TOGETHER WITH DIGESTIBLE AI.

Hello again, human brain, and welcome back to your daily munch of AI news.

Here’s what’s on the menu today:

• Your brain = military data🪖

  DARPA announced Theory of Mind.

• Your AI needs to chill🤤

  Anthropic released notes on AI agents.

New here? Subscribe! 😎

Peep today's Spot the AI at the bottom. 👇

MAIN COURSE

Military wants to read your mind🪖

Earlier this month, DARPA announced its Theory of Mind program. It aims to build AI that can understand and predict adversary behavior.

What's DARPA?

It’s the US Defense Advanced Research Projects Agency. We covered its 12-ton, AI-powered tank in April (pictured above).

What’s the point?

Theory of Mind will create algorithms that can model how adversaries think and help plan better national security strategies.

Who's behind it?

Eric Davis, who joined DARPA in 2024 from Galois, where he worked on AI for government and intelligence agencies.

Have they tried this before?

Yupp. This is similar to past programs like:

• Total Information Awareness (2002): Aimed to spot terrorist threats

• LifeLog (2003): Wanted to track all aspects of people's lives

Sketch.

Indeed. Once developed, this tech could potentially be used beyond military purposes, like law enforcement or business.

Why do I care?

Today's AI and data tools are far more powerful than in previous attempts. This reveals a larger push to use AI to predict and influence human behavior at scale.

SPONSORED BY DIGESTIBLE AI.

Making AI simple, accessible, and digestible.

Digestible AI is a content ecosystem that simplifies complex AI trends into actionable insights, tutorials, and tools—perfect for professionals, engineers, and AI enthusiasts.

Stay informed and discover practical ways to apply AI in your world (no tech jargon required)

SIDE SALAD

Your AI is trying too hard🤤

A couple of weeks ago, Anthropic shared lessons from a year of building AI agents. The key finding? Simpler is better.

What's an AI agent?

A system where AI models can use tools and make decisions on their own to complete tasks.

What works best?

• Start simple. Don't add complexity unless needed

• Make tools easy for AI to understand

• Test thoroughly in safe environments

• Keep humans in the loop

Where are agents useful?

2 main areas show promise:

• Customer support: handling conversations and actions

• Coding: solving real software problems

What are the patterns?

• Chaining: Breaking tasks into steps

• Routing: Directing tasks to specialists

• Parallel: Running multiple AIs at once

• Orchestration: One AI managing others

What about frameworks?

Start with basic APIs instead of complex frameworks. Understanding the basics helps avoid problems.

YOUR DAILY MUNCH

Recommended Read 📖

Learn how to make AI work for you

AI won’t take your job, but a person using AI might. That’s why 800,000+ professionals read The Rundown AI – the free newsletter that keeps you updated on the latest AI news and teaches you how to use it in just 5 minutes a day.

Sign up to start learning.

Think Pieces 🧠

How does Deepseek V3 censor China-related content? It freely criticizes other world leaders, yet discussing CCP or Xi Jinping triggers propaganda or abrupt refusals.

Startup News 💰

OpenAI confirmed its for-profit shift to its corporate structure. Sam Altman says this change is essential for the company’s long-term goals and to attract more investor funding.

Research 👨‍🔬 

MMFactory — a universal framework that proposes programmatic solutions for vision-language tasks while factoring in performance and resource constraints.

Molar — integrates text and non-text modalities with ID-based collaborative filtering to enhance sequential recommendation.

MEMES FOR DESSERT

SPOT THE AI

3 of these are real giraffes. 1 is fake. 🦒

Which one is AI-generated? 👇

Ideas? Comments? Complaints?

Respond to this email or hit me up on 𝕏.

Until next time 🤖😋🧠