Loading activity…
@ml_angelopoulos
Co-Founder and CEO of @arena PhD @Berkeley_EECS EE @StanfordEng SR @GoogleDeepMind
AI
Agent Arena reports that 86% to 96% of agents' tokens are context they have already seen.
AI
Arena says Amazon scientist Ido Rosen discusses evaluation pitfalls and how open-weight models could affect frontier-lab pricing.
YouTubeAI
AI:AM's YouTube description bills a segment comparing live human preferences and agent evaluations with static benchmarks.
AI
A post says Carbon scored 88.2% on SEC-bench Pro and is rolling out first to cybersecurity specialists.
AI
Arena says users can now connect GitHub, Gmail and Google Calendar within a single Agent Mode task.
XAI
A user says investor lists routinely miss brilliant investors with better track records, then jokes about the name Sarah.
AI
Kimi claims a 1-million-token context window for Pro and two to three times Pro's speed for Highspeed.
XAI
Agent Arena says GPT-6 Luna takes three of four optimal spots below $0.10 per task at three reasoning levels.
AI
The poster says seed leads E14 Fund and Obvious Ventures are continuing their support.
XAI
Arena places the model 11th overall in Agent Arena, with a 5.1% net improvement across 7,500-plus real-world sessions.
AI
Arena says the index is built from 90K-plus real-world agent sessions across 30 models.
Technology
Meta says its 816B-parameter model can run locally across two M5 Ultra Mac Studios.
AI
finkd says the model runs well on new Mac Studios, crediting work with Apple and John Ternus.
AI
The incoming professor says they're recruiting PhD students and postdocs for fall 2027 in reinforcement learning and deep learning.
AI
Arena says its index tracks unauthorized actions, false attribution and deceptive completion across 90,000-plus agent sessions and 27 models.
YouTubeAI
Agent Arena says Jev Router had similar performance to DeepSeek V4.1 Flash (Max) but cost 38% more.
AI
An essay argues that falling prices and surging token use favor platforms that control distribution over benchmark-leading labs.
AI
A user argues that cost per token should no longer be used as a metric.
AI
Google announces a 1 million-token output limit, up from 64,000, for its new model. Initial access goes to trusted cyber defenders through the Fairwind Program.
