Tiberium

20 ms·

391.
by Tiberium 1y ago
Yes, I tested the wrong version on accident :(
392.
by Tiberium 1y ago
del
393.
About 30% of Humanity's Last Exam chemistry/biology answers are likely wrong(futurehouse.org)
5 points by Tiberium 1y ago|2 comments
394.
by Tiberium 1y ago
That's true, but at least Steam did not hide NSFW games from search/recommendations (if you had mature content enabled), which Itch.io did without warning. And I assume that NSFW games likely represent a far higher % of revenue&#x
395.
by Tiberium 1y ago
Imgur really hates VPNs, and will show that bogus error page if you're not using a "nice" connection.
396.
by Tiberium 1y ago
I'm really not following -- why does the ban specifically focus on a single domain instead of attempting to solve the core issue? Do the maintainers not know that accounts for any big email provider (gmail, outlook, you name it) can be
397.
Groq is now serving Kimi K2 at 185 tok/s(twitter.com)
4 points by Tiberium 1y ago|0 comments
398.
Mineclonia – open-source Minecraft clone based on Luanti (formerly Minetest)(codeberg.org)
5 points by Tiberium 1y ago|0 comments
399.
TikTok Sans(fonts.google.com)
3 points by Tiberium 1y ago|0 comments
400.
by Tiberium 1y ago
OpenAI o3's writing style is quite distinct :)
401.
by Tiberium 1y ago
Can you also compare the performance with https://github.com/huggingface/tokenizers/ ? Would be helpful, since the benchmark in the tiktoken readme seems to be very outdated.
402.
by Tiberium 1y ago
> as I'm definitely in the market for a simple compiled language that can offer Python some really fast functions So, Nim? https://github.com/yglukhov/nimpy
403.
by Tiberium 1y ago
Any reason to prefer the WSJ link over the official source at https://www.army.mil/article/286317/army_launches_detachment... ?
404.
by Tiberium 1y ago
What finish reason are you getting? Perhaps your code sets a low max_tokens, so the generation stops while the model is still thinking, without giving any actual output.
405.
by Tiberium 1y ago
It might be a bit confusing, but there's no "only thinking flash" - it's a single model, and you can turn off thinking if you set thinking budget to 0 in the API request. Previously 2.5 Flash Preview was much cheaper wit
406.
by Tiberium 1y ago
For the curious: "жопа" (which "JEPA" sounds like) means "ass" in Russian. Also V ("В") means "in" (although if we get into specifics, the casing would need to be "жопу" or "ж
407.
by Tiberium 1y ago
That's simply not true, it's not just "max thinking budget o3" just like o1-pro wasn't "max thinking budget o1". The specifics are unknown, but they might be doing multiple model generations and then someh
408.
by Tiberium 1y ago
I've heard lots of people say that, but no objective reproducible benchmarks confirm such a thing happening often. Could this simply be a case of novelty/excitement for a new model fading away as you learn more about its shortcomi
409.
by Tiberium 1y ago
Regarding "we're now creating Portuguese and Russian language versions of the songs" - have you contacted any native speakers of those languages to listen to those songs? Since the translation part isn't that hard, makin
410.
by Tiberium 1y ago
Regarding 4 - in the full version shared by the author at https://archive.md/A95Ju there is info on what they meant by "terminal-inspired"
411.
by Tiberium 1y ago
To be fair, they do explicitly say "build a minimal, terminal-inspired email client" and then "Email client can either be in the terminal (i.e. a TUI) or a web app" below, so even if it's a web app, they expected it
412.
by Tiberium 1y ago
I understand that their responses were very minimal and not helpful, but it clearly says in the requirements to make a "terminal-inspired" email client. However, in the video you shared [1] I see a somewhat generic email web app,
413.
by Tiberium 1y ago
You're right, that's why I'm questioning the reason Anubis implemented it this way. Lots of big AI companies are at least honest about their crawlers and have proper user agents (which Anubis outright blocks). So "unethi
414.
by Tiberium 1y ago
From looking at some of the rules like https://github.com/TecharoHQ/anubis/blob/main/data/bots/head... it seems that Anubis explicitly punishes bots that are "honest" about their user
415.
by Tiberium 1y ago
It's worth noting that AI Studio is the API, it's the same as OpenAI's Playground for example.
416.
by Tiberium 1y ago
Imagen supports image references in the API as well, just on Vertex, not on Gemini API yet.
417.
by Tiberium 2y ago
That's the web version (which has tools like search plugged in), other models in their official frontends (Gemini on gemini.google.com, GPT/o models on chatgpt.com) are also "real time". But when served over API, most of
418.
by Tiberium 2y ago
Gemini 2.5 Pro has the same knowledge cutoff specified, but in reality on more niche topics it's still limited to ~middle of 2024.
419.
by Tiberium 2y ago
The thread is about 2.5 Flash though, not 2.5 Pro. Maybe you can try again with 2.5 Flash specifically? Even though it's a small model.
420.
by Tiberium 2y ago
del

More ›