nerdsniper

6 ms·

1. 61.▲
   by nerdsniper 2mo ago

   I got 1419 miles by putting the pin just east of Presidio, TX. The Gulf of Mexico doesn't have many Americans in it on the way out to the Florida Keys. 1,570 miles by pinning the lower Keweenaw Peninsula which null-counts a great circl
2. 62.▲
   by nerdsniper 2mo ago

   I'd love something like this for the Apple Watch but with local (iPhone) or self-hosted processing using whisper/parakeet/etc.
3. 63.▲
   by nerdsniper 2mo ago

   Fine-tuning is done against a dataset. Distillation is done against a model.
4. 64.▲
   by nerdsniper 2mo ago

   I mean they could just be routing known benchmark questions (which all of SWEBench are) to a full-performance variant.
5. 65.▲
   by nerdsniper 2mo ago

   > buying exercise equipment prescription only Like eyeglasses for near-sighted people. Obviously prescriptions should never be needed for far-sighted people. Only for near-sighted people.
6. 66.▲
   by nerdsniper 2mo ago

   Silly complaint. Per HN guidelines as long as a workaround is available, paywalled articles are explicitly allowed. Try this service: https://archive.ph/eJpUg
7. 67.▲
   by nerdsniper 2mo ago

   I feel like this would be much less of an issue if the datacenters were all fully solar+battery powered.
8. 68.▲
   by nerdsniper 2mo ago

   Currently it either means “better than the median human at every task including things like counting to 1000 or folding clothes” or it means “for each task, better than the top-n human at that task”. It’s clearly not actually that “generali
9. 69.▲
   by nerdsniper 3mo ago

   Or a seedbox on another continent.
10. 70.▲
    by nerdsniper 3mo ago

    That seems like entirely reasonable variance to me for AI models. For my purposes, that absolutely counts as a solid "replication". I'd probably accept +/- 5 percentage points even.
11. 71.▲
    by nerdsniper 3mo ago

    Edit: It was pointed out to me that Opus 4.8 got "21%" for successfully fully completing ~1-in-5 tasks, but also got "55.7%" for obtaining significant partial credit on some of the ~4-in-5 tasks it could not fully comple
12. 72.▲
    by nerdsniper 3mo ago

    Is this really the value-add comment you’re going with?
13. 73.▲
    by nerdsniper 3mo ago

    I took multivariate calculus (I to III), differential equations, statistics, discrete math, and linear algebra. I didn't take those courses you're talking about, but that's my point. I have a top <1% education in math in t
14. 74.▲
    by nerdsniper 3mo ago

    Right. With a college degree, I understand most of these words. I don’t have a clue about the impact of the research or how it fits into the corpus, but I understand the gist of the article title. Despite my degree being in engineering, whi
15. 75.▲
    by nerdsniper 3mo ago

    Source? Wow.
16. 76.▲
    by nerdsniper 3mo ago

    Delve. Not Deel. Very different startups.
17. 77.▲
    by nerdsniper 3mo ago

    I think both of you meant “Delve”, not “Deel”. Deel is a pretty successful HR startup that’s still growing at a good rate and AFAIK free of major scandals. Again, Deel is HR, not SOC2. Delve was the SOC2 company described in the article lin
18. 78.▲
    by nerdsniper 3mo ago

    > Heavy industry is mixed, with decent success in consumer vehicles now but little purchase in stuff like heavy equipment or shipbuilding Uhh, China does over half of the world’s shipbuilding. China is also responsible for 25% [1] of the
19. 79.▲
    by nerdsniper 3mo ago

    EDIT: OH MY GOD THAT IS A HUGE CHANGE!! On MacOS: Settings -> Accessibility -> Display -> Reduce Motion. I had no idea that was on. I like motion, I wouldn't have set this myself! ------------------ Original: How does that mak
20. 80.▲
    by nerdsniper 3mo ago

    I'm having trouble identifying any physics at all in this UI design system.
21. 81.▲
    by nerdsniper 3mo ago

    Yah thats definitely Minnesota, eh?
22. 82.▲
    by nerdsniper 3mo ago

    The trick for better quality low-skill soldering is: 1) Use a leaded solder with flux mixed in like the Multicore 502 [0] 2) Use more flux. 3) Use little alligator clips on adjustable arms to hold the stuff together for you so you can keep
23. 83.▲
    by nerdsniper 3mo ago

    They have little solar panels.
24. 84.▲
    by nerdsniper 3mo ago

    Sure - it's just hard for rando's to get tons and tons of real human interaction data to run a GAN against. "How to do the training" isn't the barrier for this, and not worth keeping a secret.
25. 85.▲
    by nerdsniper 3mo ago

    CF doesn’t block archive. Archive poisons CF. Explanation direct from the CEO of CloudFlare: https://news.ycombinator.com/item?id=19828702
26. 86.▲
    by nerdsniper 3mo ago

    Beating this would require a large amount of sophistication, not a small amount. Basic machine learning clustering will expose bots mouse+keyboard+touch behavior and discriminate them from humans. It will also likely discriminate against an
27. 87.▲
    by nerdsniper 3mo ago

    Also decent embeddings models could run locally and just trickle through the user folder if that was so necessary.
28. 88.▲
    by nerdsniper 3mo ago

    That would be “gray hat”. It would be illegal, though it has happened in the past and to my knowledge no one has yet been prosecuted for illegally fixing vulnerabilities. But it’s definitely not white hat.
29. 89.▲
    by nerdsniper 3mo ago

    Nearly every stock consumer router keeps getting these RCE’s. Perhaps that’s the point, they can get any arbitrary household in trouble if it’s expedient to do so.
30. 90.▲
    by nerdsniper 3mo ago

    I'd be curious to look for any investigation into whether extended paxlovid treatment fixes a significant subset of long COVID.