"it doesn't silently go unnoticed", "would be silently inert", "instead of silently overwriting", "you can never silently overwrite"
The biggest tell for me is overuse of the term "silently". "quietly" is another one you often see from Claude in particular. Models love adverbs for whatever reason, whereas a human writer would use them in moderation for emphasis or prefer terms like "by accident".
I read this when it came out as a long time Ballard fan. It fleshed out a little more about his life I didn’t know, but becomes a weird read for the final third where the focus is on the co-author dying and the author switches to his wife, an originally unintended co-author. They certainly had a story of their own to tell, but it felt rather odd.
This has long been one of my favorite papers, if only for the opening example. When I walked through it, it was an eye opener to a different level of thinking about code, one that assembly and democoders are probably working in every day.
Cache writes/reads are the majority of the picture for agentic development. When you see these people saying a coding run "used 200 million tokens" or whatever, most of that is cache reads, so it should be the headline price IMO (and is one reason why DeepSeek API is so striking with its minuscule cache pricing).
Probably easier to think of it more as a fast, native-backed library for a JS runtime like Node to build Markdown processors/pipelines or even just do parsing. (Though it does have a WASM build for browser/edge use.)
I wouldn't be surprised if it were more accurate based on the errors I've seen. I always eyeball the books and was confused when a £15k building popped up on our asset sheet. It turns out a "workshop" had been categorised as a building we had purchased, rather than the training session it actually was.
This is the importance of having layers and multiple sets of eyes on things, though. Even if it had got past me, my accountant would have surely queried it at year end, but that could be true of an LLM mistake too.
Back in 1999, the UK had its first total solar eclipse for several decades and VH1 played the music video (though, not this one ;-)) on loop for an hour while it was happening.
I was intrigued, because even a broad location given for an IP address is "geolocation data", but the law says "precise geolocation data" which limits it to device-reported data, I assume.
(I'm naive in this area, but..) I wonder if the various "proof of age" laws coming into play will clash with the GDPR in insidious ways. Like requiring identity providers to hold definitive "proof" of why they made an assessment rather than merely proving and discarding. I assume/hope there is some cryptographic way to do this rather than hang on to passport and ID images, however.
I'm pro-AI, but when I can't even give Claude a photo of my fingers clearly on four different piano keys and get an accurate result of which keys I'm pressing, I'm bearish on its visual analyses. (With that said, I imagine it's been trained on more photos of MRI scans than people playing piano, so I could see how it might provide interesting ideas for a radiologist to consider.)
They could treat the extreme spec machines separately from the prosumer ones, like they did with the Xserve. Let business customers spec up to 768GB (say) who are prepared for a $20-25k price tag, while keeping them away from the stores and usual consumer supply chains (Amazon et al). It may not be a big enough market segment for them to care about anymore, though.
(I know this is not how business works, but..) I worked out if they ate a $200 per Mac bump themselves, their reserves would run out in 58 years at current sales rates :-D
More realistically, though, I'm surprised they didn't eat it up until new releases when they often increase prices. All the current models will be gone in a year and they'd probably barely notice that. Perhaps they've been eating it up for the past year or two and push came to shove.
Nothing mentioned about the use of AI by elementary school teachers who may well be using it to generate sub-par worksheets or to rapidly, and potentially inaccurately, mark work.
Every special event flyer I get from my kids' school now seems to be AI generated. I'd be surprised if quizzes and worksheets don't head the same way.
Not quite what you suggested, but I did some experiments several months ago "enhancing" the samples in tracker music with some models, and they sounded terrible. There really is something about the sound of tracker files that's just right. But sure, you could generate lo-fi samples, there's a lot of computer generated samples in music, but putting them together into a pleasing combination is the hard bit.
If you're just having fun with it, there are a whole bunch of other things that produce interesting options, like asking it to theme according to a movie (think Clockwork Orange, Backrooms, anything with a strong aesthetic), or throw screenshots and photos at it and use it as a "design system" (magazine/print layouts can work well with this on stronger models).
As a few people have asked for screenshots, I spun it up. Here's a video of the basic gameplay: https://peterc.org/misc/fpscob.mp4 .. it's clunky, but it does play.
Yeah, the patched llama.cpp. The reason is I saw that using the Q4 quant on vLLM is discouraged and the int8 won't fit on my 3090 Ti, but I could certainly give it a go. I also skipped Transformers as it needs to download the full weights and quantize them locally and I didn't fancy waiting for a 50GB download.
I'm not getting anywhere near the speeds advertised on my 3090 Ti, alas, but it's fun watching it "fill out" its answers. I did Simon's "SVG pelican on a bicycle" test on it and the result was quite minimalistic but fit the brief: https://gist.github.com/peterc/7672e74ec1437945e5fca5ce2c1c9... -- this was on the Q4 quant running on patched llama.cpp. I will be interested to see if Simon's looks much different.
Stuttering John used to do this back on Howard Stern by asking celebrities questions that were far out of the expected gamut at red carpet events. This was all for shock/comedy value, but "who are you and what makes you famous" type questions can really throw celebs off script: https://www.youtube.com/watch?v=8P0hENpnMXk
I'm not the OP and I imagine all cases are different, but my dad was a software developer who had early cognitive decline in his 60s (he died of vascular dementia recently) and he used to talk about it a lot. He said it was like his tolerance for complexity kept closing in.
Where he could once hold an entire system and its details in his head (almost an essential skill in the 80s/90s), he could only instead focus on smaller pieces at a time. Any new tooling or approaches that came along, he was fascinated to hear about them, but no longer felt able to pick them up. He could still solve algorithmic problems and debug "in the small", but it was like he had to do math on a Post-it note where once he had a huge sheet of paper.
Its image processing is terrible. I ran several tests against it against Qwen 3.5 0.8b (yes, 7% the size) and Qwen beat it every time with Gemma often getting things entirely wrong. I even gave it a plain image saying "This is a test" and it thought for 6 minutes trying to analyze it and failed. Qwen 3.5 0.8b confidently got it in under a second.
It may be that the Q6 quant I got is borked (or my LM Studio is), but either way, the 0.8b's performance is mind boggling in comparison.
Can't speak for browser demos, but I just got the ternary model working on my M5 generating images. The 1 bit didn't work, as it has a known bug with XCode 24.5 and I wasn't in the mood for installing 24.4 alongside.
This year seems to be turning a bit of a corner. Of the top box office movies so far this year there's Michael, Project Hail Mary, Hoppers, Wuthering Heights, GOAT.. with Obsession and Backrooms rapidly rising.
Last year it was basically F1 and Minecraft (and while not sequels, both are arguably well known "franchises" outside of movies - but I guess MJ and Wuthering Heights are too ;-)).
There are some older studies that showed nicotine (not via smoking, which makes sleep apnea worse) had a positive effect on sleep apnea as well due to the stimulation increasing muscular activity in the airways. Obviously lots of downsides too so it never caught on, but this seems to have a similar mechanism.
I'm only an ardent viewer of crash investigation stuff, not a pro, but it seems to be a good way to show specifics of warning noises, engine sounds, unusual cabin noises (if relevant) and sometimes even structural failures happening over time in a more direct way from the cockpit voice recorder without sharing the actual "audio".
Criticism noted! But it is experimental, unconstrained FFI is inherently unsafe, and not all JavaScript developers are "mildly educated web devs".
I doubt we'll see this go into mainstream dependencies likely to be used by such developers – it hasn't in Ruby which has had easily accessible FFI mechanisms for years.