Microsoft AI Surpasses Human Doctors👨‍⚕️

Artificial intelligence continues to make headline-worthy leaps across industries—from revolutionizing healthcare diagnostics to transforming education and creative workflows.

🔬 Microsoft's MAI-DxO has stunned the medical world by outperforming experienced doctors in solving complex diagnostic cases. Tested on real patient records from the New England Journal of Medicine, the AI achieved an 85.5% accuracy rate, dramatically surpassing the 20% average of human physicians. Using a coordinated panel of LLMs, this system mimics the step-by-step reasoning of a team of doctors, offering not only accuracy but cost-efficiency. While still under validation, it's a major stride toward medical superintelligence.

📚 Google's Gemini in Classroom is now free for all educators with Workspace for Education accounts, offering 30+ AI tools that streamline lesson planning, grading, and content creation. With features like NotebookLM, Gems, and Read Along, teachers can personalize learning, track student progress, and create custom experiences—all with AI as a co-pilot. It's a bold step toward making AI-powered education more accessible and impactful globally.

🖼️ Meanwhile, a new benchmark by Bria.ai evaluates leading text-to-image models for enterprise use. The results? Google Imagen 4 tops in image quality but lacks transparency, while Bria 3.2 emerges as the most balanced model—delivering solid visual results, full compliance, open access, and developer-friendly architecture. For B2B use cases where licensing, prompt control, and safety matter, Bria 3.2 stands out as the top contender.

Let’s explore.

Microsoft Saying AI Outperforms Doctors in Complex Diagnoses

Microsoft's new AI Diagnostic Orchestrator (MAI-DxO) represents a significant advance in medical artificial intelligence, demonstrating the ability to sequentially investigate and solve some of the most complex diagnostic challenges in medicine—cases that even expert physicians often struggle with. When tested on 304 challenging case records from the New England Journal of Medicine, MAI-DxO achieved a diagnostic accuracy of 85.5%, which is more than four times higher than the 20% mean accuracy achieved by a group of 21 experienced physicians from the US and UK. The system operates by emulating a virtual panel of physicians, orchestrating multiple large language models to collaboratively ask questions, order tests, and refine diagnoses step-by-step—mirroring real-world clinical reasoning rather than relying on static, multiple-choice answers. Notably, MAI-DxO also delivered these results more cost-effectively, reducing unnecessary diagnostic testing and overall resource expenditure compared to both human doctors and standalone AI models. While this research marks a major step toward "medical superintelligence," the technology is not yet approved for clinical use and requires further validation in real-world healthcare settings.

Gemini in Classroom: Free AI Tools Empower Educators

Google has announced that its AI-powered suite, Gemini in Classroom, is now available at no cost to all educators with Google Workspace for Education accounts, offering over 30 new tools designed to save teachers time and enhance both teaching and learning. Gemini in Classroom enables educators to quickly generate lesson plans, quizzes, rubrics, and engaging content, as well as collaborate with AI to brainstorm ideas and differentiate instruction for students. New teacher-led AI experiences, like NotebookLM and Gems, allow teachers to create interactive study guides, custom AI-powered helpers, and podcast-style audio overviews tailored to classroom materials, giving students more personalized support. The platform is also rolling out advanced analytics, enabling teachers to track student progress against learning standards, identify students needing extra help, and gain insights into engagement through a new Analytics tab. Additionally, the Read Along feature now supports custom content creation and multiple reading modes—including silent reading and listening—while providing real-time feedback and comprehension data. These updates reflect Google’s commitment to providing safe, responsible, and flexible AI tools that empower both educators and students worldwide.

Enterprise Text-to-Image Model Benchmarking

A recent benchmarking study by Bria.ai analyzed five leading text-to-image (T2I) models—Adobe Firefly 4.0, Bria 3.2, Google Imagen 4, Flux.1-Dev, and Stability 3.5 Large—specifically for commercial B2B use cases, focusing on output quality, technical implementation, and risk and regulation1. The evaluation found that while Google Imagen 4 leads in overall image quality and visual appeal, it falls short in data transparency and developer accessibility due to its closed-source nature and lack of detailed training data disclosure. In contrast, open models like Bria 3.2, Flux.1-Dev, and Stability 3.5 Large deliver strong output quality, reliable text rendering, and prompt alignment, with Bria 3.2 standing out for its compact architecture, full licensing compliance, and robust safety stack. Adobe Firefly 4.0, despite being trained on licensed data, ranked lowest in output quality and flexibility, offering limited API access and no public fine-tuning options. The study underscores that for enterprise adoption, models must balance high visual quality with legal defensibility, transparency, and ease of integration. Bria 3.2 is highlighted as a model that combines strong performance, open access, and comprehensive compliance, making it especially suitable for regulated or brand-sensitive commercial environments.

Hand Picked Video

I'm demonstrating how to create YouTube thumbnails using AI. I start by uploading a simple sketch to ChatGPT, then refine it with specific prompts like "viral thumbnail ideas.

Top AI Products from this week

• Rybbit - Next-gen, open source, lightweight, cookieless web & product analytics for everyone.

• co.dev MCP - Instantly turn ideas into full-stack Next.js apps with Codev—now powered by MCP Server for seamless integration with Prisma, PayPal, and 25+ other services.

• Handit.ai - Handit evaluates every AI agent decision, auto-generates better prompts and datasets, A/B-tests improvements, and lets you control what goes live.

• Mindly - Capture anything in seconds on your Mac. Mindly auto-organizes, tags, summarizes, and connects your ideas, so nothing ever gets lost or forgotten. Upgrade your workflow with a lightning-fast second brain that’s always ready.

• PrompTessor - Transform your AI interactions with PrompTessor advanced prompt analysis and optimization platform. Get detailed analysis, actionable feedback, and performance metrics to maximize your AI tool effectiveness.

• The Influencer AI - Create consistent, tailor-made AI influencers from custom traits or 10 photos. Then instantly craft photos, try-ons & talking video with them—any style, any pose—an all-in-one tool that cuts weeks of shooting and editing down to minutes. Great for ecomm & ads.

This week in AI

• Figma New Export & Vector Tools - Figma Make now imports styles from Design libraries for consistent prototypes; Vector Edit adds variable width, bend, lasso, and paint tools for precise editing—both on paid plans, Full seat required.

• Airtable Relaunches as AI-Native - Airtable now features Omni, an AI agent for conversational app building, data analysis, and workflow automation—AI tools are included on all plans, even free

• AI Band The Velvet Sundown - The Velvet Sundown, an AI-generated psych-rock band with no real members, has gained over 500,000 Spotify listeners in under a month, releasing 2 albums and a third soon. Their AI-created music and images spark debate on AI transparency in streaming.

• Apple Eyes Anthropic, OpenAI to Power Siri - Apple may replace its own AI with Anthropic's Claude or OpenAI's ChatGPT for Siri, marking a major shift as it struggles to match rivals in generative AI

• Grammarly Acquires Superhuman for AI Suite - Grammarly is acquiring email startup Superhuman to boost its AI productivity platform; Superhuman CEO and team will join, expanding AI tools for email and workflow.

• Cloudflare Unveils 'Pay per Crawl' for AI Bots - Cloudflare’s new “Pay per Crawl” marketplace lets websites charge AI bots for scraping, block them, or set custom rates—giving publishers control and new revenue streams.