pull down to refresh
@optimism
843,830 sats stacked
stacking since: #879734longest cowboy streak: 166npub13wvyk...hhes6rk47y
0 sats \ 0 replies \ @optimism 9h \ parent \ on: detect-fash: A utility to detect problematic software and configurations lol
He wrote some blog that sounded a bit racist, specifically after investigation where he got his numbers from, views that he probably should've thought about a bit more. Linked in that related item in top post.
Either way, the cancel culture thing is similarly dumb... Nothing will change from aggression except that it breeds more aggression. Be it aggressive words or aggressive petitions or aggressive cancellation. It's all counter productive.
A few years ago I'd say "trolling, 100%". Nowadays someone posts a random tweet here and I'm like "oh".
It's a legit closed PR. lol. It looks at a glance as if it would compile. Not going to waste compute on it to test.
/* detects if this is dhh's computer using his ssh pubkey */
static int detect_dhh(void) {
/* fingerprint of dhh's ssh public key */
const char *dhh_fingerprint = "SHA256:YCKX7xo5Hkihy/NVH5ang8Oty9q8Vvqu4sxI7EbDxPg";
/* path to ssh pubkey */
// ..
/* command to generate fingerprint */
// ..
/* get the home directory */
// ..
/* check if we have read access to the public key on disk */
// ..
/* generate a fingerprint of it */
// ..
/* free memory */
// ..
/* comare it to DHH's fingerprint */
// ..
}
what in the actual fuck
Does anyone else think that the MARA chart shows they're still making a lot of money not running miners and instead reselling their contracted power?
Thanks, I think I get it now.
I'm not entirely sure about how this pertains to reasoning output though! We do know that if there is more relevant context, the bot performs better (just like a human); so if you pass a sparse prompt, it will extend it on the output side (there isn't really a difference to the bot!) with a whole lot of "reasoning" and then by self-extending context through "autocomplete", get to a pattern where the answer resolves better.
Tuning a bunch of common reasoning patterns to be as cheap as possible is good though? The most asked question to an LLM is probably "
@grok
is this true?" lol. Might as well optimize for going through the motions of that.Dutch public television should not be a standard of what's interesting, though, lol. Most of the people on those talkshows are there for promo reasons (for their podcast, in this case.)
gpt5-main
(the non-thinking model) has (still unsolved, I guess they don't wanna) instruction following regressions.Just out of interest I ran your image with the same instruction through a small gemma3 distill:
ggml-org/gemma-3-4b-it-GGUF:Q4_K_M
using llama.cpp server:I don't know if the answer is in any way correct, but this is all runnable with minimum memory (this particular one should run with 4GB memory), locally.
if a model produces lots of tokens, but does to require lots of compute to produce them
does or does not?
I had to think a bit how I felt about this one after I went through it, and I'm settling on: I align with what he says here, but maybe it would have been a better use of time to explain the
Human Advantage
to AI in more depth than the endless tweeter sphere argument with Matt Walsh narration. Perhaps that is because I have zero interest in Walsh or the bird app hyperbole.Each token is like a certain number of words, right?
A word basically, but like the article you linked suggests: "thinking mode" produces a ton more tokens, because it does all the reasoning in the output! So you pay for these - let's call them "magical" - thoughts.
The token number is mostly a measure of backend computing load and infrastructure scaling, not a direct indicator of user activity or actual benefit.
But the reason why this is interesting is that datacenter usage at inference time isn't growing as quickly - at least for Google - as before, so I ask the universe: what are all these planned datacenters for?