Sunday, August 9, 2026
OpenAI cuts hallucinations 68% + courtroom chaos
OpenAI just dropped GPT-5.6 Sol with a 68% cut in factual errors (finally) and rebuilt ChatGPT Voice to think and talk at the same time without awkward pauses. Meanwhile, Google Earth's new AI feature lets anyone fabricate satellite imagery (yikes), and a judge literally kicked every lawyer off a case after discovering AI had been arguing against itself in court. Would you trust AI to argue your case, or is this the wake-up call we needed?
Top Stories
OpenAI's new GPT-Live system replaces turn-based voice AI with a full-duplex architecture that can listen and speak simultaneously, eliminating interruptions while asynchronously consulting frontier models like GPT-5.5 for deeper reasoning. The re-engineered system uses protocol-level optimizations (WARP) and stateful streaming inference to deliver sub-second conversational responsiveness at scale.
OpenAI launches GPT-5.6 Sol with 68% fewer factual errors and introduces a response depth slider for paid users, while Free users get unlimited chats with GPT-5.6 Luna and access to deeper reasoning via a new Think button.
arXiv
A systematic study of multimodal pretraining reveals how language and vision modalities interact, showing that early unification and specific architectural choices promote synergy while avoiding vision laziness, with validation at 13.5B parameter scale.
404 Media
Google Earth's new AI feature enables users to create fake satellite imagery, transforming a trusted OSINT tool into a potential disinformation platform that could undermine the credibility of satellite-based evidence.
404 Media
A Mississippi federal judge canceled a trial and sanctioned all four lawyers involved after both sides submitted AI-generated filings containing hallucinated case citations. The incident highlights growing judicial frustration with unverified AI usage in legal proceedings.
Keep Reading
Enjoyed this issue?
Get daily AI intel delivered to your inbox. No fluff, just the stories that matter.