In partnership with

Welcome back!

Google DeepMind just set a new visual benchmark with Nano Banana Pro, a model that reasons about pictures. By combining structured logic and real-world data, it’s able to generate amazing visual explanations. Let’s unpack…

Today’s Summary:

• 🚀 Nano Banana Pro redefines visual reasoning

• 🔥 OpenAI unveils GPT-5.1-Codex-Max

• 🌦️ WeatherNext 2 forecasts the world hourly

• 🇪🇺 Europe plans to ease AI and cookie rules

• 🤯 Gemini 3 refused to believe it’s 2025

• 💬 ChatGPT group chats roll out globally

• 🛠️ 2 new tools

TOP STORY

Nano Banana Pro sets new bar for structured visuals

The Summary: Google DeepMind has released Nano Banana Pro, an image generation and editing model built on Gemini 3 Pro. Unlike previous versions, it can reason about structure, text, and real-world facts before rendering. The model supports up to 14 input images, 4K output, and real-time grounding through Google Search. Its precision makes it suited for enterprise design, education, and creative production.

Key details:

• Ranked #1 in compositional image generation and infographic accuracy

• Supports flawless multilingual text rendering inside images

• Handles up to 14 image inputs while keeping up to 5 people consistent across several scenes, previously a major limitation in AI imaging

• Priced at $0.13 per 2K image and $0.24 for 4K

• Pro rollout includes Google Ads, Workspace Vids, Slides, Vertex AI

• Consumer rollout via Gemini app and Pro/Ultra tiers

Why it matters: Nano Banana Pro delivers professional image generation and complex infographics. It composes visuals with logic, checking facts, structure, and meaning before a single pixel appears. For teams and enterprises, it can power instant, production-ready visuals built with the accuracy once reserved for code.

Prompting tips

FROM OUR PARTNERS

Teach Anything in Seconds with This Free AI Extension

Effortless Tutorial Video Creation with Guidde

Transform your team’s static training materials into dynamic, engaging video guides with Guidde.

Here’s what you’ll love about Guidde:

1️⃣ Easy to Create: Turn PDFs or manuals into stunning video tutorials with a single click.
2️⃣ Easy to Update: Update video content in seconds to keep your training materials relevant.
3️⃣ Easy to Localize: Generate multilingual guides to ensure accessibility for global teams.

Empower your teammates with interactive learning.

And the best part? The browser extension is 100% free.

Check out Guidde

OPENAI

GPT-5.1-Codex-Max enables 24-hour autonomous development

The Summary: OpenAI has released GPT-5.1-Codex-Max, a new coding model designed for very complex, day-long software engineering tasks. Built with a mechanism called compaction, it can sustain coherent reasoning across millions of tokens and complete projects lasting over 24 hours. The model is faster, more efficient, and now native to Windows environments. It replaces GPT-5.1-Codex as the default engine across all Codex interfaces.

Key details:

• Benchmarks: SWE-Bench Verified: 77.9% vs Gemini 3 Pro 76.2%

• Uses 30% fewer “thinking tokens” for equal or better accuracy

• Runs up to 42% faster on real-world tasks

• Early testers reported it tends to follow procedural instructions very literally, to the point of making some unexpected complications

• Available today within Codex CLI, IDE extension

Why it matters: OpenAI’s quick release after Gemini 3 shows how tightly matched the AI-coding race has become. Codex-Max gives OpenAI an endurance advantage for complex tasks running for hours. The escalation between OpenAI and Google is shaping not just performance charts but the very workflow of modern software development.

System Card

FROM OUR PARTNERS

Turn Your Ideas Into AI Agents

Build AI agents with your voice. Automate in minutes.

With Lindy, you can build AI agents and apps simply by describing what you want, like:

"Create a booking platform for my business."
"Automate my sales outreach."

From inbound lead qualification to customer support, Lindy has tons of agents to streamline your workflows.

Get $20 Worth of Free Credits

GOOGLE

WeatherNext 2 brings hour-level global forecasts

The Summary: Google DeepMind released WeatherNext 2, an AI model that predicts global weather at hour-level resolution. Using a new Functional Generative Network, it simulates hundreds of realistic scenarios from a single data input. The model now powers forecasts across all Google products, including Search, Gemini, and Maps, and is open to developers through Earth Engine, BigQuery, and Vertex AI.

Key details:

• Forecasts hundreds of scenarios in under a minute using one TPU, compared to hours on supercomputers

• Surpasses predecessor model on 99.9% of weather variables

• Uses “noise injection” to create realistic variability

• Available to researchers and developers via Google Earth Engine, BigQuery, and Cloud Vertex AI early access

Why it matters: Weather forecasting has long depended on physics-based models that strain even supercomputers. WeatherNext 2 lowers the baseline, delivering ensemble predictions at massive speed and resolution. The model’s design could help in disaster preparedness, supply chain planning, and renewable energy output.

Research paper

QUICK NEWS

Quick news

• Europe plans to relax AI rules and cut cookie banners

• Gemini 3 refused to believe it was 2025

• Group chats in ChatGPT are now rolling out globally

TOOLS

🥇 New tools

• Guideflow - Create interactive demos with AI

• UPCV - Create a professional resume in 3 minutes