HigherEd AI Daily: March 19 – Microsoft Challenges OpenAI, a Model That Trains Itself, and Free Courses From Anthropic

Leave a Comment / Higher Ed AI Tools / By
Hello,
Wednesday, March 19 brings a legal showdown between Microsoft and its own partner, a self-improving AI model that ran 100 training cycles without human input, and 13 free courses from Anthropic that educators can start today.
SHORT ON TIME? TODAY'S ESSENTIAL LINKS
AI Industry and Policy

Microsoft May Sue Its Own Partner Over a $50 Billion Amazon Deal

Microsoft is considering legal action against OpenAI and Amazon over a reported $50 billion cloud deal. Microsoft invested heavily in OpenAI under an agreement that grants it exclusive cloud-hosting rights. The new Amazon arrangement for OpenAI's Frontier product may violate that exclusivity clause. The dispute is ongoing and no lawsuit has been filed.
Why it matters for campuses. Universities that rely on Microsoft 365 Copilot or OpenAI's API for teaching tools are sitting between two giants in a legal dispute. Campus IT departments should watch this closely because service terms and pricing may shift depending on how this resolves.
AI Research

MiniMax Built an AI That Trained Itself Through 100 Improvement Cycles

Chinese AI company MiniMax released M2.7, a model that participated in its own training process. The model ran more than 100 self-evolution rounds, improving performance by 30 percent on software engineering benchmarks without human intervention. M2.7 can now autonomously handle data pipelines, training environments, and debugging routines. It scored 56.22 percent on SWE-Bench Pro.
Why it matters for campuses. A model that improves itself raises new academic integrity questions. If AI can write and refine its own training data, the assumption that human authorship underlies any AI output becomes harder to verify. This is a discussion worth having in your class now.
Teaching and Learning Technology

Gamma Imagine Enters the Classroom Design Space With a Direct Challenge to Canva

Gamma launched Gamma Imagine on March 17, adding AI image generation for branded assets to its existing presentation platform. Users type a text prompt and receive marketing materials, charts, and visual assets styled to their brand. Gamma now has 100 million users and integrates with ChatGPT, Claude, Zapier, and Atlassian. The tool competes directly with Canva and Adobe.
Why it matters for campuses. Educators who already use Gamma for lesson presentations can now generate branded visuals without leaving the platform. For faculty who create course materials or department communications, this removes one more step from the production process.
Free Resources for Educators

Anthropic Academy Offers 13 Free AI Courses With Certificates, From Beginner to Developer

Anthropic expanded its free learning platform this week. Anthropic Academy now offers 13 courses covering AI fluency, Claude Code, advanced prompting, and the Model Context Protocol. Certificates are included at no cost. The courses range from general AI literacy suitable for any faculty member to developer-level modules for those building classroom tools.
Why it matters for campuses. Professional development budgets are tight at most institutions. Thirteen free courses with certificates from a leading AI lab give faculty a credible, low-cost way to build AI fluency on their own schedule without institutional approval or expense.
Tool of the Day: Alma
Category: Productivity and AI Memory  |  Status: Free to start
Alma is a persistent memory layer for AI. It connects to your existing AI tools and remembers your preferences, context, and work history across sessions. Instead of re-explaining your role and goals each time you open a new chat, Alma carries that information forward so every session starts from where you left off.
Try this before Friday: Set up Alma with three facts about your teaching context—your subject, your student level, and one recurring task. Then open your next AI conversation through Alma and notice how much less setup time you spend.
A Final Reflection for Today
MiniMax released a model that ran 100 improvement cycles on itself before a human reviewed the results. That is a research milestone worth teaching. Not because it threatens your job, but because your students will work alongside systems like this. Knowing how they improve is part of being AI-literate in 2026.
Dr. Ali Green
Founder, Ask The PhD Community
askthephd.com
Not sure where you stand with AI in your classroom?
Take the 90-Second AI Readiness Assessment

Start the Assessment

Ask The PhD Community | askthephd.com | askthephd.substack.com

Leave a Comment