If you’ve been following me, you know I recently had a story go viral about my high school, The Harker School, launching a venture fund where the students are the VCs.

This week, I went back to campus in San Jose to meet the head of the program and talk to three of the students.

And honestly? I was blown away.

My Tea: the kids are alright.

To be part of Harker’s Venture Initiative, students take Honors Venture Capital and Advanced Finance. They learn how to analyze term sheets, evaluate cap tables, talk to founders, and actually make investment decisions—with guidance from teachers, advisors, and industry experts.

TBH, these kids may already be more qualified than half the VCs out there (sorry not sorry to my VC readers 😉 )

The other fascinating takeaway was how students are using AI in school today. Two of the students I spoke with are now college freshmen, and they told me professors are now designing exams with AI in mind.

Some exams are built so ChatGPT, Claude, and Gemini may all give you different answers. In calculus, for example, the problems are getting harder than previous years because professors know students have AI support.

Here’s what struck me: despite what everyone thinks, AI isn’t just helping students study…or cheat. It’s actually changing what they are capable of learning.

It’s raising the bar for education itself.

How’s that for some Tea? 

My interview with the students launches next week. You’ll be the first to know.

1. Apple raises MacBook and iPad prices as AI memory costs surge

My Tea: Welcome to chipflation. AI demand is making the hardware behind it more expensive—and consumers are starting to feel the pain. 

2. OpenAI plans to delay its IPO to 2027 to hit a $1T valuation

My Tea: A 2026 IPO would mean a lower valuation than Anthropic. And Sam’s not showing up unless he’s the prettiest girl at the party. 

3. Anthropic gets government approval to release its Mythos 5 model to limited group of businesses

My Tea: The AI model–government regulation dance is reaching a crescendo. Side note: President Trump called Dario a weirdo, and I'm just so here for the Tea.

4. OpenAI unveils Jalapeño, its first custom AI chip built with Broadcom

My Tea: With compute constraints hitting everyone in AI, this is OpenAI's move toward vertical integration—and a hedge against its dependence on NVIDIA.

5. Europe proposes new tech sovereignty laws to build its own AI, cloud, and chip infrastructure

My Tea: Countries no longer want to depend on US tech giants for their digital future. AI is becoming national infrastructure.

Season 1 of my Linkedin show “Innovation in Action”, hosted in partnership with IBM, came to a close this week.

We covered everything from management theory in an AI world to lessons from my own career that feel even more relevant today.

Below is a fan favorite from the season, on a management framework I learned at Harvard Business School that feels especially relevant in an AI world.

A photo dump from last week in New York: visits to LinkedIn and IBM, plus a special event with Abridge’s CEO Dr. Shiv Rao (more Tea on that this week).

I spoke with Pega Systems’ Head of Product Marketing and Strategy about tokenmaxxing, AI cost management, and one of the biggest questions in enterprise AI right now: how do you scale AI without letting the costs spiral?

As if Anthropic hasn’t had enough drama lately, China has entered the chat.

Anthropic has accused Alibaba of using nearly 25,000 fake accounts and 30 million Claude interactions to extract and replicate Claude’s capabilities.

The technique is called model distillation: essentially, when one AI model learns from another AI model’s outputs.

And this is one of the reasons the most advanced models are becoming a national security concern. You can restrict access. You can limit releases. But if people can still query a model at scale, the model can become an involuntary teacher.

My Tea: this is the real loophole in the AI race. The government can try to control chips, model weights, and access, but intelligence is getting easier to compress, copy, and redistribute.

And China has historically been very, very good at learning fast, catching up, and making whatever it is, cheaper.

Personally, as a Silicon Valley girlie, I am a believer in open source. But that’s also because I think the better strategy is to gatekeep what can’t be easily replicated: distribution, trust, brand, talent, infrastructure…and taste.

More on taste…Silicon Valley’s new buzzword…next week.

—

Today’s Tech Tea was written from a Silicon Valley coffee shop: founders to my left, investors to the right, Stanford students in front. As much as I rolled my eyes at all the AI buzzwords floating in the air…god, I love it here.

Strawberry matcha helped. -M