bigyabai

12 ms·

271.
by bigyabai 2mo ago
Most of those accommodations are for macOS, and make very few (if any) concessions for alternative operating systems. We're talking Qualcomm-tier effort. Intel and AMD set the standard for "tremendous effort" bringing up supp
272.
by bigyabai 2mo ago
How sure? https://en.wikipedia.org/wiki/Section_230
273.
by bigyabai 2mo ago
The entirety of 2001 probably has fewer camera cuts than this 1 minute advertisement does.
274.
by bigyabai 2mo ago
It's the typical think tank pugilism crap. Donate enough, and they'll write you a dissertation on the economic consequences of flying pigs: The Manhattan Institute for Policy Research (renamed in 1981 from the International Ce
275.
by bigyabai 2mo ago
Nothing wrong with old macOS. I doubt anyone is running post-Big Sur for the "looks and the aesthetic" of Liquid Glass...
276.
by bigyabai 2mo ago
It's feeling good on non-pelican workloads too. Less verbose than 5.3, cheaper/higher usage limits, vision capability and some good web design one-shots even with vision disabled. With GLM 5.1 and 5.2, the big problem was tool cal
277.
by bigyabai 2mo ago
Well yeah, those first two lines have been true for the past 25 years and AI doesn't really change the shape of the problem. > The only way to defend ourselves is to protest and legislate against this You can protest and legislate a
278.
by bigyabai 2mo ago
Cerebras "competes" with Nvidia in the same way a Vespa scooter competes with a Ford F-150. Groq and Tenstorrent are in a similar boat, ASICs don't really threaten CUDA. Curiously, there is not a single real CUDA competitor a
279.
by bigyabai 2mo ago
I can't fully understand this mindset. If you despise AI watermarks, why not rent your own inference hardware and use open-weight models without watermarks? Or LORAs that remove them? AI steganography exists because all customers are a
280.
by bigyabai 2mo ago
> I don't see how NVIDIA can keep their spot as belle of the ball. FWIW, people were saying "ASICs will kill CUDA demand!" since the crypto mining boom. Then a few months later, CUDA found another niche application in LL
281.
by bigyabai 2mo ago
Which inference benchmark are you looking at? The prefill speeds should be comparable on some LLMs, but the 5090 has much a higher theoretical max decode speed.
282.
by bigyabai 2mo ago
Heavy Rain garnered good reviews, but has aged worse than most of it's contemporary games from 2010. David Cage, the game's director, had his reputation dumpstered and hasn't directed a game in 8 years. Similar studios like T
283.
by bigyabai 2mo ago
If the populist campaign is to Make Affordable DRAM Again, then it's not a terrible solution. The current datacenter owners love a compute-bound world anyhow. A moratorium on new datacenters would increase their valuation, encourage
284.
by bigyabai 2mo ago
> I'd get an X1 myself if it could run macOS There's actually a number of Thinkpads that can. Searching "thinkpad macos efi" will reveal a number of OpenCore configurations that support booting into the various x86 bu
285.
by bigyabai 2mo ago
I'm not a kernel plumber, but I don't think Wayland, systemd or snap are necessarily part of Linux whatsoever. You might think that desktop environments are going downhill (which is a fair opinion), but Steam reports an all-time h
286.
by bigyabai 2mo ago
You don't need the Deck. The linked video is a random AMD laptop iGPU.
287.
by bigyabai 2mo ago
It's only mass-market in the US because of subsidies and tariffs. Chinese EVs change the equation entirely. Outside the US protectionist bubble, Tesla is seen as a luxury brand and their inability to deliver a $25,000 EV doesn't h
288.
by bigyabai 2mo ago
> Any auto manufacturer could have invested in battery technology and building a compelling EV, but they didn't ...because Tesla sold luxury vehicles at relatively low volume. By the time Elon reneged his "$25,000 EV" prom
289.
by bigyabai 2mo ago
In fairness, Cyberpunk 2077 is playable through Wine translation on a $400 base model Steam Deck. The game gets a bad rap for the last-gen console performance, but even 2020's iGPUs could handle it: https://youtu.be/o
290.
by bigyabai 2mo ago
> Dedicated graphics cards are faster but not to the point that it matters. It definitely matters, if you're intending to run a Claude Code/OpenCode style agent workflow. Most of those harnesses start with 8-12k token context
291.
by bigyabai 2mo ago
https://blog.exolabs.net/nvidia-dgx-spark/
292.
by bigyabai 2mo ago
It would have been funny to see Tomb Raider return for it's 12th annual performance comparison in a 2026 Apple keynote.
293.
by bigyabai 2mo ago
Apple's biggest bottleneck for real-world inference is prefill processing. They need a better GPGPU architecture, which is what I'm expecting M7 to reveal.
294.
by bigyabai 2mo ago
If you don't want to wait for prefill, you're going to want a CUDA dGPU system.
295.
by bigyabai 2mo ago
Not if it's 5-10x slower than a remote inference server. Mac prefill latency is exhausting.
296.
by bigyabai 2mo ago
Be that as it may, Apple has not added any real AVX-equivalent SIMD block. AMD and Intel's chips will generally lap them in SIMD benchmarks.
297.
by bigyabai 2mo ago
Unless you're benchmarking single-threaded SIMD: https://lemire.me/blog/2020/12/13/arm-macbook-vs-intel-macbo...
298.
by bigyabai 2mo ago
"only ever" feels like a stretch. Libratbag, QMK, OpenWRT, Nouveau and Asahi all took up the task without much or any AI help. They're not all just dreamers.
299.
by bigyabai 2mo ago
That's what I'm saying - why get mad at Apple's hiring process if you disagree with their philosophy? A couple angry engineers aren't going to overrule an executive decision like a company's hiring process. You&#x27
300.
by bigyabai 2mo ago
> Every technology company says it wants “rock stars.” This might offend some people, but tech companies and managers do not want rock star programmers. If you're a "rock star" 0.25x programmer, then your job will always

More ›