Microsoft's cancer-mapping AI

PLUS: Create a complete brand kit with Nano Banana Pro

Good morning, AI enthusiasts. Advanced tumor mapping used to cost thousands and take days of lab work. Microsoft just figured out how to do it with AI and a $10 tissue slide.

With the company’s new open-source GigaTIME model drawing insights from 40M cell samples in seconds, cancer research may have just gotten a serious accessibility upgrade.

In today’s AI rundown:

  • Microsoft AI turns tissue samples into cancer maps

  • Mistral’s Devstral 2 and new coding assistant

  • Create a brand kit with Nano Banana Pro

  • OpenAI, Anthropic, Block team up on agent standards

  • 4 new AI tools, community workflows, and more

LATEST DEVELOPMENTS

MICROSOFT

Image source: Reve / The Rundown

The Rundown: Microsoft just released and open-sourced GigaTIME, a new AI model that can extract thousands of dollars worth of tumor insights from a basic $10 tissue slide — analysis that previously required expensive lab work and days of processing.

The details:

  • The model learned from 40M cell samples provided by Providence Health, matching simple slides with advanced immune system scans.

  • Researchers tested it on over 14,000 cancer patients, creating a virtual library of 300,000 detailed tumor images across 24 different cancer types.

  • The ‘virtual population’ analysis surfaced over 1,200 patterns linking immune activity to factors like cancer stage and patient survival.

Why it matters: Cancer research is entering a new phase where AI can unearth meaningful insights from routine data that would have required costly, time-intensive lab work just years ago. GigaTIME is part of an AI-led shift towards models that make population-scale analysis affordable and fast enough to influence treatment decisions.

TOGETHER WITH HUBSPOT

The Rundown: Stop drowning in tasks and start delegating like top performers who use AI to handle 80% of their routine work. HubSpot’s free kit provides the exact templates, prompts, and systems that 10,000+ professionals use to complete a full day's work by lunch.

The AI Assistant Kit includes:

  • Ready-to-use "AI Assistant Command Center" for managing all your AI tools

  • Step-by-step implementation guide to master AI delegation in under 60 minutes

  • Built-in ROI calculator to track your time savings and productivity gains

  • Advanced prompts and templates to turn ChatGPT into a 24/7 productivity partner

MISTRAL

Image source: Mistral

The Rundown: French AI startup Mistral just launched Devstral 2, the next-gen version of its coding-focused model family — alongside Vibe CLI, the company’s first move into autonomous coding agents.

The details:

  • Mistral’s 123B Devstral 2 scores a 72.2% on the SWE-bench Verified, nearly matching the top open-weight model (Deepseek V3.2) while being 5x smaller.

  • The Small 2 variant challenges other open-weight rivals despite being just 24B parameters and able to run on a single GPU or laptop CPU.

  • Devstral 2 is notably released with a ‘modified’ MIT license that puts restrictions on use for companies with $20M+ in monthly revenue.

  • Mistral’s Vibe CLI is a terminal-native coding agent that can scan codebases and handle multi-file changes, free to use under an Apache 2.0 license.

Why it matters: Mistral is shipping this month, with a strong Devstral release coming just days after the Mistral 3 launch. While the main model moves near the open frontier for coding, the small variant looks to be a real development upgrade for models able to be deployed locally and offline on consumer devices.

AI TRAINING

The Rundown: In this tutorial, you will learn how to create a full brand kit PFD, logo variations, and social media templates using Google’s Nano Banana Pro, customized to your business, target audience, or company guidelines.

Step-by-step:

  1. Go to AI Studio and prompt: "I need a complete brand kit. Business: [NAME]. What we do: [ONE SENTENCE]. Target: [WHO + PAIN]. Personality: [3 ADJECTIVES]. Interview me with 3-5 questions, include tagline ideas"

  2. Answer the questions, then prompt: "Use Nano Banana Pro to generate 4 logo options in a 2x2 grid" — pick the best and request variations (secondary, monochrome black/white, icon-only)

  3. Prompt: "Generate the logo system slide for my brand kit deck," then ask it to outline remaining slides and generate each one at a time

  4. Assemble slides in Google Slides and export as PDF

Pro tip: Use the prompt pattern to get consistent outputs from Nano Banana every time: “interview me → generate one asset → outline remaining assets → generate”

PRESENTED BY INVISIBLE

The Rundown: AI won’t run your company in 2026, but it could let ten people do the work of a thousand if you fix adoption. Invisible’s 2026 Agnetic Field Report reveals insights from both sides – inside the labs training over 80% of leading models and inside enterprises making them work at scale.

Invisible’s 2026 predictions:

  • How enterprise agents get better through multimodal reasoning, not just more features

  • What AI safety looks like when agents are touching money, contracts, and customers

  • How RL environments leap from Atari games to an enterprise test bed

Download the 2026 AI Trends Report to see what's next for enterprise AI.

AI RESEARCH

Image source: AAIF

The Rundown: OpenAI, Anthropic, and Block just co-founded the Agentic AI Foundation, a neutral entity governed under the Linux Foundation — with each company pooling open-source tools to build standards across the AI agent ecosystem.

The details:

  • Each founding company is donating a core project — Anthropic's Model Context Protocol, OAI's AGENTS.md, and Block's Goose agent framework.

  • The MCP has already been adopted by ChatGPT, Cursor, Gemini, and VS Code, with over 10,000 active public servers since its launch last year.

  • Supporting members include Google, Microsoft, AWS, Bloomberg, and Cloudflare, expanding industry alignment for the AI agent framework.

  • The Linux Foundation will provide neutral governance, drawing on experience hosting other tech infrastructure projects like PyTorch and Kubernetes.

Why it matters: The AAIF is an important development to avoid AI agents being built across the industry behind walled gardens, with the commitment to shared standards helping both accelerate adoption and allow the tech to work more efficiently on the same rails for a better experience for users and developers.

QUICK HITS

  • 🗺️ Unwrap Customer Intelligence - Turn unstructured customer feedback into data-backed insights that inform your product roadmap*

  • 🚀 Vibe CLI - Mistral’s new open-source CLI powered by Devstral

  • ⚙️ GLM-4.6V - Zhipu AI's open-source multimodal family with native tool use

  • 🧑‍💻 Claude Code - Deep-context AI coding assistant with new Slack integration

*Sponsored Listing

Meta is reportedly planning to drop a new frontier model codenamed “Avocado” in Q1 2026, which “could be a proprietary model,” unlike its previous open-source launches.

The U.S. War Department introduced GenAI.mil, a new AI platform for the U.S. military, with Google's Gemini debuting as the first AI model available for use.

Microsoft announced a $19B CAD investment to expand AI infrastructure across Canada through 2027, alongside a plan to keep Canadian user data within the country.

The EU opened a new investigation into Google over whether its AI search summaries and AI Mode improperly use content from websites and videos without compensation.

Anthropic is partnering with Accenture to train 30,000 of the firm’s consultants on Claude to help enterprises move AI pilot projects forward.

The U.S. DOJ detained two men for allegedly running a smuggling network to ship Nvidia chips to China, part of an investigation that has seized more than $50M in GPUs.

COMMUNITY

Every newsletter, we showcase how a reader is using AI to work smarter, save time, or make life easier.

Today’s workflow comes from reader Laura L. in Minneapolis, MN:

"My daughter recently turned 18. We fed into AI meaningful locations around our house and asked it to generate scavenger hunt clues leading from spot to spot, ending at the treasure - a college-ready laptop. AI and I went through a couple of iterations together, with the results ranging from way too obvious "look in your boots" clues to elaborately-worded hints in Victorian English before we landed on clever rhyming clues."

How do you use AI? Tell us here.

That's it for today!

Before you go we’d love to know what you thought of today's newsletter to help us improve The Rundown experience for you.

Login or Subscribe to participate in polls.

See you soon,

Rowan, Joey, Zach, Shubham, and Jennifer — the humans behind The Rundown

Reply

or to participate.