ətaqa

21 posts

ətaqa

ətaqa

@taqa286109

Katılım Mart 2025
1 Takip Edilen0 Takipçiler
ətaqa retweetledi
Technopolitik
Technopolitik@Technopolitik_·
SubQ is going after a real weakness in frontier AI: long-context systems that get expensive and unreliable as context grows. With 12M-token context and a lower cost curve, the implications reach industry, enterprise systems, and defense applications.
Alexander Whedon@alex_whedon

Introducing SubQ - a major breakthrough in LLM intelligence. It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA), And the first frontier model with a 12 million token context window which is: - 52x faster than FlashAttention at 1MM tokens - Less than 5% the cost of Opus Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention). Only a small fraction actually matter. @subquadratic finds and focuses only on the ones that do. That's nearly 1,000x less compute and a new way for LLMs to scale.

English
3
14
37
15K
ətaqa retweetledi
Documenting AGI
Documenting AGI@DocumentingAGI·
Anthropic reportedly deemed Mythos too risky to release, yet a small private Discord group allegedly got access on day one by guessing the URL from naming conventions exposed in the Mercor breach and using a contractor’s eval credentials.
Documenting AGI tweet media
English
16
35
554
9.3K
ətaqa retweetledi
Documenting AGI
Documenting AGI@DocumentingAGI·
OpenAI has launched GPT-5.5, calling it a new class of AI for work and agents. It’s now available in ChatGPT and Codex.
English
5
31
71
20.2K
ətaqa retweetledi
Byte
Byte@ByteEcosystem·
NEW: 🇺🇸 US silver mining stocks are trading at low levels relative to silver, despite many cos. reporting record FCF.
Byte tweet media
English
9
28
349
18.6K
ətaqa
ətaqa@taqa286109·
Childhood nostalgia hits different
ətaqa tweet media
English
0
0
0
1

Keşfet

@pokearenagame @resilience24 @Sebitax86 @CBDdamon @LeadingTheBull @vialaseric @0tnlmrV5uPfuzZY @mguzi12