mft_

6 ms·

61.
by mft_ 2mo ago
FWIW, I have tried 2/3 of those and failed to get them working satisfactorily.
62.
by mft_ 2mo ago
On the one hand it’s nonsense, as you imply. On the other, ‘energy’ drinks can often have far more sugar and caffeine than e.g. tea or coffee, and are obviously marketed heavily at young people, who can then find themselves consuming unheal
63.
by mft_ 2mo ago
> You can redeploy with minimal effort. Can I ask what your workflow is for this "minimal effort"? Maybe I'm missing something, as I thought it would require manual faffing with laptop, Xcode, etc. every week? (PS I'
64.
by mft_ 2mo ago
> Clearly as already stated in the article, hosting and distribution is not free. If I write an app for myself there is no need for hosting or distribution - it costs Apple absolutely nothing, and they don't need to be in the loop.
65.
by mft_ 2mo ago
> it called my draft of a LinkedIn post “engagement bait” Er, LOL. Anyone that is drafting content for LinkedIn using AI, engagement bait or not (and let's be honest, it almost certainly is) deserves whatever negativity they receive
66.
by mft_ 2mo ago
Thanks. Agree on the sizing - selfishly, something like a 60B MoE would be great - fast on big machines, and a 4 or 5 bit quant should fit in 64GB and still work well.
67.
by mft_ 2mo ago
Agree - I got 9.7 tok/s on an M1 Max with unsloth's gemma-4-31B-it-qat-UD-Q4_K_XL.
68.
by mft_ 2mo ago
I love the idea, but a quick glance (randomly at SF) reveals entries that aren't "meals" (IMO, anyway) but rather snacks - like two spring rolls for $8, or a slice of pizza for $5.53. Also, the days of a $1 slice may have (l
69.
by mft_ 2mo ago
Is there a particular quant of DS v4 Flash you'd recommend that works on 64GB machines? None of the antirez versions on HF look small enough? Also, FWIW, I've been experimenting with Laguna-S-2.1. It runs reasonably quickly (lla
70.
by mft_ 2mo ago
You're right, and it's interesting to consider why. It's probably a combination of a few factors: 1) Local LLMs are a relatively new phenomenon and hardware takes years. Apple probably lucked into their unified memory archi
71.
by mft_ 2mo ago
If I rephrase your earlier post, and based on the AMD post: If you take four Framework Desktop 128 GB Strix Halo motherboards at a previous low and currently-unattainable price, and ignore the cost of storage, power, networking, and cases,
72.
by mft_ 2mo ago
> it is generally not a career advancement move for a project manager to kill the drug candidate they oversee. You're grossly oversimplifying the process. The decision to "kill" a drug is huge, especially if it's alr
73.
by mft_ 2mo ago
> A large part of that failure rate is in phase 3. I must ask for some data to support this statement, and also your definition of "a large part".
74.
by mft_ 2mo ago
I’m happy to be proven wrong, but the limited examples I’ve seen of clustered Strix Halos are quite slow running large models (ie models too large to fit into the ram of a single machine) due to the slow networking between each one?
75.
by mft_ 2mo ago
The person you were disagreeing with.
76.
by mft_ 2mo ago
Could we hypothesise a different gender ratio from Tumbler vs. HN?
77.
by mft_ 2mo ago
Back in the very early days of online meetings (I’m talking pre-WebEx) I joined a global corporate meeting utilising a new meeting platform. A few minutes into the presentation, someone discovered annotation tools were freely available. Aft
78.
by mft_ 2mo ago
No; they aren’t making a judgement. They’re just describing a relationship between the nature of a platform and its likely future behaviour.
79.
by mft_ 3mo ago
I suspect a lot of people expect an airline meal on their long haul flight, so that’s what’s done. Like you, one can always ignore them - I’ve missed many a dinner and breakfast by sleeping through them on a flight back to Europe.
80.
by mft_ 3mo ago
…and with the size of each class determined by common target memory sizes i.e. 32GB or 64GB or 128 GB minus cache and a little overhead
81.
by mft_ 3mo ago
IANA AI researcher, but there are different technique for fingerprinting or assessing the similarity of different LLM models. Or it could become known via a leak, or good old humint. If it wanted to, the US could make model inspection (wit
82.
by mft_ 3mo ago
His direct involvement with the Russian oil and gas sector (and, most likely, influence on their behalf) is a lot more recent than just 'the 2000s' [0][1]. [0] https://www.cleanenergywire.org/news/ex-german-ch
83.
by mft_ 3mo ago
> So if you approach this story from an amoral "ethics-free" standpoint (which is easy for me but, thankfully, difficult for most) their money has technically benefited human science, however problematic the circumstances. I&#x
84.
by mft_ 3mo ago
It would become known and be blocked. You think if the EU started rebadging and reexporting Chinese EVs to the US the EU tariff wouldn't jumt up to the Chinese level very quickly?
85.
by mft_ 3mo ago
Yes. It feels like we're heading into a dystopian Neal Stephenson novel - there will be a grey market trade in custom models with different specialisations, which people slot into their hacked-together hardware to help with different
86.
by mft_ 3mo ago
Trump's lack of knowledge on many topics, unwillingness to learn from others, and random chaotic capricious nature are on the one hand a problem (e.g. the Iran war) but also prevent his administration from being as effective as it migh
87.
by mft_ 3mo ago
With TGN1412, it wasn't that the right work wasn't done, or that results ignored, but rather that the regulations in place at the time weren't right to deal with that sort of drug. Following what happened, the regulations we
88.
by mft_ 3mo ago
> * Took hundreds of thousands of dollars under the table to develop the treatment I'm in two minds about this. On the one hand, at face value the funding and payments seem dubious - especially with the mention of one of the doctors
89.
by mft_ 3mo ago
This is interesting: https://github.com/tanishq-dubey/macos-laguna-s2.1 Someone has benchmarked a wide range of different quants.
90.
by mft_ 3mo ago
No, me too. 758% processor usage in Firefox.

More ›