samus

11 ms·

301.
by samus 11mo ago
The consumer GPU market is not treated as a primary market by GPU makers anymore. Similar to how Micron went B2B-only.
302.
by samus 11mo ago
I just gave Qwen3-Next-Instruct a system prompt that instructs it to always answer with a Prolog program that computes the answer. When I asked whether the Collatz sequence terminates, it indeed answered with a Prolog program that computes
303.
by samus 11mo ago
Or expect the LLM to only output Prolog, with the assumptions and knowledge as clauses?
304.
by samus 11mo ago
If the generated code uses a suitable programming language, like the safe subset of Haskell, then the risk is significantly lower. Anyway it makes sense to execute this code in the user's browser instead of on the server.
305.
by samus 11mo ago
Yes, it is. But it's rather complicated and not all applications need its full power. https://lwn.net/Articles/810414/
306.
by samus 11mo ago
Yet for many syscalls there is an official library - in most cases a wrapper in libc, but especially io_uring is known to provide a C library that most applications ought to use instead of the raw syscalls.
307.
by samus 11mo ago
There aren't any because it depends a lot on what your use case is, what speed you expect, how accurate you want it to run, how many users want to use it, and how much context size you need. - If you have enough system RAM then your VR
308.
by samus 11mo ago
I also had the impression from the report that shareholders were miffed about this Q3 snag, so they had to publish this even though they were about to treat this as business as usual.
309.
by samus 11mo ago
Rust is by no means allowed in the core yet, only as drivers. So far, there are only a few drivers. Currently, only the Nova driver, Google's Binder IPC and the (out of tree) Apple drivers are of practical relevance.
310.
by samus 11mo ago
That's indeed reasonable, but the opposite can happen just as well: there is a vulnerability in the western calendar, but I'm on an old major.minor version that receives no security patches anymore. So now I have to upgrade that d
311.
by samus 11mo ago
These people are not customers though. The maintainers do their best, but overall the project seems to be understaffed though, so customers (for example Google, as it seems they occasionally chip in) get priority.
312.
by samus 11mo ago
> Also in general Google does investigate software they don't make money off. An organization of this size might actually have trouble making sure they really don't use code from that project. Or won't do so in the future.
313.
by samus 11mo ago
That's a one-time bonus they get for discovering a bug, not from using the project on production. Google also gets this reward by the way. Therefore it's still imbalanced.
314.
by samus 11mo ago
They are different because the independent researchers don't make money off the projects that they investigate.
315.
by samus 11mo ago
Why should Google not be expected to also contribute fixes to a core dependency of their browser, or to help funding the developers? Just publishing bug reports by themselves does not make open source projects secure!
316.
by samus 11mo ago
Sorry to put it this bluntly, but you are not going to get what you want unless you do it yourself or you can convince, pay, browbeat, or threaten somebody to provide it for you.
317.
by samus 11mo ago
So what is Google gonna do if security fixes don't happen in time and the project takes a "reputational hit"? Fork it and maintain it themselves? Why not send in patches instead? Maintaining a reputation might be enough rewar
318.
by samus 1y ago
> No it's not. There are plenty of things that can't be 100% reliable no matter how well they're made. A perfect bridge is still going to break down and eventually fall apart. The best possible motion-activated light is go
319.
by samus 1y ago
That's a known limitation of floating point numbers. Nothing buggy about that.
320.
by samus 1y ago
> it's complex to generate correct machine code, but we trust compilers to do it all the time. Generating correct machine code is actually pretty simple. It gets complicated if you want efficient machine code. > So, then - why
321.
by samus 1y ago
> He doesn't point to a single thing that pg_vector doesn't do or doesn't do well. He just complains it hard to do. He very clearly complains that IVFFlat indexes have to be periodically rebuilt, that HNSW has high overhe
322.
by samus 1y ago
That is a workaround and precisely the point the author makes. It increases operational complexity and creates a divide between records in the vector DB and the relational DB.
323.
by samus 1y ago
In that case you might not even really need optimized vector search though.
324.
by samus 1y ago
HTML5 is unlike XML in very important regards, which makes it actually quite difficult for tools to handle. But easier than XML. XHTML was quite annoying to write as far as I remember.
325.
by samus 1y ago
With a Qwen3-32B-A3B (Q8) I'm getting 10-20 t/sec on KoboldAI, e.g., llama cpp. Faster than I can read, so good enough for hobby use. I expect this model to be significantly faster, but llama.cpp-based software probably doesn&#x27
326.
by samus 1y ago
This is a Mixture of Experts model with only 3B activated parameters. But I agree that for the intended usage scenario VRAM for the KV cache is the real limitation.
327.
by samus 1y ago
We very much can, especially such a Mixture of Experts model with only 3B activated parameters. With an RTX 3070 (7GB GRAB VRAM), 32 GB RAM and an SSD I can run such models at speeds tolerable for casual use.
328.
by samus 1y ago
> Some do, like windows can be replaced with gnome, but mobile phones cannot. > Probably you are not typing or reading this from a European os on your phone, yet alternatives exist. Both Linux and Android are OSS, though admittedly t
329.
by samus 1y ago
Why would you think the Austrian government is not doing its own IT development? There already is a company specifically for that: the Bundesrechenzentrum.
330.
by samus 1y ago
Output really doesn't have to be the same datatypes as the input. Text tokens are good enough for a lot of interesting applications, and transforming percels (name suggested by another commenter here) into text tokens is exactly what a

More ›