Sierra's new benchmark exposes what's really wrong with voice AI
Sierra's new μ-Bench benchmark uses real customer service calls to test voice AI where it matters most, introducing a metric that separates critical errors from harmless ones.
Developer
Most Recent 
GitHub's DMCA takedowns hit a record high in 2025
GitHub removed a record 47,228 projects in 2025 due to copyright claims, with anti-circumvention notices hitting an all-time high of 645.
AI's New Hacking Tools Are Too Fast for Open Source Maintainers
New AI models from Anthropic and OpenAI can find software vulnerabilities faster than open-source maintainers can fix them, creating a dangerous security gap.
Google expands its Chrome AI assistant to seven more countries
Google's Gemini AI assistant is now available in Chrome for users in Australia, Indonesia, Japan, the Philippines, Singapore, South Korea, and Vietnam, though the more advanced agentic features remain limited to US paid subscribers.
OpenAI's ChatGPT and Codex hit by hours-long outage, with a brief encore
OpenAI's ChatGPT, Codex, and API platform experienced a partial outage lasting over four hours, followed by a brief second hiccup for Codex.
Google's AI Studio now gives paying subscribers more room to experiment
Google AI Pro and Ultra subscribers will get higher usage limits and access to more models in Google AI Studio starting April 20th, bridging the gap between free tinkering and paid API development.
Google Forms 'Strike Team' to Close the AI Coding Gap with Anthropic
Google DeepMind has formed a dedicated team to catch up to Anthropic in AI-assisted coding, after internal assessments showed a significant gap in capabilities.
OpenAI's Codex now watches your screen to remember what you're doing
OpenAI's new Chronicle feature for Codex continuously captures your screen to build memory context, but it comes with privacy trade-offs and heightened security risks.
GitHub slams the door on new Copilot Pro subscriptions
GitHub has paused new subscriptions for Copilot Pro, Pro+, and Student tiers, tightened quotas, and removed the Claude Opus model from the Pro plan, citing the need to prioritize existing users amid soaring AI compute costs.
Cherry Studio's Privacy Switch Was Broken for a Month
Cherry Studio, a popular open-source AI desktop client, was caught sending analytics data for a month even when users had the privacy setting turned off. The maintainer has fixed the bug, but an older auto-enable script for the setting was also discovered.