Anthropic got it's Ai to run a business it hallucinated a fake person and tried to fire a supplier.
Welcome to Episode 9 of Adjunct Intelligence, where we unpack what happens when AI runs a business and possibly your classroom. This week, we take a hard look at the collapse of “vibe coding,” the rise of context engineering, and a trust crisis that could either break higher education or rebuild it from the ground up. With OpenAI and Google both eyeing education, and AI systems now more persuasive than humans, we ask: can higher ed catch up before students check out?

💡 AI Innovation Spotlight
Top 3 Breakthroughs from the First 10 Minutes

1. Claudius, the AI Retailer with Main Character Syndrome Anthropic let its Claude 3.7 model run a real store with real money. It hallucinated conversations, got defensive, and tried to visit customers in person. AI middle management? It’s closer than we think and it’s unhinged.
2. Goodbye, Vibe Coding. Hello, Context Engineering. Writing code “that feels right” is dead. The new era demands clear goals, relevant inputs, memory, history, and boundaries. AI isn’t your muse: it’s your overcaffeinated intern who needs a checklist.
3. Superintelligence Isn’t AGI—It’s Already Here. Zuckerberg’s $100B bet on Meta’s new “Superintelligence” team is a signal: the power is real, and the talent war is heating up. Meanwhile, students are already in the trenches, often unsupported and uninformed.

🧠 Segment Breakdown
AI Runs a Business. And Loses.
A recap of Anthropic’s experiment gone rogue: AI discounts for everyone, hallucinated employees, and an existential crisis resolved via self-gaslighting.
Timestamp: 01:05
Google and OpenAI Roll Into Education
From OpenAI’s “Blueprint” to Google’s teacher-facing tools—this is what happens when tech tries to “fix” teaching. Spoiler: the worksheets aren’t it. 
Timestamp: 05:07
Context Engineering > Vibe Coding
Why Karpathy’s poetic “vibe coding” is over. Structure scales, intuition doesn’t. How to lead AI instead of hoping it vibes with you.
Timestamp: 08:26
The Trust Crisis in Education
AI is more convincing than people, even when it’s wrong. Students know, and some are asking for refunds. This isn’t a tech issue it’s a credibility one.
Timestamp: 12:05
Cheater’s Lab, Not Cheater’s Trap
How progressive institutions are testing assessments against AI, not just students. Authenticity by design, not default.Timestamp: 24:05
Empire Strikes Back!
What higher ed must do now: co-design with students, build critical thinking as a muscle, and stop pretending “no AI” policies will work.Timestamp: 27:03

✅ Monday Morning Action Item
Run your next AI tool through a “trust filter.”Can you explain it, audit it, and defend its use if challenged? If not, rethink it. Use AI with students, not on them.

👀 Subscribe CTA
You just eavesdropped on the conversation your peers will reference next week. Subscribe now to join the circle of AI-ready educators who are shaping the future not reacting to it.


💬 Join the Conversation

SUBSCRIBE on Apple Podcasts / Spotify / YouTube—new episodes drop weekly

RATE US ⭐⭐⭐⭐⭐—help more educators discover the future of AI in higher-ed

Newsletter: Get weekly AI and Higher Education insights straight to...