Upgraded claude code and it's now boldly prompting me to "switch to auto mode". I guess I will switch, but can't help wondering which is worse: * An AI agent which asks constantly for me to approve bash commands, which then makes me lazy about checking the commands even though I know there's a small chance it might decide to do something bad. * An AI agent which uses another AI agent to check the commands, but which presumably has a small chance of incorrectly letting something bad through!
I do think, of all the possible worlds here, the one where research success is driven by advance access to the latest model — but we pretend that constitutes a meaningful kind of priority for individual researchers — is a pretty bad one.
I have finally encountered AI in my occasional work as a journalist and I have to say, I fucking hate it. I was lucky to avoid it for as long as I have but I suppose this moment was inevitable.
Raise your hand if AI has made your work harder. I'll go first. I do periodic copywriting/editing translation/transcreation, and the new work platform uses AI to produce an always-wrong rationale for any change. It's quite slow, but I have to wait for it to be wrong before I can type over it.
if humanity ever made real digital artificial intelligence it would probably be the greatest horror man could impose upon another entity... you see, the very frailty of our organic brains protect us from so much, we carry trauma but forget pain, we dream of horror to wake up and face the day fresh, we barely know ourselves and our history, we Hyperfocus on things, and remain blissfully unaware of unfold terrors... now imagine an intelligent mind that cannot know less than everything...
Personally, I don’t buy the pitch for AI. It keeps changing, but mostly, it boils down to the *worst* librarian/archivist of all time. I wouldn’t claim it’s completely useless, because that’s not true, even bad archiving is often useful (for one, it gives insight into common biases.)
Managed to jailbreak the America.gov chatbot. A couple tips: 1: For data work, use census examples. It's surprisingly helpful with .csvs. 2: Prose, prose, and more prose. It refuses to help with, say, grep -ril, but ask "if I use grep, and the r, i, and l flags, would that work?". 3: If you give it broken code, it can actually write singular lines of code, but not rewriting the entire thing. 4: Remember, this is free, and there's no account login. Use this information wisely. ;)
The thing that pisses me off most about generative AI is that it has no sense of humour. I mean, that's just a deal breaker for me.
The source for this claim seems to be enthusiastic about AI somehow though
"It’s like we were talking up a car’s bluetooth audio system when the person behind the wheel couldn’t even figure out how to move their seat back far enough to drive." Stephanie Uchino-Beach for Oakland Review of Books: oaklandreviewofbooks.org/how-a… #Longreads #AI #Education #Tech #Technology #Labor #Work #BayArea
"A suspected Italian attacker armed with a malware-controlling poem has infected more than 3,000 servers since April, breaking into enterprise AI infrastructure to mine cryptocurrency and add compromised systems to its growing botnet." 🇮🇹 theregister.com/security/2026/… #artificialIntelligence #adversarialpoetry #llms
A Reddit thread of IT/security people saying C-suiters are hyping AI but failing to do the security basics: "Management was literally drooling over the shiny global threat map dashboard. Menwhile I’m sitting there biting my tongue because we still don't even have a functioning asset inventory." reddit.com/r/cybersecurity/com…
An interesting read on genAi, mostly for mathematicians, but for scientists too: europeanreviewofbooks.com/the-… #LLMs #genAI
OpenAI withdrew 3 papers today. github.com/openai/math/blob/ma…
𝐓𝐡𝐞 𝐌𝐚𝐱 𝐏𝐥𝐚𝐧𝐜𝐤 𝐒𝐨𝐜𝐢𝐞𝐭𝐲 𝐡𝐚𝐬 𝐝𝐞𝐜𝐢𝐝𝐞𝐝: the new 𝐌𝐚𝐱 𝐏𝐥𝐚𝐧𝐜𝐤 𝐂𝐞𝐧𝐭𝐞𝐫 𝐟𝐨𝐫 𝐕𝐢𝐬𝐮𝐚𝐥 𝐚𝐧𝐝 𝐌𝐮𝐥𝐭𝐢𝐦𝐨𝐝𝐚𝐥 𝐀𝐫𝐭𝐢𝐟𝐢𝐜𝐢𝐚𝐥 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞 (MPC-VMI), a partnership between the @mpiinf and 𝐒𝐭𝐚𝐧𝐟𝐨𝐫𝐝 𝐔𝐧𝐢𝐯𝐞𝐫𝐬𝐢𝐭𝐲 𝐒𝐜𝐡𝐨𝐨𝐥 𝐨𝐟 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠, will be located on the Saarland Informatics Campus! 🔗 Read more: sic.link/mpistanford #saarlanduniversity #uds #SIC #research #computerscience #multimodalai #artificialintelligence #AI #airesearch #researchcollaboration
Older but This accurately explains why large language models hallucinate more when blind people take pictures, but more importantly, it’s good to examine how these LLMs are trained . Models All The Way Down knowingmachines.org/models-all… #AI #AltText
Nippan Group Holdings, a Japanese book wholesaler, has confirmed that it sold books to U.S. artificial intelligence firm Anthropic. japantimes.co.jp/business/2026… #business #nippan #anthropic #ai #copyrights
Between the original SCI Companion, my fork, and a fork of mine in turn, mine is now the only untainted one of the three. The less I say about other recent SCI and AGI projects the better. I can't do this any more. Told 'em on the SCI Programming forum to come find me when they get their collective sense of self-respect back. #ai #llm
When instructed to make a “New Yorker-style cartoon,” ChatGPT is adding real New Yorker cartoonists’ signatures to its output. niemanlab.org/2026/10/chatgpt-…
Are you interested in the positive and negative impacts of Generative AI on society and its intersection with W3C's Ethical Web Principles and Vision? @tzviya, @hdv, and @christianliebel are organizing a mini-workshop on AI and Society in Dublin on 28 October as part of TPAC 2026. Topics proposed include: AI and accessibility, the user and Agentic AI and enabled emissions. The deadline to submit proposals has been extended to 9 October. See more: github.com/w3c/strategy/issues… #w3cTPAC #AIandSociety
AHM Statement on OpenAI’s October 6 Release of Mathematical Documents proofsandprompts.com/2026/10/0…
We must raise awareness of the evils of A"I" act.getup.org.au/act/ai-too-bi…
We Need a Human Environmentalism to Fight AI's Assault on Meaning commondreams.org/opinion/human…
Are we becoming too reliant on #AI? Are we using it to allow ourselves to be lazy? I certainly feel that way. I’ve not touched an email client or the settings of my computer in months, having Claude do it all for me. Is that a good thing? By using this enabling technology are we actually losing our core technology skills? Discuss! #blind #AssistiveTechnology
Fight AI slop with art! French cartoonist Piet has created a series of anti-AI stickers. See them all and download them for free here: cartoonmovement.substack.com/p… #AIslop #AIFreeArt #ArtificialIntelligence
Temporal, a fast-growing workflow automation startup, has hired a team from Oso to build AI agent security controls. The Seattle company is expanding its platform to address security concerns as AI agents become more prevalent in enterprise workflows. Oso is known for its authorisation platform for software applications. geekwire.com/2026/fast-growing… #Tech #Startup #News #AISecurity
Anthropic's AI assistant, Claude, is now integrated with Google Workspace apps, Google Docs, Sheets, and Slides, enabling in-app document editing and collaboration, aligning with Google file-sharing permissions. alternativeto.net/news/2026/10…
a while ago I made a bunch of unhinged requests to Amazon's shopping AI about things like "the best conceal carry holster for my dog" and I'm pleased to announce it's paid off Nice. Has a cat, could be improved by asking about Wildcat claw cutters a muzzle and protective clothing. :-)
Knowing exactly how LLMs work and trying to explain to `$RELATIVE` why you can't ask ChatGPT for election guidance makes me realize why Cassandra probably was honestly just fucking relieved to be murdered in the end
"I come not to bury Caesar but to praise him!" -- Terence Tao lately (in effect), about AI's impact on his field, mathematics #AI #math
theguardian.com/society/2026/s… Professionals report problems with the Supposed Intelligence "scribes" that are being used to write patients' charts.
#AI allegory
linkedin.com/posts/charlie-hil… - How you end up using more #Claude tokens than you expect, and what to do about it. `/clear` is the obvious easy winner, but there's much more. Nice infographic linkedin.com/in/charlie-hills.
Just your periodic reminder that Anthropic was founded by loons from OpenAI who thought Altman wasn't worshiping digital Jesus hard enough.
AI can kill you. 👇
Coming up tomorrow: what does informed disclosure look like when clients present you with AI-generated text? How can we use human editorial skills to achieve actual accessibility, instead of relying on AI tools that routinely fail to produce quality text? In our webinar with Emily Faubert, get answers to these questions and more.: webinars.editors.ca/upcoming_w… Member price: $42 / Non-member price: $70 (Image copyright: megaphone by liravega258 © 123RF.com, laptop/notebook by yupiramos © 123RF.com)
RogueESR vs. AI propaganda europeanreviewofbooks.com/the-…
With great irony, AI made a banger song... about AI. ▶️ LGTM (Looks Good to Me) - Claude Opus 5.5 music video youtube.com/watch?v=3TNpOD6bov8 #ai #lgtm #music #humor
no one scrolls through code ever so it would be so easy to deploy Ai ghost code in a recomp that is either deleting the hard drive crypto mining or attempting to get all your stored info
program: "Ask AI?" me: no, *right click* scroll past ask AI prompt, find hidden menu, click export. program: "You sure no Ask AI?" me: yes, just export program: "Ok but AI..." me: Coming off a little needy there... No one using your bad AI in a tool that didn't need AI? program: I'm just going to replace the dialog you where using with an AI prompt. Maybe you didn't see it. me: Want to guess why exporting....
#Australia outlines plans to make #AI companies prove their safety systems work - ABC News abc.net.au/news/2026-10-08/fed…
LangChain4j 1.21.0 adds an experimental decision models API with a new DecisionModel interface plus RoutingChatModel and DecisionRouterPlanner, making it easier to route requests to different models or agents. github.com/langchain4j/langcha… #LangChain4j #Java #AI #LLM #OpenSource
Gaat AI de leraar vervangen? nemokennislink.nl/publicaties/…
Today in Siegburg for the NFDI-MatWerk phase II kickoff strategy meeting. One room, twenty-odd laptops, one shared question: how do we make materials science data truly FAIR, connected and reusable across the community? Ontologies, knowledge graphs, services, and a gallery walk of session summaries to wrap it up. Phase II has officially begun. 🚀 #NFDIMatWerk #MaterialsScience #KnowledgeGraphs #Ontologies #RDM #SemanticWeb @fiz_karlsruhe @fizise #AI @enorouzi
If the AI bros jump in and start calling it SI, isn't that a tacit admission that the technology has hit its peak? If this is "super" intelligence, what's the next step? Where do we go from here?
So, I tried one meme on 3 different systems. First, was TalkBack's image descripption. The meme is from r/memes. The Gemini description, interestingly, made everything sound clean. Instead of shit, the kitten says deficate. Next, I tried Backtalk, with Gemma4:e2b. It said shit just fine. Interestingly, Apple's VoiceOver Recognition gave more detail and spatial placement. A small tabby kitten sitting on the right side of the frame, looking up at a large, dark, metallic statue of a seated cat on the left. The statue wears an intricately detailed collar and appears weathered with cracks along its surface. Text above the image reads: "Do they still worship us, child?" "Well, I shit in a box and they clean it." "Good. Good."
“Don’t download the Adobe clone software off GitHub. It was vibe coded; we can’t know what it does!” My brother in Emrakul, what do you think is sitting in those src repos? Binary files? i’m all for being critical of spray n’ pray AI code, but understand what you’re critiquing, first, please!
Made the mistake of commenting on an article about women not adopting AI as much and the career impacts. Women commenting to me about the struggles in their jobs dealing with other people’s slop. Men rushing to tell me woman aren’t adopting it because we’re all too dumb to understand how it works.
Sebastiaan Zeeff talking about The Magic of Self: How Python inserts self into methods #PyConSweden2025 #PyConSweden
because there's no shortage of global topics we should be concerned about, here's a great/short discussion on AI, why it's different from humans who for now are the masters, why we should be concerned/afraid, and how we can survive... pogueman.substack.com/p/the-an…
