Iolaum

3 ms·

1.
by Iolaum 5d ago
I wonder if that means that SpaceX evals show that they consider astra better than fable or that they hate Sam&co so much they don't want to show their stuff.
2.
by Iolaum 6d ago
Even after Astra rollout?
3.
by Iolaum 6d ago
There is difussion.cpp which is intended for those types of models. I set up krea-2-turbo with the help of ChatGPT 2 months ago, if you have a capable computer that's what I would suggest once it becomes supported.
4.
by Iolaum 8d ago
Things like that - and other examples posted here - are why I 'm sticking with OpenCode despite it having some papercuts that annoy me. The incentives are not there for them to do shady stuff like vacuum your files, inflate your token
5.
by Iolaum 8d ago
While I liked your comment enough to upvote, I then realized that Chairman isn't an operational position. It's more a position where the person can exercise ownership of a company (to put it in layman terms). In your analogy he wo
6.
by Iolaum 8d ago
Why mostly unusable 125b? I assume you are talking about qwen3.8-flash-next. Support for it on some places, like llama.cpp, is still wip (depending on configuration) but it looks like a very capable model in it's category.
7.
by Iolaum 10d ago
Some years ago Firefox was the go to browser if you wanted to have "some" privacy in your browsing (together with uBlock origin). With news like this I m really wondering if my views are outdated and I need to to some good researc
8.
by Iolaum 15d ago
lol! OP writes: PM finally said "too fast, go around" but PF ignored him (or likely did not hear)
9.
by Iolaum 15d ago
Some US labs may become uncompetitive if they have the wrong business strategy, that doesn't mean every AI company will be. If one part of the stack gets commoditized, build your moat elsewhere.
10.
by Iolaum 18d ago
Yes but what about taking an open weight Chinese model and do some extra pre-train and post-train to remove any potential hidden backdoors?
11.
by Iolaum 22d ago
$20 per month should get you far enough if you use luna, it's a decent model. If you plan to try astra you need to try make it use as little context as possible to protect your usage limit - or bank a reset mid task to allow it to keep
12.
by Iolaum 23d ago
Our new model is super-duper awesome, try it ... Ooops, while deploying it all the clouds are out ... :/
13.
by Iolaum 23d ago
Let's say you are a Graphene OS user. Why are you even using such an app? Also you are misrepresenting the threat model. The problem is not an app deleting it's own legitimate data. P.S. On GrapheneOS you can block apps from getti
14.
by Iolaum 23d ago
Zuck hinted at it n his twitter post but I doubt it.
15.
by Iolaum 24d ago
Graphene puts a HEAVY emphasis on security. Also your argument about a user inspecting and editing application files feels like a strawman argument. For example many spyware use malicious links to infect the devices, not malicious apps.
16.
by Iolaum 24d ago
Even in the EU spyware use is prevalent (and i 'd guess everywhere else in the world). There have been many scandals of government authorized commercial spyware been deployed against journalists. Is it really that niche a mobile OS tha
17.
by Iolaum 24d ago
subscribed to their newsletter ;)
18.
by Iolaum 24d ago
But not open weight I guess, right?
19.
by Iolaum 25d ago
Is there really a cost now that you can learn things from AI? And there's always slack et al for backup. Maybe it's "just" a matter of learning in a different way?
20.
by Iolaum 26d ago
I thing it will be the same like servers. Some of us have servers (raspberry pi's?) at home, but that's a very small part of the "server market".
21.
by Iolaum 29d ago
What? lettuce is not part of the traditional greek salad recipe.
22.
by Iolaum 1mo ago
I expected this sentence to end with assassination teams, lol. Talk about duall use technologies :p !!
23.
by Iolaum 1mo ago
They actually do when all their proprietary model customers are also building their own chips.
24.
by Iolaum 1mo ago
Do they really need HF for that? Dgx Spark and Strix Halo have very close specs and deliver similar performance. If nvidia makes their stack be more efficient with for example 50% more tockens on similar hardware specs agaist competitors, t
25.
by Iolaum 1mo ago
While nvidia 's drivers are closed source there's enough interest in running LLM's that you are not locked in using nvidia. Strix Halo chips have been around before dgx spark came out and deliver very similar performance.
26.
by Iolaum 1mo ago
While nvidia may not want open source drivers, they want open weight models and this acquisition allows them to keep pushing for that. Commoditize your complement.
27.
by Iolaum 1mo ago
Claiming AGI when they don't even have the best performing LLM out is a bit bold :p
28.
by Iolaum 1mo ago
indeed, can't wait for it to be supported by llama.cpp (or other engines)?
29.
by Iolaum 1mo ago
?! It does work but you need separate (sub)agents (and potentially loop it until the reviewer agent finds nothing). You also need to set it up a bit so all citations mentioned are easy to identify. Ideally this is also done programmaticall
30.
by Iolaum 1mo ago
I still think it's more realistic that we 'll have a nuclear war or society collapse because of climate change or populist politics than an AI going rogue. Of course one doesn't exclude the other and we may well get both ..

More ›