Tiberium

20 ms·

421.
by Tiberium 2y ago
Not really, repo map only gives LLMs an overview of the codebase, but aider doesn't automatically bring files into the context - you have to explicitly add the files you wish for it to see in their entirety to the context. Claude Code&
422.
by Tiberium 2y ago
Because chatgpt-4o-latest already has all of those improvements, the largest point of this release (IMO) is to offer developers a stable snapshot of something that compares to modern 4o latest. Altman said that they'd offer a stable sn
423.
by Tiberium 2y ago
Yes, and they don't make snapshots for chatgpt-4o-latest, but they made them for GPT 4.1, that's why 4.1 is only useful for API, since their ChatGPT product already has the better model.
424.
by Tiberium 2y ago
There's a HUGE difference that you are not mentioning: there are "gpt-4o" and "chatgpt-4o-latest" on the API. The former is the stable version (there are a few snapshot but the newest snapshot has been there for a w
425.
by Tiberium 2y ago
Very important note: >Note that GPT‑4.1 will only be available via the API. In ChatGPT, many of the improvements in instruction following, coding, and intelligence have been gradually incorporated into the latest version If anyone here d
426.
by Tiberium 2y ago
It's sad that they used 4o's image generation feature for the cat example which does some diffusion or something else, results in the whole image changing. They should've instead used Gemini 2.0 Flash's image generation
427.
by Tiberium 2y ago
Gemini 2.5 Pro was only released on 25th of March ;)
428.
by Tiberium 2y ago
I think o1 pro runs multiple instances of o1 in parallel and selects the best answer, or something of the sort. And you do actually always pay for thinking models with all providers, OpenAI included. It's especially interesting if you
429.
by Tiberium 2y ago
Even the basic interface is different, actually - "input" vs "messages", no "max_completion_tokens" nor "max_tokens". That said, changing those things is quite easy.
430.
by Tiberium 2y ago
"Max" as in "Claude 3.7 Sonnet MAX" is apparently Cursor-specific marketing - by default they don't use all the context of the model and set the thinking budget to a lower value than the maximum allowed. So essentia
431.
by Tiberium 2y ago
There actually were almost no benchmarks for o1 pro before because it wasn't on the API. o1 pro is a different model from o1 (yes, even o1 with high reasoning).
432.
by Tiberium 2y ago
Not sure what you mean? Those are the exact watches Repebble is selling https://store.repebble.com/
433.
by Tiberium 2y ago
You're right about them not wanting to give away requests for free, but in fact they don't rely on OpenAI directly - thanks to the agreement Microsoft has with OpenAI they get model weights for all OpenAI models before OpenAI &quo
434.
Xerox scanners/photocopiers randomly alter numbers in scanned documents (2013)(dkriesel.com)
1 points by Tiberium 2y ago|0 comments
435.
by Tiberium 2y ago
They already removed that and one of xAI's developers clarified in a lot of posts that this change was published by someone without going through all of their process. https://x.com/ibab/status/189377401737648
436.
by Tiberium 2y ago
H800 is the export variant that they had access to. They directly reference it in the repo: >Achieving up to 3000 GB/s in memory-bound configuration and 580 TFLOPS in computation-bound configuration on H800 SXM5, using CUDA 12.6.
437.
by Tiberium 2y ago
Better URL: https://novasky-ai.github.io/posts/sky-t1/
438.
by Tiberium 2y ago
The extremely interesting part is that 3.5 Sonnet is above o1 on this benchmark, which again shows that 3.5 Sonnet is a very special model that's best for real world tasks and not some one-shot scripts or math. And the weirdest part is
439.
by Tiberium 2y ago
That's actually exactly what the author of NES86 did at first: he ported Lunix to the NES - https://www.youtube.com/watch?v=SVqN_FixG0M (repo https://github.com/decrazyo/lng-fds ). But then he rece
440.
NES86 – IBM PC Emulator for the NES(github.com)
129 points by Tiberium 2y ago|27 comments
441.
Linux Running on an NES, but for Real This Time [video](youtube.com)
3 points by Tiberium 2y ago|0 comments
442.
by Tiberium 2y ago
*Electron, you almost got it :)
443.
by Tiberium 2y ago
It's not that big of a news because they already had gemini-exp-1206 on the API - they just didn't say it was Gemini 2.0 Pro until today. Now the AI Studio marks it as 2.0 Pro Experimental - basically an older snapshot, the newe
444.
by Tiberium 2y ago
* from the recent commit "V1.07 works on Win95. But you have to install a different LibCrypto, LibSSL, and also install the WinSock2 update." Download for the Win95 version: https://github.com/DiscordMessenger&#x2F
445.
Discord client that works on Win95*, Win98 and above(github.com)
253 points by Tiberium 2y ago|195 comments
446.
by Tiberium 2y ago
Are you playing arcade or realistic battles? There's a huge difference between the two. And there are also simulator battles.
447.
by Tiberium 2y ago
Useless SEO garbage: it's just an iframe for https://huggingface.co/spaces/deepseek-ai/Janus-Pro-7B + lots of AI-generated text based on the official docs
448.
by Tiberium 2y ago
Would you say the same for OpenAI releasing new models?
449.
by Tiberium 2y ago
Finetunes of smaller models on R1 output are pretty bad, but the main R1 model is very, very far from being "garbage".
450.
by Tiberium 2y ago
They trained their own V3 and then trained R1 Zero from V3 purely as RL, it didn't completely work so then they took some CoT examples and trained R1 on RL + some SFT. You're thinking of the finetunes based on R1 outputs - those a

More ›