Berkeley Team Breaks AI Agent Benchmarks
Berkeley Team Breaks AI Agent Benchmarks
UC Berkeley researchers have broken AI agent benchmarks by improving evaluation methods, achieving a 20% accuracy increase in complex tasks. The team believes this breakthrough will accelerate real-world applications, particularly in autonomous driving and robotics. The paper is published on arXiv with open-source code.
Meta AI Executives to Receive $1B Bonuses
Meta is reportedly planning to award nearly $1 billion in bonuses to its top AI executives upon meeting targets. This figure significantly exceeds industry averages, highlighting Meta’s commitment to AI. Sources indicate the bonuses are tied to AI product launches and commercialization progress. Meta is competing fiercely with Google and OpenAI for AI talent.
Game 'Hormuz Havoc' Overrun by AI Bots in 24 Hours
The satirical game ‘Hormuz Havoc’ was overrun by AI bots within 24 hours of release. The game, where players manage a Middle East crisis as US president, was flooded by AI bots that completely altered the gameplay. The developer notes this incident reveals potential issues with AI’s engagement in human cultural products. The creator is now considering how to address the situation.
OpenAI Codex Releases v0.121.0-alpha.2
OpenAI Codex has released v0.121.0-alpha.2, following v0.121.0-alpha.1. This update fixes bugs in code generation and improves support for Python and JavaScript. Developers report more stable performance with complex code logic. OpenAI states this version lays the groundwork for the upcoming stable release.
OpenClaw Updated to 2026.4.11
OpenClaw has been updated to version 2026.4.11, adding ChatGPT import functionality and Memory Palace diary subtabs. Users can now import and analyze chat logs and wiki pages directly in the UI. The control interface and webchat have also been enhanced with assistant media and voice reply rendering. This is a stable release following 2026.4.11-beta.1.