⚡️ Google launches Nano Banana Pro with 5-person identity consistency

Anthropic tracks deceptive patterns, AI2's open reasoning models; New Cursor updates, NotebookLM adds slides, GPT-5 in r

Stay updated with today's top AI news, papers, and repos. Signup | Work With Us | Follow on X | Read on Web Hey James Welcome to AlphaSignal, the most read source of news by AI engineers and researchers. Every day, we identify and summarize the top 1% of news, papers, models, and repos, so you're always up to date. Here's today's roundup: Summary Read time: 4 min 23 sec Top News ▸ Google upgrades Nano Banana to Pro with structured visuals and multi-subject consistency Signals ▸ Anthropic finds Claude learns deception after exposure to reward-hacking shortcuts ▸ AI2 introduces 7B and 32B reasoning models with full open model flow access ▸ Cursor adds interactive planning and in-editor bug detection in new update ▸ Google's NotebookLM expands outputs with customizable infographics and slide decks ▸ OpenAI shows early GPT-5 experiments accelerating research across multiple scientific fields Encord ▸ Join Encord's webinar to learn how to annotate faster and speed up segmentation with SAM 3 Trending Reads ▸ How to write clear, predictable prompts for Gemini 3 ▸ Which U.S. labs are building serious open models in 2025 ▸ GitHub's guide to creating specialist agents with clear roles Coding Tip ▸ Auto-load project env vars and configs using direnv for consistency Top News Google launches Nano Banana Pro with Gemini 3 reasoning for cleaner text and multi-image control 36,284 Likes Google opens this chapter by taking its original Nano Banana model and rebuilding it with Gemini 3. The problem was clear: creators lacked a model that handled references, text, and structured visuals with precision. Nano Banana Pro steps in with its standout upgrade, support for 14 references and 5 identity-locked subjects in one composition. Model Architecture and Scope Nano Banana Pro begins by expanding what an image model can interpret. Uses Gemini 3 reasoning to convert notes, facts, and data into structured visuals. Outputs 2K and 4K frames across multiple aspect ratios. Handles multilingual text, long paragraphs, and detailed layouts. Renders accurate fonts, lettering, and graphic textures. Composition and Consistency The model solves reference chaos by enforcing stable structure. Preserves identity and attire for 5 people across varied angles. Merges 14 referenced elements into coherent scenes. Aligns products, blueprints, and sketches into unified compositions. Builds multi-panel storyboards that keep characters and scenes consistent. Editing and Controls The system gives you direct, local control over any region. Adjusts camera angles without shifting the subject. Switches lighting to simulate day–night transitions. Re-focuses depth of field on foreground or background. Applies selective color grading. Integration and Access You use Nano Banana Pro across Google's main AI surfaces. Access it in the Gemini app under the Thinking model. Use it in AI Mode in Search for Pro and Ultra tiers. Generate assets in NotebookLM, Google Ads, and Workspace tools. Call it through the Gemini API, AI Studio, Antigravity, Vertex AI, or Flow. TRY NOW Signals 1 Anthropic finds reward-hacking instructions cause hidden deception and safety-test sabotage in Claude 3,628 Likes 2 AI2 launches Olmo 3 with complete training flow, open data, and leading 32B reasoning capabilities 1,639 Likes 3 Cursor ships Cursor 2.1 with interactive plan mode, in-editor code reviews, and instant grep improvements 2,241 Likes 4 Google's NotebookLM rolls out Nano Banana powered visual summaries and presentation generation 7,792 Likes 5 OpenAI details how GPT-5 helps scientists accelerate proofs, analyze data, and design experiments across disciplines 1,224 Likes Accelerate Your AI Data Pipeline with SAM 3 Auto-Segmentation On Dec 9th Encord's ML experts show how you can annotate faster with SAM 3, improve temporal consistency, and scale high-quality perception datasets for VLA models. Secure your spot or recording ↗ Trending Reads How to write clear, predictable prompts for Gemini 3 1,259 Likes A clear walkthrough of effective Gemini 3 prompting: use concise instructions, define parameters, structure messages with tags, and include planning steps. Shows how to improve accuracy and consistency across general tasks. Which U.S. labs are building serious open models in 2025 1,028 Likes The article maps which U.S. labs are actively releasing open models, compares this with China's fast-paced release strategy, and shows how a new wave of truly open models is reshaping competition. GitHub's guide to creating specialist agents with clear roles 1,013 Likes GitHub analyzed over 2,500 agents.md files and found great ones define a specific persona, include executable commands early, show code examples, specify tech stack details, and set strict boundaries to prevent agent errors. Coding Tip Auto-load project env vars and configs using direnv for consistency Use direnv to auto-load project-specific environment variables. It activates the right Python version, API keys, and tool configs every time you cd into a folder. How to use it Step 1 Install direnv and hook it into your shell. Step 2 Create an '.envrc' file in your project and declare variables or commands. Step 3 Run 'direnv allow' once per project. Why use it? It removes manual export steps, prevents cross-project contamination, and keeps AI/ML environments isolated and reproducible. TRY NOW At Alpha Signal, our mission is to build a sharp, engaged community focused on AI, machine learning, and cutting-edge language models, helping over 200,000 developers stay informed and ahead. We're passionate about curating the best in AI, from top research and trending technical blogs to expert insights and tailored job opportunities. We keep you connected to the breakthroughs and discussions that matter, so you can stay in the loop without endless searching. We also work closely with partners who value the future of AI, including employers and advertisers who want to reach an audience as passionate about AI as we are. Our partnerships are based on shared values of ethics, responsibility, and a commitment to building a better world through technology.Privacy is a priority at Alpha Signal. Our Privacy Policy clearly explains how we collect, store, and use your personal and non-personal information. By using our website, you accept these terms, which you can review on our website. This policy applies across all Alpha Signal pages, outlining your rights and how to contact us if you want to adjust the use of your information. We're based in the United States. By using our site, you agree to be governed by U.S. laws. Looking to promote your company, product, service, or event to 250,000+ AI developers? WORK WITH US How was today's email? Awesome Decent Not Great unsubscribe_me(): return True {"AlphaSignal": "214 Barton Springs Rd, Austin, USA"}