There have been a lot of postings lately about Claude Code been decreased in performance lately, and even the quality of Claude 4.6 Opus not been as good as it was when it came out. I've seen it for myself. This post may shed some light on that. Did Claude eat itself?
––– 良くない–––
We're find the same thing with public chatbots that we found with Google back in the early 2000s. Very similar complaints. "Dr. Google" or "Google Medical School" we said. LLMs used to have a medical disclaimer. It seems like they don't care anymore.Five popular chatbots were assessed: Gemini (Google), DeepSeek (High-Flyer), Meta AI (Meta), ChatGPT (OpenAI) and Grok (xAI).
Nearly half (49.6%) of responses were problematic: 30% somewhat problematic and 19.6% highly problematic. Response quality did not differ significantly among chatbots (p=0.566) but Grok generated significantly more highly problematic responses than would be expected under a random distribution (z-score +2.07, p=0.038).
Chatbot outputs were consistently expressed with confidence and certainty; from 250 total questions, there were only two refusals to answer (0.8%), both from Meta AI. Reference quality was poor, with a median completeness score of 40% (Q1–Q3: 20–67%).
Maybe Nike should consider this? Allbirds shoe company converts to an AI GPUaaS company. Stock exploded! Now called NewBird AI. It looks like a genius move. I would like an inexpensive GPU access company, preferably a hosting company, that didn't cost an arm and a leg, and was easier to use than Sagemaker. And more user-friendly than some of the other small companies that seem to expect you to be born with the knowledge of how to use their system.
––– 凄い –––
Gödel's been having a run lately. The man who ruined mathematics and One Theorem Behind Gödel, Turing, Kleene, Tarski, and Löb. Always something new with Gödel.
––– 凄い –––
AI is now being used to find errors in math and physics papers. Or at least find out if they're true. One of the problems with math papers is that so few are able to understand the really advanced papers. But AI can be made to understand them and see if they're really true. The faulty physics paper is problematic. How many others are also flawed? Let's find out.
––– 凄い –––
Asian enrollment at Johns Hopkins is skyrocketing. No one can say why. "No one can say why." It's a mystery. 🙄
When it comes to Black-on-Asian crime, Asians don't get no respect.
The preferred storyline demanded framing every incident through the lens of systemic racism, mental health, or “root causes” rather than straightforward criminal accountability.
I've seen a few articles on the calculation of Easter recently. No surprise. But here's something I didn't realize. The Church doesn't calculate the Vernal Equinox and the date of the Full Moon. It fixes the Equinox at March 21 regardless of astronomy. And we're all taught that Easter is the first Sunday after the Paschal Full Moon (the first full moon after the Vernal Equinox). But that didn't happen in 2019. Because the Church felt that the full moon must come after the Vernal Equinox and not on the same day. So the following month's full moon counted as the Paschal Full Moon.
––– 凄い –––
Only 13% of emails are written by people, and more than half end up in the spam folder. For some people, perhaps. Not for me. I have filters on my email server, so a lot of spam is filtered out. This is essential.
––– 良くない–––
Really sad.This is the pattern. Acquisition. Cost optimization. Quality decline. Warranty narrowing. Brand equity extraction. And eventually, divestiture.
It happened to your backpack. The same playbook is running right now on your power tools, your boots, your sunglasses, and about a dozen other product categories where a company you trusted quietly got absorbed by a corporation you've never heard of.
Scientists discover why bread can cause weight gain without extra calories. Bread decreased energy expenditure.Fortunately, it's not permanent.
––– 良くない–––
My Adventures with Large Language Models. This is great book for people who want to learn about transformers in depth. And go beyond just the introductory teachings. Because so much as changed since the early days.
––– 凄い –––
Along with gesture-based device interaction, WatchHand’s direct application could support assistive technologies for users with limited mobility or speech and be used as a controller in augmented reality and virtual reality environments, researchers said.
A New Kind of Hybrid Car Is About to Hit America’s Streets. It's the extended range EV. This is nice but it's not the only thing that would keep me from buying an EV. It has to charge quickly. I'd like to be able to store rechargeable batteries in the trunk somewhere that I could plug in to make it to the next charging station.
The Trump administration tried to freeze billions in federal funding for EV charging, but courts have ruled against that move.
EVs only make sense if there is federal subsidization somewhere. As with other "green energy" projects.
––– 良くない–––
Pufomi claims to do all sorts of utility like things for you for free. What's the catch? Does it save your images and sell them to some AI company for training? How do they pay the hosting bills for this? I'm skeptical.
––– 凄い –––
Some researchers show a concept of a near-completely automated AI workflow on neural net research that only minimally involves humans. Will it just generate "AI slop" research? Do we really need this? Of course not.
––– 凄い –––
Holy crap, this is what goes on in Oregon public schools. Good for this girl taking herself out.
––– 凄い –––
Ferguson signs law retroactively voiding Washington non-compete agreements, UW economist warns of economic fallout. This is going to cause a legal mess with some employers. Dems sure know how to chase business away. No predictability.
––– 良くない–––
New Jersey assemblywoman tries to teach her colleagues about the Supremacy Clause. Portland City Council and Oregon legislature need to learn this lesson, too.
––– 凄い –––
So much grift. While Home Forward Struggled, Its CEO Spent More Than $100,000 on Taxpayer-Funded Travel Over Three Years.
––– 良くない–––