• AI Academy
  • Posts
  • 🤖 Google's "Nano-Banana" Sets New Standards

🤖 Google's "Nano-Banana" Sets New Standards

Plus: Microsoft launches its first AI models and Claude gets to Chrome

September 04, 2025

Hello AI Enthusiast,

This week Google dropped "nano-banana" (yes, that's what they're calling it), an image generator that finally handles text and keeps characters consistent. Microsoft launched their first homegrown models, trying to ditch OpenAI dependence. And Anthropic found a clever way into browser wars with a Chrome extension instead of building their own browser.

Let's dive in.

The Big Picture 🔊

Google Launches Gemini 2.5 Flash Image

Google released Gemini 2.5 Flash Image (which they're calling "nano-banana"), an image generator that can maintain character consistency across multiple images, blend photos together, and make targeted edits through text prompts. Available now at $0.039 per image, it includes features like removing objects, changing backgrounds, and merging elements from different photos while keeping the same character or style consistent.

Gioele Mottarlini
Gioele MottarliniCOO and Image Addict

Character consistency and multi-image fusion are genuinely useful for designers and photographers. The "nano-banana" branding is quirky marketing that actually works.

The resolution limitation matters though, these work for social media but won't cut it for print campaigns yet. The SynthID watermarking helps with fact-checking, but won't stop deepfakes from spreading on social platforms.

Microsoft Launches First In-House AI Models

Microsoft unveiled its first homegrown AI models: MAI-Voice-1 and MAI-1-preview. The voice model can generate a minute of audio in under one second and already powers features like Copilot Daily's AI news host. MAI-1-preview is designed for everyday consumer queries rather than enterprise use.

The company plans to integrate MAI-1-preview into Copilot, which currently relies on OpenAI's models, and is testing it publicly on the LMArena benchmarking platform.

Biljana Prlichkova
Biljana PrlichkovaPRM and Outdoor Enthusiast

Microsoft's consumer focus is genuinely surprising since their bread and butter comes from enterprise customers. But they're sitting on massive amounts of consumer data from Windows, Edge, and Office that could make their models way more useful for everyday tasks.

Most people encounter Copilot without realizing it through Edge or Office anyway, so Microsoft's betting on capturing users who just stick with whatever's already installed.

Microsoft's building their first AI models while most companies still can't figure out if their teams actually know how to use ChatGPT properly.

Our Corporate AI Training teaches practical skills, effective prompting, spotting useful output versus nonsense, and identifying where AI saves time versus where it creates more work.

Anthropic Launches Claude for Chrome Extension

Anthropic launched Claude for Chrome, a browser extension for 1,000 Max subscribers ($100-200/month) that lets Claude see your screen and complete tasks on your behalf with permission. The agent can chat in a sidebar while maintaining context of everything you're browsing. It joins the growing browser AI race with Perplexity's Comet and rumored OpenAI browser, especially as Google faces antitrust pressure to sell Chrome.

Andrea Mattiello
Andrea MattielloCM and Board Lover

Anthropic's Chrome extension approach is way smarter than standalone browsers or ChatGPT's web agent that goes off and does random stuff unsupervised. You stay in Chrome, Claude sits in a sidebar, and you maintain control while it helps with specific tasks on the page you're already viewing. Much smoother than switching to a completely new browser.

Having Claude there when you're comparing products or reading through complex pages could be genuinely useful.

Bits and Bobs 🗞️

  • The UN has launched two AI governance bodies to link science with policy and foster global, ethical collaboration.

  • Imperial College London developed an AI stethoscope that detects heart failure, valve disease, and arrhythmias for earlier diagnosis and treatment.

  • China’s State Council unveiled an “AI Plus” plan targeting 70% AI use by 2027, 90% by 2030, and full integration by 2035 across all sectors.

  • OpenAI announced the general release of its Realtime API, adding advanced speech-to-speech, image input, SIP calls, and new voices for voice agents.

  • An Anthropic report warns that threat actors misuse AI, like Claude, for cybercrime, prompting stronger safety measures.

  • Meta is restricting teen chatbot use to educational and creative characters, blocking sensitive topics as interim safety measures.

  • Elon Musk’s X has sued Apple, claiming its OpenAI partnership gives ChatGPT an unfair edge that threatens competition.

  • xAI released Grok-code-fast-1, a new model for faster agentic coding workflows with high efficiency, now available through its API.

  • Microsoft’s Copilot is coming to select 2025 Samsung TVs and monitors, offering free voice-powered AI assistance for entertainment and daily use.

That's a wrap on our newsletter! Before you go, here’s a quick recap of our offerings:

  • AI Academy Membership: Get 12 months of access to all our cohort-based programs, live webinars, on-demand courses, and tutorials.

  • AI Agent Bootcamp: Accelerate processes and solve business problems by mastering prompts and building AI Agents, without coding.

  • Practical Introduction to ChatGPT: A free course on using ChatGPT confidently, understanding its workings, and exploring its potential.

  • Customized Corporate Training: Equip your team with the skills they need to unlock the potential of AI in your business.

Catch you next week! 👋