Good morning. Here is what mattered in AI today. Google’s Gemini Omni 1.2 Flash leads with longer, cheaper video generation, followed by NVIDIA’s widening cooperation with Korea, Microsoft’s data on AI-assisted Windows patching, and an unconfirmed report of a large gas-generator purchase for Anthropic’s data centers. Our explainers cover how IBM’s Palette builds slide decks and how verifiable rewards train reasoning models.
Daily DigestDigest
Daily Digest: Gemini Omni 1.2 Flash and NVIDIA in Korea
The AI stories that mattered on November 9, 2026: cheaper 15-second video, NVIDIA in Korea, Windows patch data and a reported gas-generator deal.
Today in 6 stories · November 9, 2026
- 01Model Releases
Gemini Omni 1.2 Flash makes 15-second video clips, 30% cheaper
Google's video model now runs to 15 seconds and costs $0.07 per second at 720p.
- 02Geopolitics & Global AI
NVIDIA expands Korea sovereign AI ties with NVLink Fusion, Dynamo
A broad cooperation framework with Korea's science ministry, with no disclosed investment figures.
- 03Cybersecurity
Microsoft details how AI speeds Windows vulnerability fixes
More than 1,100 CVEs fixed in October, but under half of sampled devices patched within two weeks.
- 04Energy
Anthropic reportedly buys $3.4 billion of gas generators for data centers
An unconfirmed FT report on 1.7 GW of on-site gas generation.
- 05Presentations
IBM's Palette trains a small model to build on-brand slide decks
A 20B model with one LoRA adapter scores 59.7 on PresentBench.
- 06Reinforcement Learning
Reinforcement learning with verifiable rewards (RLVR) explained
How checkable rewards train reasoning models.
Sources
This story was reported and written in our own words from the sources below. Primary sources are preferred.
- Gemini Omni 1.2 Flash — deepmind.google