Interesting but I can't deal with the AI generated prose.
Sentences like "This was not a garment for walking between a carriage and a doorway. It was ..." show me no human has proof read or edited the text.
Yes you can pin it to core 1 whilst pinning all other tasks to core 0. Then will never be interrupted or preempted (except by interrupts created on core 1)
As expected ANC headphones cancel less noise at low frequencies so I guess the 780Hz is a trade off between high enough frequency to be a bell and low enough frequency to get attenuated a little bit less than high frequencies.
The research paper is pretty poor quality and this is mainly a marketing exercise.
I don't understand why people get so hung up on Chrome using so much memory. A lot of this memory is "discardable" so will get dropped when the system is under memory pressure and the amount of memory allocated for this type of usage will depend on how much memory your system has available. If Chrome is using lots of memory then it's almost always because your system has lots of available memory. It allows the browser to cache large images and video assets that would otherwise have to be re-downloaded over the internet.
Does the KV cache really grow to use more memory than the model weights? The reduction in overall RAM relies on the KV cache being a substantial proportion of the memory usage but with very large models I can't see how that holds true.
This series of graphs https://www.bmj.com/content/bmj/387/bmj-2024-082194/F1.large... shows that whilst those two professions are at the bottom of the distribution they are not particularly outlying and cherry picking of those professions has occurred. The statistical analysis should have adjusted for picking the best 2 occupations of the 443 in the study. That would likely show very little statistical significance.
This is good work. I wish branch predictor were better reverse engineered so CPU simulation could be improved. It would be much better to be able to accurately predict how software will work on other processors in software simulation rather than having to go out and buy hardware to test on (which is the way we still have to do things in 2026)
Only 2 students actually used an LLM in his exam, one well and one poorly so I'm not sure there is much you can draw from this experience.
In my experience LLMs can significantly speed up the process of solving exam questions. They can surface relevant material I don't know about, they can remember how other similar problems are solved a lot better than I can and they can check for any mistakes in my answer. Yes when you get into very niche areas they start to fail (and often in a misleading way) but if you run through practise papers at all you can tell this and either avoid using the LLM or do some fine tuning on past papers.
An interesting idea but in theory just three correct pass codes and some brute force will reveal the secret key so you'd have to be very careful about only inputting the pass code to sites that you trust well.
It's definitely computable on a piece of paper and reasonably secure against replay attacks.
Is it just me or does this have a familiar "edited by Chat GPT" feel to it? I can only take this chatty writing style in moderation but it seems everyone is using Chat GPT to edit their work in the same way.
