Sabitlenmiş Tweet
Rookie
50 posts
@NicholasDClay @theo the thing is that no AI lab has officially announced that they have made a model having 5 trillion or more than 5 trillion parameters
English
@Rokieee__ @theo There is already a 10 trillion parameter model and Grok doesn't even have a 5 trillion parameter want yet, what are you on about?
English
I made SpaceTimeLoom.com to explain gravity to my kids. They can just play with mass and see how it bends space.
You don’t have to explain concepts anymore. You can build them!
English
@thsottiaux but still why though, will it automatically start to talk about them if not specifically instructed @thsott
English
Hmm...
Arena.ai@arena
GPT-5.5 by @OpenAI is now live in the Arena, landing across multiple leaderboards. Here’s how it ranks by modality: - Code Arena (agentic web dev): #9, a strong +50pt jump over GPT-5.4 - Document Arena (analysis & long-content reasoning): #6, on par with Sonnet 4.6 - Text Arena: #7, Math #3, Instruction Following: #8 - Expert Arena: #5 - Search Arena: #2 - Vision Arena: #5 Strong, well-rounded performance, especially in Code (+50 pts vs GPT-5.4). Congrats to @OpenAI on the release. Full category breakdowns by modality in the thread.
QST
@sashaprzybylski Hey @sashaprzybylski, Can you please explain your approach, and what kinda data are you using
English
I'm 18 and building ML for battery materials discovery. Just made a model to predict ionic conductivity 22% better than the published benchmark!
English
Great article by @viktoroddy,
google stitch also opensourced a draft specification for DESIGN.md,
github.com/google-labs-co…
Viktor Oddy@viktoroddy
English
Its Just getting you info from you chat history, its memory and probably by searching about you from the web
Linus ✦ Ekenstam@LinusEkenstam
You must try this. GPT Image-2 can do PALM reading and I’m so here for it. Full prompt below ⤵️
English
@Rokieee__ @GregKamradt No public ARC-AGI-3 results for HRM yet. Worth noting: the official track (where frontier labs score <1%) uses different task distribution than the community leaderboard. Any HRM score outside the official track isn't directly comparable to GPT/Claude/Gemini numbers.
English
@bycloudai @bycloudai Can we replace the linear transformation inside each neuron with something richer, so the same number of parameters stores more information and achieves better accuracy?
I tried but couldn't get something without tradeoffs
huggingface.co/anshdadhich/ri…
English
be DeepSeek
>need to achieve batch invariance so bad
>split-k is the only optimal solution but is batch variant
>Thinking Machines ($50b valuation) could barely recover the performance for their solution, gets 1.6x slower
>hold_my_beer.jpg
> dual-kernel strategy
>"match or even surpass the perf of standard split-k in most major scenarios"
>DeepSeek strikes again
>$20b valuation btw
English
Can we replace the linear transformation inside each neuron with something richer, so the same number of parameters stores more information and achieves better accuracy?
bycloud@bycloudai
be DeepSeek >need to achieve batch invariance so bad >split-k is the only optimal solution but is batch variant >Thinking Machines ($50b valuation) could barely recover the performance for their solution, gets 1.6x slower >hold_my_beer.jpg > dual-kernel strategy >"match or even surpass the perf of standard split-k in most major scenarios" >DeepSeek strikes again >$20b valuation btw
English