🍭 Massive Model Money Moves 💸

Good morning. If your chatbot starts acting distant, it’s not you. It’s scaling.
Claude might be seeing other models.

Let’s dive in 👇

🍭 What’s Cookin’:

• AI spending hits Manhattan Project levels

• Anthropic says someone tried to steal Claude

• Google drops Gemini 3.1 Pro

Curiosity
🏗️ AI Spending Is Manhattan Project-Level Big

The Bite:

AI spending is entering mega-project territory.

Governments and tech giants are pouring hundreds of billions into AI infrastructure at levels that rival some of the largest public projects in modern history.

When you adjust for inflation, today’s AI buildout is already competing with the Manhattan Project, the Apollo Program, and the Interstate Highway System.

And we’re still early.

Snacks:

• The Manhattan Project cost ~$2B in the 1940s (~$30B+ today)

• The Apollo Program cost ~$25B in the 1960s (~$250B today)

• The U.S. Interstate Highway System cost ~$500B+ in today’s dollars

• Private tech capex is now rivaling historic wartime-scale public spending

• AI infrastructure investments are projected in the hundreds of billions annually

Why it Bites:

This isn’t a product cycle. It’s an infrastructure era.

The Manhattan Project built the atomic age.
Apollo Program built the space age.

The Interstate Highway System reshaped commerce, cities, and culture for decades.

AI is now getting that level of capital commitment.
Not as a one-off mission, but as a permanent layer of global infrastructure.

The difference?

Those projects were single national efforts.

AI is a continuous, global buildout.
Funded by governments and corporations.
Happening in real time.

We’re not watching a tech trend.

We’re living inside the construction phase of the AI era.

And history books rarely announce themselves while they’re being written.

Steal This Prompt
🎨 Cartoon Stickers

Turns any idea, character, or photo into a bold, clean, sticker-pack-ready cartoon with punchy outlines and expressive personality.

Use it to:

• Create custom Slack / Discord emoji packs

• Turn your team into branded mascot stickers

• Design merch-ready sticker sheets in minutes

Workflow:

1. Click this link (Prompt)

2. Paste into Midjourney / any image model

3. Replace the #s with your subject, style, and expressions

4. Watch it cook: instant sticker pack energy

   Try the prompt on Snack Prompt

Anthropic
🔫 Says Someone Tried To Steal Claude

The Bite:

Anthropic says it caught attempts to “distill” its Claude models.

This means someone allegedly queried Claude at scale to extract its intelligence and train a competing model.

The company claims it identified and blocked these large-scale extraction attempts using new detection systems.

Anthropic frames this as a serious threat to model integrity and intellectual property.

But here’s the wrinkle: model distillation is also a totally normal AI training technique.

And almost every major model was trained on massive amounts of scraped internet data.

So… what exactly counts as stealing here?

Snacks:

• Distillation = training a smaller model on outputs from a stronger model

• Anthropic says it detected “coordinated” extraction attempts

• They built new systems to flag unusual, large-scale query behavior

• Distillation is widely used inside AI labs for efficiency

• Most frontier models were trained on scraped web data

Why it Bites:

This isn’t a hacking story. It’s an AI industry boundary dispute.

Anthropic is drawing a line:
Using our model outputs at scale to train yours crosses it.

But here’s the tension: large AI labs trained on oceans of internet data, often without explicit permission from creators.

Now they’re saying their own outputs are protected territory.

That’s where it gets uncomfortable.

If scraping the web to train a model is innovation, is scraping a model to train a model theft?

Anthropic argues there’s a difference between public data and proprietary systems. Fair.

Critics argue this is just the next phase of AI eating itself.

The real fight isn’t about distillation.

It’s about who gets to own intelligence once it’s been compressed into a model.

_{Everything Else}
🧠 You Need to Know

🏗️ How AI Spending Compares to History’s Mega-Projects
→ New analysis compares current AI investment levels to historic mega-projects like the Manhattan Project and the Apollo program, highlighting the unprecedented scale of private-sector AI funding.

💸 AI Dominates Capital Allocation as $50M Funding Looks Tiny
→ AI startups continue capturing the majority of venture funding, while $50M rounds now pale compared to 2021’s mega-deal era.

🧠 Google Launches Gemini 3.1 Pro
→ Google introduced Gemini 3.1 Pro, highlighting improved reasoning, coding performance, and multimodal capabilities.

🎓 USP Medical School Gets Global AI Recognition
→ The University of São Paulo’s medical school received international recognition for integrating AI into medical training and education programs.

🛡️ Anthropic Details Defense Against Model Distillation Attacks
→ Anthropic published new research outlining techniques to detect and prevent distillation attacks aimed at extracting proprietary model knowledge.