agnos

Openrouter UI updated

@HarshithLucky3 This dmhss called me lazyss ;_;

ASI is here It can control the universe 🤯🤯

@gill_kyle I hate the codex UI from my core

The new cursor is.... Codex? 🤨

@DandumontP Bruh I asked grok did hakintosh got any new update after Apple arm transition Grok: it’s April fools joke

J’ai récupéré un MCU de Tesla Model 3 et j’y ai installé Mac OS X. Parce que pourquoi pas. Le journal du lapIA journaldulapin.com/2026/04/01/11-…

@RoundtableSpace Wait a 0.9B OCR model is beating Vision Transformers at OCR?? that's actually insane

THIS TINY 0.9B MODEL CALLED GLM OCR BEATS GEMINI AT DOCUMENT READING AND IT’S OPEN SOURCE

Claude code source code has been leaked via a map file in their npm registry! Code: …a8527898604c1bbb12468b1581d95e.r2.dev/src.zip

I forgot my agent is running bg and messaged in telegram after many failed attempts

🪄 Introducing MagiAttention v1.1.0: Defining the Performance Ceiling for Hopper & Blackwell Architectures. MagiAttention is a distributed attention engine sharing the bottlenecks of ultra-long context and heterogeneous masks. - Repo: github.com/SandAI-org/Mag… - Blog: sandai-org.github.io/MagiAttention/…

The TurboQuant paper (ICLR 2026) contains serious issues in how it describes RaBitQ, including incorrect technical claims and misleading theory/experiment comparisons. We flagged these issues to the authors before submission. They acknowledged them, but chose not to fix them. The paper was later accepted and widely promoted by Google, reaching tens of millions of views. We’re speaking up now because once a misleading narrative spreads, it becomes much harder to correct. We’ve written a public comment on openreview (openreview.net/forum?id=tO3AS…). We would greatly appreciate your attention and help in sharing it.

@loktar00 @zai This is correct: @Zai_org 😭😭

I've started really using my @zai max coding plan recently with hermes... 45 million tokens, most within the last 3-4 days and not even 1% of my monthly usage.. people are really sleeping on GLM models.

@amritwt Confirmation:

@amritwt speculation: Cursor used kimi's model to beat opus in their benchmark

CURSOR MADE A MODEL BETTER AND CHEAPER THAN OPUS

@askalphaxiv Didn’t he come from Meta? This should have started before his tenure.

Yann LeCun and his team dropped yet another paper! "V-JEPA 2.1: Unlocking Dense Features in Video Self-Supervised Learning" In this V-JEPA upgrade, they showed that if you make a video model predict every patch, not just the masked ones AND at multiple layers, they are able to turn vague scene understanding into dense + temporal stable features that actually understands "what is where". This key insight drove improvements in segmentation, depth, anticipation, and even robot planning.

Here is SeedProteo, our latest diffusion-based model for de novo all-atom protein design from ByteDance Seed! Our server is now live — feel free to give it a try! seedfold.io/dashboard/prot…

This guy still alive? Didn’t antigravity killed him?

@julesagent This guy still alive? Didn’t antigravity killed him?

Gemini 3.1 Pro is now available in Jules. What a great time to merge some PRs.

Holy shit... @AntLingAGI just open-sourced a trillion-parameter AI model. It's called Ring-2.5-1T and it runs on consumer-grade GPUs at half the cost of comparable models. No lab access. No waitlist. No $10M compute cluster.

Why no one is talking about Gemini 3.1 flash? Even Google’s benchmark

Future is clearly ascii Humans don’t just write text!

@arcprize @hive_echo @OpenAI Gemini 3.1 pro preview just aura farming

GPT-5.4 and GPT-5.4 Pro from @OpenAI on ARC-AGI Semi Private ARC-AGI-2: - GPT-5.4: 74.0%, $1.52/task - GPT-5.4 Pro: 83.3%, $16.41/task