Good morning. It’s Monday, March 16th.
Quick note: We have a new look, and new layout! Going for a cleaner, more information-dense read.
The Alpha includes the must-read stories of the day, while The Feed replaces the old Etcetera section - and includes everything that happened in AI in the past 48 hours. New Tools are 5 new startups/projects that I’ve come across that I thought were worth sharing.
Since starting this newsletter in 2022, I’ve published 556 posts and sent a whopping 26,296,598 emails. I’m grateful to have die-hard readers from across the world, including employees and at OpenAI, NVIDIA (including the CEO… hi Jensen!) Meta, Anthropic, Apple, Google, and many more.
As always, please share your thoughts by replying to this email!
-Jeff
AI Breakfast
In Partnership with MedOS
Most AI still lives on a screen.
MedOS is bringing it into the hospital.
Built by the Stanford–Princeton AI Coscientist team, MedOS is a clinical AI co-pilot that combines multi-agent AI reasoning, XR smart glasses, and collaborative robotics to assist doctors during procedures.
A clinician wearing the glasses can interact with MedOS in real time while it analyzes what’s happening, surfaces insights, and supports decisions during care.
The system is designed to reduce medical error and clinician burnout, not replace doctors.
MedOS has already moved beyond the lab with deployments at the Stanford Blood Center and Stanford’s Department of Pathology, and will be featured at NVIDIA GTC 2026.
Thank you for supporting our sponsors!
AI consultant uses ChatGPT, AlphaFold, and Grok to find a possible treatment for his dog's cancer
Australian consultant Paul Conyngham utilized ChatGPT, AlphaFold, and Grok to develop a bespoke mRNA vaccine for his dog’s incurable mast cell cancer. After sequencing the tumor genome at UNSW Sydney, AI models identified target proteins and repurposed FDA approved substances for the final design.
Following administration, tumor shrinkage of 75 percent was reported. However, experts urge caution, noting the dog concurrently received conventional immunotherapy, making the specific efficacy of the AI generated vaccine difficult to isolate. Read more.
xAI full rebuild + Tesla terafab launch begins in 7 days for 40x AI5 chips
Elon Musk is vertically integrating his AI ambitions by launching the Tesla terafab in seven days to produce fifth-generation AI5 chips, while simultaneously executing a total foundational rebuild of xAI to fix a "not built right" infrastructure.
The AI5 chip targets a 40x performance jump over previous iterations, focused on high-volume inference for FSD and Optimus. New leadership from Cursor (Milich and Ginsberg) will report directly to Musk to bridge the competitive gap with rivals.
Within the software stack, the upcoming Grok Build will integrate yolo mode and new connectors to accelerate autonomous execution and enterprise app utility. Read more.
Google explains Nano Banana 2 capabilities alongside Groundsource launch
Google is bridging the gap between generative AI and physical reality by launching Groundsource, a Gemini-powered pipeline that converted decades of news into a 2.6 million record flood dataset, and a new grounding tool for Google Maps in AI Studio.
A major component of this update is a deep dive into the capabilities of Nano Banana 2 (Gemini 3.1 Flash Image), explaining how to use its new visual grounding via Google Search to verify architectural or biological accuracy and when to toggle its reasoning mode for complex spatial logic.
This technical breakdown highlights support for extreme 1:8 aspect ratios and a cost-effective 512px batch-to-upscale workflow that offers 95% of the Pro model's power at a fraction of the cost. Developers in AI Studio can now ground Gemini API responses directly in live, verified location results from Google Maps. Read more.
Claude Opus and Sonnet 4.6 reach general availability for 1M token context
Anthropic has announced general availability of a 1 million token context window for Claude Opus 4.6 and Sonnet 4.6. Standard pricing now applies across the entire window, eliminating the prior long context premium. Media capacity has increased sixfold, supporting up to 600 images or PDF pages per request.
Opus 4.6 achieves a 78.3% score on the MRCR v2 benchmark, demonstrating superior retrieval accuracy for large codebases and complex agent traces. These updates are live on the Claude Platform, Amazon Bedrock, Google Vertex AI, and Microsoft Azure Foundry, with 1M context now the default for Claude Code users. Read more.
Junior AI is an autonomous AI employee that integrates with your tools, learns your company context in minutes, and proactively executes tasks 24/7 to scale your team without increasing headcount.
Agent 37 provides instant, managed OpenClaw instances with full terminal access and 850+ pre-wired integrations, allowing you to deploy and run autonomous AI agents in sixty seconds.
Struct automates on-call runbooks by cross-referencing logs, metrics, and code to proactively root-cause alerts, providing impact analysis and suggested fixes directly within your existing engineering stack.
GStack provides Garry Tan's exact Claude Code setup, turning the assistant into a specialized agent team with six opinionated slash commands for plan reviews, shipping, and browser automation.
ElevenCreative by ElevenLabs is a single platform to generate, edit, and localize premium audio and video, using studio-grade AI voices, music, and visuals to reach audiences in 70+ languages.
Thank you for reading today’s edition.
Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.
Interested in reaching smart readers like you? To become an AI Breakfast sponsor, reply to this email or DM us on X!
Thinking of starting your own newsletter? AI Breakfast readers who sign up with Beehiiv receive a 14-day free trial and 20% off for 3 months.
