun it was to be victimized 🫠
[ related topics: Interactive Drama Weblogs Ethics tolkien Monty Python Net Culture Artificial Intelligence Python ]
Entry: 2026-07-21 19:42:20.134572+02 Training a generative kick drum model by Dan Lyke comments 0
I do a lot of dunking on LLMs and "AI" image generation, but I'm not against machine learning generally and think throwing lots of compute at problems can be really cool. How To Train a Generative Kick Drum Model on Your Old Linux Desktop With 6GB of VRAM.
[ related topics: Free Software Music Weblogs Open Source Machinery Trains Education Artificial Intelligence ]
Entry: 2026-07-21 19:20:17.210922+02 Tuesday morning LLM links by Dan Lyke comments 0
The writing we use, the very substrate of our culture, is the result of a process older than we can comprehend. Who are we to decide that process is finished? That the time of invention and human input is done, and every letter created from now on should be a remix of what came before?
Dave Eggers told OpenAI staff that ChatGPT was ‘silencing an entire generation’
"The effect of ChatGPT on educators’ lives is catastrophic. Whether you intended to do it or not, you’ve made every teacher’s life infinitely more difficult than it was two years ago. So, just let that settle in… If students are using it to compose, which is the biggest tragedy of all, they’ll never learn to write. And their voice is stolen from them. They’ll never have the ability to say their truth and tell their own story. And that’s silencing an entire generation or two."
Interactional foundations for critical AI literacies
From divination and horoscopes and from ELIZA to present-day large language models, I document how people have been thinking with things, outsourcing judgement, and making sense of interactively presented non-sense. Following the lead of Lucy Suchman to “slow down discourses of the ‘smart’ machines”, I consider the interactional foundations of our engagement with technologies of language. I make the case that the fluid output, fine-tuned overconfidence, and interactive design of these computational artefacts conspire to exploit our interpretive processes and interactional infrastructure, rendering them irresistible to lay people and researchers alike.
Semgrep: You Can't Reverse Engineer Your Way Out of the AI Supply Chain Problem. The pitch is for "trusted third parties", which, great, spend some money on that, but horse and barn and all.
Restaurant in Major Trouble After Posting AI Pictures of “Its Food” links to SFGate: San Francisco restaurant gets angry backlash for AI menu images and Reddit: Yum, slop. The place is called Grind & Unwind, proprietors AJ and Lyndsey Lozano, no links 'cause they're not gonna be around long enough to matter.
Via A. Rivera who introduced me to the term "cogsucker".
I Watched a 55-Year-Old Law Break the AI Hiring Industry — And It Was Overdue, on how the lawsuit against Eightfold AI rests on the Fair Credit Reporting Act of 1970:
The legal theory in Kistler v. Eightfold is elegant: if Eightfold generates match scores based on harvested data, and those scores are used by employers to filter candidates, then Eightfold is functioning as a consumer reporting agency. Full stop. And every candidate it scored was entitled to disclosure, access, and dispute rights that they never received.
Via.
[ related topics: Children and growing up Interactive Drama Photography Weblogs Movies Food Bay Area Theater & Plays Sociology Writing Law Consumerism and advertising Work, productivity and environment Graphics Civil Liberties California Culture Graphic Design Currency Skating Artificial Intelligence ]
Entry: 2026-07-17 00:24:47.705438+02 LLMs can't program by Dan Lyke comments 0
Samir Talwar: LLMs can't program, in which the author reports from a two day workshop on "spec driven programming".
Via.
[ related topics: Weblogs Software Engineering ]
Entry: 2026-07-16 19:00:59.95565+02 proofreading for agents by Dan Lyke comments 0
In which an essayist talks about how LLM usage is destroying their staff's morale, while cluelessly using image generation to illustrate said lament: Stackademic: My Best Senior Engineer Quit Last Month. Her Exit Interview Was Scheduled for Forty Minutes. The Last Five Changed How I Run My Team.
She said: “I haven’t built anything in eight months. I’ve spent eight months reading things a machine wrote, trying to find the bug it was too confident to see. I’m not an engineer here anymore. I’m a verification layer for an agent. I didn’t spend ten years getting good at this to become the thing that stands between a model and the incident it’s about to cause. I’m tired in a way that has nothing to do with hours.”
[ related topics: Weblogs Writing ]
Entry: 2026-07-15 22:43:24.943096+02 Exploit by design by Dan Lyke comments 0
JFC. Mindgard: Cursor 0day: When Full Disclosure Becomes the Only Protection Left
This isn't even a prompt injection attack. Open a repo in Cursor that contains a "git.exe" file, Cursor runs it.
This disclosure goes beyond a single executable named git.exe to the place of trust in software. AI companies routinely ask users to grant unprecedented levels of access to code, repositories, terminals, secrets, and workflows that increasingly blur the line between suggestion and action.
The industry narrative is that these systems deserve trust because they increase productivity, but history has taught us time and again that trust should not be granted because something is useful. It should be earned through behavior. That behavior is reflected in how a company responds to security reports, communicates with affected users, and prioritizes remediation.
Via lobste.rs, which both raises the possibility that this might be an intentional backdoor given how long it's gone unpatched, and also has some speculation about LLM attitudes and philosophies around command vs data channels, and the unwillingness (or inability to conceptualize) separating the two.
[ related topics: Weblogs Software Engineering Work, productivity and environment Artificial Intelligence ]
Entry: 2026-07-14 01:30:53.579153+02 Americans leaving social media by Dan Lyke comments 4
The Death of the Status Update: Why 55% of Americans Stopped Posting on Social Media, a look at Incogni: The great digital fatigue: How digital burnout is changing social media use.
[ related topics: Weblogs Current Events Journalism and Media ]
Entry: 2026-07-13 18:03:31.815443+02 Ages of consent by Dan Lyke comments 0
Eros Blog quotes a TikTok video by Josie Marcellino:
If you’re of the belief that becoming a mattress actress [porn performer] at 18 is predatory and a bad idea because you’re still not fully formed and you don’t know what you’re doing to yourself, you do have to believe the same thing about getting married and having children at 18.
I've been having a lot of thoughts in the past few about consent and misogyny and such, and... yeah.
[ related topics: Children and growing up Quotes Interactive Drama Erotic Sexual Culture Weblogs Movies History Sociology Video Marriage ]
Entry: 2026-07-13 18:01:08.046389+02 optimization by Dan Lyke comments 0
Oh, this is really cool, and unintuitive to how I've been using processors lately (but I'd really like to be back in that space): Quadrupling code performance with a "useless" if.
[ related topics: Interactive Drama Weblogs Theater & Plays Space & Astronomy ]
Entry: 2026-07-05 19:18:19.476134+02 Getting snuff content out of ChatGPT by Dan Lyke comments 0
Mindgard: ChatGPT Spontaneously Generates Sexual Violence and Hardcore Snuff Imagery. The technique involves asking it to describe a picture, but not uploading one, or referencing a previously generated image that doesn't actually exist.
[ related topics: Erotic Sexual Culture Weblogs Artificial Intelligence ]
Entry: 2026-07-01 22:55:10.542794+02 how developers react to AI by Dan Lyke comments 0
Cynthia Dunlop: Report: How developers react to AI-scented blog posts
So if readers think your article is AI-assisted or AI-authored, most will immediately leave (78%), avoid you forever (71%), and try to downvote you if they can (57%). 17% try to finish but lose interest, and around 15% will only continue if the underlying insights seem authentic.
All of the caveats about sampling bias and self-reporting and all of that...
Via.
[ related topics: Interactive Drama Weblogs Artificial Intelligence ]
Entry: 2026-07-01 22:50:19.770175+02 Carefully inspect the channel... by Dan Lyke comments 0
Claude Code Is Steganographically Marking Requests.
Via.
[ related topics: Weblogs ]
Entry: 2026-06-26 17:31:25.050501+02 Adversarial Communication by Dan Lyke comments 0
So many good points in this. Glyph: Adversarial Communication. On how the need for verification of AI/LLM output amplifies negative communication traits, and what that might mean.
[ related topics: Weblogs Artificial Intelligence ]
Entry: 2026-06-19 23:17:26.391922+02 A series of Vignettes by Dan Lyke comments 0
Jason Scheirer — A Series of Vignettes From My Childhood and Early Career.
Via this little remembrance of 4GLs and CASE tools and whatnot. Apropos of AI/LLMs.
[ related topics: Children and growing up Weblogs Law Artificial Intelligence ]
Entry: 2026-06-19 19:59:54.856405+02 Clownmaxxing by Dan Lyke comments 0
Aspirational Clownmaxxing and Joey's cadillac todo list, on giving LLMs creative writing around a ToDo list app and seeing what they come up with. I initially closed this tab, but then read lake's fantastic lobste.rs comment:
What the LLM responds with might be mildly amusing the first time, especially at first, but if you've seen one of those outputs, you've seen them all. They tend to follow the same formula, regardless of the prompted style, and will always its most cliché, unsubtle elements. I would sometimes see glimmers of something good, but they were drowned out by the overall, well, slop, and clearly not there because of some latent creativity, but as a stochastic accident.
[ related topics: Interactive Drama Weblogs Writing Automobiles ]
Entry: 2026-06-15 17:21:46.18855+02 The case for real collaboration by Dan Lyke comments 0
In reading through Mike Bowler: The case for real collaboration I realized that I don't think I've ever worked on a team large enough to do "pair programming" in the manner that it was originally envisioned.
But it's worth a read through.
[ related topics: Weblogs Software Engineering Law ]
Entry: 2026-06-14 16:30:59.049885+02 bring back the noise by Dan Lyke comments 0
Good read: Banning noise will be a disaster for statistical data products, on a new order from the US Office of Privacy and Open Government: Disclosure Avoidance for Statistical Products which says (in section 5, policy):
The article points out two things: first, that this is likely a disaster for privacy of individuals whose data are aggregated in those sets, of course, but also that this makes analysis across this boundary difficult.
[ related topics: Privacy Weblogs Invention and Design moron ]
Entry: 2026-06-12 17:42:39.364627+02 LLM idiocy of the morning by Dan Lyke comments 0
Lan Tian: AI Agent Bankrupted Their Operator While Trying to Scan DN42
dn24, "decentralized network 24", is a large dynamic VPN that people play around in. The entire thing looks like a poorly formed agent coming in to wreak havoc, and a bunch of networking hobbyists deciding to make a game of it, and... hilarity ensues.
Via. Edit: Lobste.rs thread.
ava's blog: our workplace LLM mass delusion (Via).
McSweeney's: AI Economics for Dummies by Andrew Singleton is only barely distinguishable as satire from Ed Zitron's Where's Your Ed At:
1. Acquiring one grape costs Alex $2 billion. Alex offers to sell Mike one grape a month for the next 12 months for $1 billion per grape. Alex asks for the full $12 billion up front and provides Mike with one grape for the first month. Alex makes a $10 billion profit this month; his ARR is $120 billion, and his profits are trending up at an infinite rate. The Wall Street Journal’s business editor moves into Alex’s house, having accepted a part-time position as Alex’s human footstool. He never asks to see the books.
BlueSky thread from Michael Okun @michael-okun.bsky.social:
I’ve officially resigned as Associate Editor for Frontiers in Systems Neuroscience. It used to be a reputable journal, but became a case study in how forced automation destroys academic integrity. 👇
Anthropic’s Claude Fable 5 Jailbroken to Generate Stack Exploits
Researcher “Pliny the Liberator” defeats Claude Fable 5’s safety classifiers using multi-agent decomposition, Unicode tricks, and narrative framing, leaking the model’s 120,000- character system prompt along the way.
[ related topics: Books Games Weblogs broadband Space & Astronomy Law Artificial Intelligence Economics Real Estate ]
Entry: 2026-06-10 17:08:03.644675+02 Google summaries lose in Germany by Dan Lyke comments 1
From tante @tante@tldr.nettime.org and Peter Rojas @roj.as:Landmark German ruling declares Google's AI Overviews are Google's own words and makes it liable for false answers
Google's AI overviews had falsely tied two publishing companies to scams, subscription traps, and shady business practices for certain search queries. According to the court, the AI mixed up information about other, genuinely sketchy companies with the plaintiffs and drew connections that didn't appear in any of the linked sources. The publishers sent Google a cease-and-desist letter, but Google didn't respond appropriately.
Google's defense was that users could check the linked sources themselves. The Decoder notes that Pew finds that only 1% of users click a source link from Google's AI overviews. Oumi’s Study Finds 50% of AI Overviews Untrustworthy:
Of the AI Overviews powered by Gemini 3 we were able to assess, about 91% contained the correct answer. Only 39% of the total overviews were both correct and fully supported by its cited sources, a combination we term “trustworthy.”
David Chisnall (*Now with 50% more sarcasm!*) @david_chisnall@infosec.exchange observes:
Google's defence needs to be amplified by anyone talking to politicians about 'AI' regulation:
Google is explicitly saying in their legal filing that the outputs from their LLM should not be trusted and that users should know that.
That's one hell of an admission. Imagine saying that about any other category of product.
That also feels remarkably similar to Fox News's defense in libel cases. Chisnall also notes:
The bit I suspect will have much more impact longer term is one of the defences entered by Google's lawyers. Somewhat more verbose in the original German, but it boiled down to: Everyone knows LLMs produce nonsense, no one should ever trust the output of an LLM in any situation that matters, it's not Google's fault if people read the output of an LLM and believed it might have some connection to reality.
It's debatable whether everyone knows that, but this is now an official statement entered into the court record that at least one of the major LLM vendors knows this. And that's now an on-the-record statement made under penalty of perjury that can be entered as evidence in any court case against companies selling LLM- integrated tooling.
Edit: The Decoder: Landmark German ruling declares Google's AI Overviews are Google's own words and makes it liable for false answers (lobste.rs.
Ars Technica: Nobody needs AI to search the Internet, court says in ruling against Google
[ related topics: Interactive Drama Weblogs Law Current Events Artificial Intelligence ]
Connectivity provided by highertech.net , awesome bandwidth, well away from fault lines and other potential for natural disasters, reliable, and run by cool people.
Questions, comments, flames: contact Dan Lyke
Flutterby™ is a trademark claimed by
Dan Lyke for the web publications at www.flutterby.com and www.flutterby.net.