AI Updates on 2025-11-22

Google's Nano Banana Pro achieves #1 ranking on both Text-to-Image Arena (+84 points over Nano Banana) and Image Edit Arena (+41 points over Nano Banana), with both Nano Banana models claiming top spots on the Image Edit leaderboard @arena
Gemini 3 Pro demonstrates state-of-the-art performance on math benchmarks, released just 3 days prior to these achievements @OfficialLoganK
Perplexity announces Nano Banana Pro and Sora 2 Pro as default generation models for Perplexity Max subscribers @perplexity_ai
NVIDIA releases Nemotron-Personas Collection, multilingual synthetic persona datasets including 6M personas for USA and Japan, and 21M for India, created with NeMo Data Designer for fine-tuning AI systems @NVIDIAAIDev
Nex-N1 series of agentic foundational models launches on Hugging Face in sizes from 8B to 671B parameters, with strengths in tool-use, web-search, and real-world agentic workflow @Xianbao_QIAN

Bret Taylor's Sierra reaches $100M ARR in under two years, demonstrating rapid growth in AI-powered customer service solutions @TechCrunch
OpenAI partners with Foxconn in strategic collaboration, expanding AI infrastructure capabilities @gdb
Google's team provides 24/7 support for customers scaling with Gemini 3 Pro and Nano Banana Pro, including higher API rate limits @OfficialLoganK
Valve demonstrates exceptional business efficiency with ~$17B revenue and ~336 employees, achieving >$50M per employee with average pay of ~$1.3M/person, representing one of the most efficient businesses globally @deedydas
Top churn reason for AI product management tool ChatPRD is "I love it and it's very helpful but it's not allowed," highlighting enterprise adoption barriers where employees cannot spend $8/month of their own money despite AI tools improving productivity @clairevo
OpenAI hosts AI Jam mentoring 1,000 small business owners to build AI tools tailored to their needs, spanning professional services, restaurants, retailers, creative services, and local businesses @gdb

Simon Willison and others discuss prompt injection vulnerabilities in GitHub MCP server and the development of common MCP Apps standard across Anthropic, OpenAI, and MCP-UI @ibuildthecloud
Andrej Karpathy seeks quantitative definition of "slop" in AI-generated content, noting intuitive ability to estimate quality but difficulty in formal measurement @karpathy
Tesla announces progress toward shipping Full Self-Driving (Supervised) in Europe after 12+ months of work, with Netherlands National approval expected February 2026, though current regulations make FSD illegal in its current form despite proven safety record @teslaeurope

Google showcases Gemini 3 applications including one-shot interactive maps, realistic physics demos, and game creation, demonstrating versatility in educational and creative use cases @GeminiApp
Figma integrates Google's Gemini 3 Pro with Nano Banana across products for dark mode illustrations, in-situ imagery placement, brand-consistent content creation, profile photo updates, 3D visualization, and moodboard-to-scene conversion @nlevin
Cursor Agent Review launches as integrated code review feature running optimized pipeline for $0.40-$0.50 average cost, providing second set of eyes on codebase with edge case detection @RayFernando1337
Perplexity announces daily updates to Perplexity Finance including in-line annotated price tickers on finance-related queries @AravSrinivas
Nano Banana Pro demonstrates capability to create recursive meta-imagery, generating "amateur photograph from 1998 of artist copying image from computer screen to oil painting, where the image is itself the photo of the artist painting the recursive image" @goodside
Wabi integrates Gemini 3 enabling creation of interactive mini apps including black hole simulations @wabi

Research paper demonstrates GPT-5 capable of new discoveries in challenging fields, though process currently requires guidance and expertise without repeatable methodology for others to follow @emollick
Google DeepMind supports leading academic labs worldwide with Gemini 3 access via API, with new researchers able to apply for credits and access @divy93t
Ethan Mollick observes AI organizational challenges regarding how AI alters economies of scope determining firm boundaries, transaction costs, and efficiency/creativity trade-offs, questioning whether this brings return to centralized CEO decision-making since the shift from U-form to M-form organizational structures in the 1920s @emollick
Ilya Sutskever highlights important work from Anthropic on AI safety and alignment research @ilyasut