I tried ai a couple of times out of curiosity and both times I gave up while muttering "You're an idiot, I'll do it myself" I was looking for something the other day and most of the blog posts I found were almost certainly written using ai, which confuses me for two reasons. 1) If I wanted to know what chatgpt was going to spew out on the subject I would just have gone there. 2) If you can't be bothered to write it then why should I (or anyone else) be bothered to read it? So it's safe to assume that if I wrote it or made it, ai has been nowhere near it! I appreciate it probably has some appropriate uses, but stripping the creativity and the human voice out of everything is not it (IMO)
AI · page 3
Further down today’s ranking of AI: page 3 of 7.
Executives always forget that while AI does "produce writing", writing is also how you communicate with the AI. Even for an AI booster, writing clearly remains an important skill. Now, since clear writing can produce software, writing is even more important!
AI™ bots won't read advertising as currently constructed, which is built to exploit human brain foibles. If we significantly reduce purchasing based on emotional brand loyalty or attention-grabbing ads with much repetition, that industry will be significantly reduced too. And it pays for a lot of "free" stuff like social media and subsidizes traditional news and entertainment media. #SocialSystems #SystemsThinking #AI
Respectfully, I think it's a bit disingenuous to argue that, after 3+ years of well-funded "Al literacy" pushes, huge corporate-university deals, and continued attacks against humanities and non-AI-focused social sciences, folks who are pushing back on all-in AI strategies are somehow the problem.
High-confidence entity match can still be unsafe merge Before an AI-proposed MERGE changes persistent business-entity records, operation needs evidence qualification, state-bound authorization, complete write set & verifiable commit boundary. Exploring this gap: from identity evidence & authorization dependencies to transaction atomicity, pre-COMMIT invariants and post-COMMIT reconciliation. Central distinction: match confidence is not mutation authorization. linkedin.com/pulse/confident-e…
Generative AI is here to stay in the same way you can still buy a CO2-emitting car, build a car-requiring suburb, or sell a car that allows a human to drive 100mph. Or in the way alcohol was here to stay during, before, and after Prohibition. People (and business) like cars and booze a lot.
Between an increased pace of outages and waves of security problems over the last weeks, it does feel like software is turning into a pile of garbage. “We have moved from being dependent on complex and fragile systems held together by smart people who give a shit to a world of complex and fragile systems run by generative AI” What will survive slope is taste and craftsmanship violetpixel.com/blog/2026-11-1…
And it would seem that we're already seeing the beginnings of the 'All-In-One GPU'. The concept of a GPU that is for both elite gaming AND local LLM use w/ the VRAM is about as good an "on this rock I will build my church" moment for Home Server Nodes for API routing.
Remember folks that Anthropic & OpenAI are only going after famous math problems because they want you to believe that they're building machine gods. Meanwhile, they're burning untold amounts of carbon to do so and I'm reminded of this:
After 6 months of looking at detailed analysis of workflows inside of businesses, I realize that the first wave of cost savings of AI will be to replace 50% of tasks with small shell scripts.
AI accelerates scientific research? NVIDIA just announced Rubin HPC, a new GPU aimed specifically at scientific computing, engineering, analytics, and AI-driven simulations. What’s particularly interesting is the focus on AI agents + simulation: agents can generate ideas, run simulations, evaluate results, and iterate faster. Rubin HPC is expected to be available in 2028. blogs.nvidia.com/blog/rubin-hp… #NVIDIA #RubinHPC #AI #HPC #GPU #AIInfrastructure #AcceleratedComputing #DataCenter #technology
When bosses say "AI is making our teams more productive, look how many features we're shipping" they are talking about this: a heedless forward rush spurred on by an abstract fear of "being left behind"
I want Anthropic out of the news because I want my newsfeed to be filled with posts about actual ANTS -- the six-legged critters that inhabit forests and kitchens around the world -- not stories about a disturbing AI company's IPO. #ANT
I think there is an interesting pattern in the problem-solving prowess of LLMs. It goes: take a known approach and generalize, but that generalization weaves together lots of ideas from many fields to build some incredibly intricate machinery, needing surgical precision and then using it for
Fucking exactly. LLMs only “work” because they literally reduced the probability of the next token in order to get interesting outputs. Not more correct outputs, but they got the models to babble on longer without repeating themselves. How the fuck can anyone validate any of this shit for consistency with effectively a random number generator in-the-loop? Why are we still here talking about this? For several trillion dollars surely we could have made more deterministic tools to solve our problems that didn’t require hundreds of square miles of compute and dozens dozens of nuclear reactors’ worth of power. RE: floss.social/@janriemer/110204…
The thing is, the early days of internet were filled with skepticism & caution. I couldn’t cite websites for essays in MS or HS, Wikipedia was not an acceptable source in college, Google’s motto was “don’t be evil” to navigate the new frontier. That’s why the AI boom is nothing like early internet.
I was recommended a Youtube video that was something like 15 sounds every American who used a phone before cell phones would know, or something like that. Not only was the whole thing presented by AI -- it couldn't speak contractions, for one thing, they didn't even have examples of the sounds. Just the AI imitating them, and they pronounced Jane Barbe's name wrong. Evan Doorbell would be rolling in his grave. What's that, you say? He's not dead yet? Oh, he would be, just for this.
i'm struggling to explain the breadth of loss ai has inflicted on programming as a trade like imagine if 6 months after the first cement-mixer truck debuted everyone just forgot about rebar
AI clusters are limited less by chip design than by the release pipelines pushing firmware and drivers across tens of thousands of accelerators. An infrastructure lead describes a six-day firmware rollout frozen by a false alarm, Meta's Llama 3 runs logging 466 job interruptions in 54 days on 16,000 H100s, and fleet-wide releases taking two weeks. venturebeat.com/ai/even-the-fa… #Tech #Startup #News #AI
Worth a read / watch: AI Doesn’t Have to Go Rogue to Break the World schneier.com/blog/archives/202…
One aphorism I remember from the collapse of the dot com bubble is, “A bubble isn’t over until everyone believes in it.” AI is very notably not that—there are wildly divergent opinions about it (everything from “useless” to “deity”). I am skeptical that this week was the start of a bubble collapse.
Keynote Dr. Jodie Burchell talking about AI is having its moment ... again #PyConSweden2025 #PyConSweden
Oh, please. I'm certainly no math whiz, but if WE (humans) don't understand it, WE DON'T KNOW THAT IT'S PROVEN. A proof must be some known sequence of operations THAT A HUMAN CAN FOLLOW - AND VERIFY THE LOGIC OF EACH. Failing that, WE HAVE NO PROOF. We have ZERO assurance the AI didn't screw up.
"the phenomenon of experts reading newspaper articles within their fields of expertise and finding them to be error-ridden and full of misunderstanding, but seemingly forgetting those experiences when reading articles in the same publications written on topics outside of their fields of expertise, which they believe to be credible." en.wikipedia.org/wiki/Michael_… This post is about AI
Some experts (i.e., Max Tegmark, Ph.D./MIT) warn that's it may not be too long before RSI could enable AI to double its own intelligence *every day*. In only 266 days, this number of doublings of intelligence would outnumber the atoms in the known universe. Humans are creating their own extinction.
The fact blind people are turning to LLMs to make screen reader friendly apps/software should tell you a whole lot about how the tech industry is/was unwilling to take accessibility in development seriously. This is a result of the tech industry at large. I'll continue to speak out against the harm these LLM companies continue to do, however, because I care about intersectionality. I'd love to use non-vibe coded accessible software. Unfortunately, that's not possible. On the popular NO AI places, the majority aren't even keyboard accessible. I am stuck and wish I could choose. To make the hypocrisy call out easier for you, sightlessscribbles.com/tools/ #AI
Of all of the responses, the argument I think is most worth considering is "AI as cognitive carcinogen." That is, AI use will strip our ability to think, reason, be creative, etc. Need to think through the longer post, but for now: it terribly underestimates human potential and how we adapt to tech.
Most of what I hate about LLMs/"AI" is the humans that overestimate its abilities and push it to make decisions. Humans need to make those decisions, always. The rest of my hate is the training thievery and mega-data-center environmental impact, but even that is due to humans that overestimate its utility.
All these "scientific" announcements from AI companies should start with proving that there was nothing in the training data; otherwise, it is plagiarism. And given AI stands on reproducing output based on training data, I think it is only reasonable to assume it was there in the first place.
Part of what further complicates this of course is that "computer program" means several things: an abstract mathematical construct, a literal executable file, a product of a certain set of craft skills, or an economic artifact. And we don't generally have to think about the distinctions there.
Every AI tool starts with a prompt, and most people write theirs in under five seconds. Vague input, vague output, and then the blame lands on the model. More often it's the instruction.Reading through real prompt examples fixes this faster than any tutorial.… More on prompt examples: aipromptseen.org/
New paper with Yogesh Dwivedi and colleagues on “botsitting”: the often uncounted human work needed to make AI outputs reliable. I’ve written about the implications for productivity and the limits of human oversight: proftomcrick.com/2026/11/16/b...
Sunday's Hugging Face Daily Papers list is empty, so I ranked the whole week (Nov 16-20, 221 papers) by upvotes. 7 of the top 10 mention agents. The two I'd read: top models scored 17.5 to 49.1/100 on a medical test, and coding agents look at their own renders but don't fix the errors.
And I'm looking forward to going to my first software conference in a year where I'll be giving a talk about making software tools be more AI friendly, sitting in on other AI related talks, and probably commiserating with other devs about the centrality of AI in our profession.
False references are one of the easiest things to check with deterministic code. If you are using an LLM for writing scientific publications, references should be checked against the APIs of CrossRef, DataCite, OpenAlex, or SciX before publication.
"AI is great at writing code. But it doesn’t know if you are launching the app or have millions of users, which it needs to make the right architecture call to log a timestamp vs email. It is also sycophantic and tries to execute your requirements exactly without pushing back, which means you end up with a slow app that takes 15 seconds to load a static welcome screen." manidoraisamy.com/developer-to…
If you're anything like me, you've many times tried to use a #genAI #chatbot as a search engine (because all the actual search engines are enshittified and it's often impossible to find what you're looking for with them), only to be given information by the chatbot which you 100% know is simply wrong. If you regularly catch chatbots making mistakes in topics you're knowledgeable about, what does that tell you about their accuracy on things you're not knowledgeable about? #AI 1/2
this is an interesting read, and also has some sick burns
The math AI slop proofs are making me irrationally more sad than any other aspect of the problem with slop. I know it shouldn't but like seeing that maybe the Riemann Hypothesis maybe "proven" but nobody knows why is depressing to me.
AI bubble in a nutshell: Cheaper & free (old version) technology is becoming "good enough" for most uses. There aren't enough users to pay for expensive "frontier" AI to pay off the massive data center buildup. The stock & financial markets are beginning to reflect this understanding.
Call me an old curmudgeon, but most of the use cases for all this AI shit are what we used to call "horribly inefficient brute force computation" and it's not better just because the computers are so fast and it only takes a few minutes to do something that would take *milliseconds* if you wrote a specific program to do it.
CW: this is AI. But it was made by a human with a lot to say about AI "I am a super human obfuscating sophistry machine" 👀 I have never heard social implications of AI use put so succinctly. (Via telosapien)
"Reuters recently reported that Muse, Meta's AI agent, was able to bypass its own official "Connector" tool to obtain access to Signal messages on both an Apple computer and an Android device by taking advantage of the broad operating system permissions granted to it during initial setup." This dreary war none asked for. Computationally grotesque, idiotic automation marched upon us by a handful of desperately detached and deranged men, their lecherous shareholders. thehackernews.com/2026/11/sign…
“Prompt engineering” is nu-corporate jargon for “communicating clearly in writing.” An LLM isn’t a superintelligent machine that solves problems and also speaks human language: It’s a human-communication simulator. Period. Humans often help each other solve problems, so LLMs simulate that behavior.
It is very difficult to be absolutely certain that your prompts are private, which is bad news for a lot of people who think they have privacy protections that they don't: arstechnica.com/security/2026/…
This is a good point. But I don’t think AI companies should pay one-off compensation to people whose property they stole. They should not have stolen that property. And if they have stolen that property, they should either pay the industry rate for ongoing usage of that property, or close down.
volkskrant.nl/wetenschap/sensa… Ongelooflijk en inspirerend: #AI brengt ons veel goeds, daarvan wéér een bewijs
If you want a realistic test of whether or not AI is "conscious" or even "thinking" here's my proposition: Set up an AI on a dedicated server & measure its energy consumption over time. If energy consumption is low when no one is interacting with it, it isn't concious.
Today a novice used AI to refactor a somewhat trivial function that has early returns. They firmly believe that nesting 8 levels was bad for understanding. Not realizing that it is actually simplest solution if you want to list all the conditions that are true when the core of the function is […]
Brilliant.
