Tops GPTs 
IBM researchers report Palette uses a LoRA-trained 20B open-weight model and a check-and-repair harness to generate slide decks from structured briefs. On PresentBench it ranks first in Economics and Education and second on the three-domain mean.
The CUGA team ran the same open-weight model, GLM-5.3-flash, across three agent harnesses on 90 AppWorld tasks and found task completion rates of 81.1% (CUGA), 46.7% (Hermes) and 30.0% (Deep Agents). CUGA emphasizes runtime policy enforcement and on-premise deployment options.
Heimstaden began using OpenAI in June 2025 to automate parts of lettings, property assessment and procurement. The company reports a 99% cut in staged-image costs in the Netherlands, 20,112 assets assessed with Codex, and 50% employee adoption of ChatGPT Enterprise.
Apple has begun developer beta testing of its Audio Intelligence features in iOS 27.3 for Apple Watch Series 12 (and Apple Watch Ultra 4 support). The opt-in tools include sound and song detection via Shazam, Live Rewind 15-second snippets, and summarized Siri Recaps.
A new Model Integration Contract (MIC) workflow packages small, reproducible fixtures, runnable usage examples and recorded baselines to catch regressions earlier. A UMT5 fixture replayed commits from July 30 to September 28, 2026 and flagged a decoder attention bug.
Google launched Dreambeans in India, its first market outside the U.S., offering AI-generated personalized visual stories on Android and iOS in 13 local languages. The app draws on connected Google services and now uses the Veo model to animate select stories.
OpenAI announced a new internal security sprint on November 2, 2026, partnering with Cathedral Systems to use models in testing and hardening systems. The company also said it will more than quadruple rewards for critical findings and expand access for frontline defenders.
OpenAI says it built Pando, a new backend for Git, to handle a surge in parallel work driven by Codex. The company reports Pando cut median push latency nearly 5× and now serves almost a petabyte of Git data per day while remaining Git-compatible.
Trajectory closed a $45M Series A at a $300M valuation to let nontechnical specialists improve open-source models through continual learning. The 18-person startup says its platform is now public and is used by companies including Harvey, Rogo and Nvidia.
In a revised S-1A, Anthropic cautions that using AI to accelerate R&D could trigger recursive self-improvement (RSI) and outpace safety measures. The company says it has not yet crossed its RSI threshold and plans to keep humans involved in frontier model development.