storus

18 ms·

1. 421.▲
   by storus 11mo ago

   With the recent Intel layoffs many Linux drivers are no longer maintained and as a consequence some older laptops don't boot anymore.
2. 422.▲
   by storus 11mo ago

   OK, so it's Valve's version of HP Reverb G2 just released 5 years later. It doesn't even match the best Pimax/HTC, not to mention Apple's one.
3. 423.▲
   by storus 11mo ago

   What are the reputable sellers these days? Wasn't there a recent German scandal where Seagate was selling used hard drives as new to its wholesalers?
4. 424.▲
   by storus 1y ago

   We might not even need RL as DPO has shown.
5. 425.▲
   by storus 1y ago

   Look up Tabata intervals, used to train speed-skating Olympians.
6. 426.▲
   by storus 1y ago

   The point was that in ME/CFS you are essentially a conscious vegetable where drinking a glass of water or crawling to the toilet feels like a massive achievement. That mitochondrial supplements allowed pushing the limit every day a bit
7. 427.▲
   by storus 1y ago

   How about boosting mitochondria via supplements? Would that be something to look at? I climbed out of ME/CFS-like neurocovid mainly thanks to boosting mitochondria as much as I could and am wondering if the same lesson could be applied
8. 428.▲
   by storus 1y ago

   4-minute HIIT run (30s full/5s walk, repeat) makes you vomit and not feel your legs. 15 minutes of HIIT 5-times a week is a wishful thinking. It's not your typical "vigorous" activity. At my athletic very best I could at
9. 429.▲
   by storus 1y ago

   I wish Stanford kept offering CS236 but they haven't run it for two years already :(
10. 430.▲
    by storus 1y ago

    Would 8 minute HIIT a few times a week do the job?
11. 431.▲
    by storus 1y ago

    I used DeepSeek to draft a legal letter for some dispute with some marketplace that didn't want to do what I paid for. Within 2 days after sending that email all was resolved. I would hate to lose that option.
12. 432.▲
    by storus 1y ago

    Dude who became powerful by the simplest business model in existence (buying S&P winners with other people's money, selling losers), suddenly gets to set the policy for the rest of the world. Index funds should never be allowed to
13. 433.▲
    by storus 1y ago

    Google's march to irrelevance continues with full steam.
14. 434.▲
    by storus 1y ago

    Only 3) could scale but then those exam takers not using AI would fail unless they are geniuses in many areas. 1) and 2) can't be done when you have 50-70% of your course consisting of online students (Stanford mixes on-campus with CGO
15. 435.▲
    by storus 1y ago

    Universities and education overall also had their foundation detonated by AI. Some Stanford classes now do 15 minute tricky exams to reduce the chance of cheating with AI (it takes some time to type it so the point is to make the exam so sh
16. 436.▲
    by storus 1y ago

    I had enormous fun optimizng C++ via self-modifying assembly to squeeze the utmost performance of some critical algorithms, and now this drive towards immutable everything feels like cutting my hands and legs off and forcing me to do subpar
17. 437.▲
    by storus 1y ago

    Maybe try NAD+ boosters next? They seem to be reducing addictive behaviors quite a bit. Liposomal NAD+ or nicotinamide riboside or IV NAD+. The theory is an energetic deficit in the brain that drugs/addictions seem to override temporar
18. 438.▲
    by storus 1y ago

    Strix Halo has awful token prefill speed. Only suitable for very small contexts.
19. 439.▲
    by storus 1y ago

    Basic datacenter technicians will be the new astronauts, swapping burnt CPUs and failed hard drives in space.
20. 440.▲
    by storus 1y ago

    Strix Halo can only allocate 96GB RAM to the GPU. So GPT-OSS 120B can be ran only at Q6 at best (but activations would need to be partially stored in the CPU mem then).
21. 441.▲
    by storus 1y ago

    That's not really true, the latest autoregressive image models create a codebook of patches that are then encoded as tokens and image is assembled out of them.
22. 442.▲
    by storus 1y ago

    What I mentioned was the case in some hard CS classes I took there.
23. 443.▲
    by storus 1y ago

    That won't work at elite schools like Stanford where a hard class average is like 98% and 94% will give you B+ due to the opposite curve being applied.
24. 444.▲
    by storus 1y ago

    Excellent! Good luck!
25. 445.▲
    by storus 1y ago

    Wouldn't this restrict memory to 128GB, wasting M3 Ultra potential?
26. 446.▲
    by storus 1y ago

    M5 is supposed to support FP4 natively which would explain the speed up on Q4 quantized models (down from BF16).
27. 447.▲
    by storus 1y ago

    DGX Spark is not for training, only for inference (FP4).
28. 448.▲
    by storus 1y ago

    M3 Ultra has slow GPU and no HW FP4 support so its initial token decoding is going to be slow, practically unusable for 100k+ context sizes. For token generation that is memory bound M3 Ultra would be much faster, but who wants to wait 15 m
29. 449.▲
    by storus 1y ago

    Is ASUS Ascent GX10 and similar from Lenovo etc. 100% compatible with DGX Spark and can be chained together with the same functionality (i.e. ASUS together with Lenovo for 256GB inference)?
30. 450.▲
    by storus 1y ago

    Intel leadership actually reads HN? Mindblown...