This week, Uncle Sam turned 250…and decided he wants to get in on the AI game.

Honestly, I’m a little torn on it.

On one hand, AI is quickly becoming national security infrastructure. These models will shape defense, intelligence, education, and the flow of information itself. So yes, it makes sense that the government wants a seat at the table.

But then comes the uncomfortable part: the government is getting involved in Silicon Valley. Like, really involved.

We know the Trump administration held back Anthropic’s Fable model release for weeks. OpenAI has now offered the U.S. government a 5% equity stake to hedge against similar action (yes, actually…).

And then this week, Palantir and NVIDIA announced they are teaming up to build custom AI models for the U.S. government itself.

And all of this has me wondering: what precedent does this set?

Silicon Valley has flourished because builders have had room to build — without fearing extreme oversight or the government picking favorites.

But AI is, maybe rightfully, changing that equation.

So the question I keep coming back to is this: are we entering a new era of technology where the government is not just regulating innovation from the sidelines, but actively shaping it?

And if so, where is the line between national security and government overreach?

Something to brew on ☕️

1. US government lifts restrictions on Anthropic’s Fable & Mythos models after weeks of scrutiny

My Tea: The models are back online, but the precedent is clear: the government has very much entered the AI chat.

2. Amazon and Microsoft bet billions on forward-deployed engineering units

My Tea: The next frontier for enterprise AI won’t be the technology — it’ll be implementation. See my video deep dive on AI’s latest trend.

3. OpenAI proposes giving the U.S. government a 5% stake in the company

My Tea: Woah. When did frontier AI go from Silicon Valley innovation to a potential national asset class?

4. NVIDIA and Palantir team up to build custom AI models for the U.S. government

My Tea: If it wasn’t obvious by now, AI has the government’s attention — not just as a regulator, but now, as a player.

5. Meta explores a cloud business to sell its excess AI computing capacity

My Tea: It’s giving…we went too hard on AI data center builds and now need to recoup costs. But hey, the stock is up on the news.

Silicon Valley’s latest buzzword is taste.

And despite what you may think, this one has everything to do with AI…and nothing to do with AI at all.

Because we are officially in a world where AI commoditizes knowledge, busy work, and first drafts. So the thing that starts to matter more is knowing what’s actually good.

Taste is not just aesthetic. It’s judgment. Restraint. It’s understanding the culture, the customer, the moment, the mood.

And that’s something you can’t train a model on. Taste isn’t hackable.

What makes taste so interesting to me is that the AI conversation has been so tethered to the exponentially scaling capabilities of the tech. 

How interesting, then, to talk about a deficiency. 

To reveal the limits of the technology — and in contrast, the very nature of our humanity.

My coverage of taste below:

We hold these prompts to be self-evident. Google ran a Fourth of July ad this week reimagining the Founding Fathers drafting the Declaration of Independence with Gemini and Google Workspace. And honestly? Bravo.

If you think AI hype is bad in Silicon Valley, I regret to inform you it has also spread to New York (well, that’s what someone from Jersey would say, anyway)

Jersey Mike’s filed for IPO this week and mentioned AI 22 times in its filing. I love the tech pivot for them, but LBH… I’m just waiting for Optimus to make my sandwich (scooped sourdough, extra turkey, dijon) — get on it, Elon.

Hope you had a restful Fourth. As one might imagine, I think of the Boston Tea Party often 🫖

Feeling a spark of inspiration from our forefathers today:

No model regulation without Tea-sipper representation. Equal weights and open source for all. -M