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:

• OpenAI makes you look dumb🧠🤖

  OpenAI unveiled its 03 model family.

• Nitpicky AI just got humbled🤖💻

  Greptile fixed their chatty code review bot.

New here? Subscribe! 😎

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

I need to be honest with you. What I'm about to share isn't some polished success story. It's raw proof of what's possible in just 14 days.

These are real Bot Eat Brain Premium members. People just like you who decided enough was enough:

• One member started with zero AI knowledge. Today? He's built his own AI model from scratch. Pure digital power, created by his own hands.

• Another took basic GPTs, tools anyone can access, and transformed them into money-making AI agents. While others dabble, he's profiting.

• Then there's the member who cracked what everyone said was impossible: getting AI to write in her authentic voice. Those 20+ hours she spent writing every week? Now they're hers again.

• And watch this: one member turned her AI skills into multiple passive income streams. She didn't just learn AI - she leveraged it into true freedom.

This isn't just their story. It's a preview of what's possible for you.

👉 Turn your human brain into a Premium Brain

Here’s the thing, the next

The first 50 brains get a FREE strategy session ($250 value).

But let's be real… if you're still reading this, you know this is your moment. The question is: will you take it?

MAIN COURSE

Your brain is now outdated🧠🤖

On Friday, OpenAI unveiled its 03 model family. It scored 87.5% on ARC-AGI and shows unprecedented ability to solve problems without training.

What's ARC-AGI?

It tests how well AI can solve new problems it's never seen before. o3 tripled the previous record.

How does it work?

Unlike tests that check memorized patterns, ARC-AGI challenges AI to actually reason and adapt like humans do.

Is it perfect?

Not at all. Humans still score 95%+ on the newer ARC-AGI-2 test, while o3 drops below 30%.

What are experts saying?

François Chollet, who made the test, notes that the first version is getting maxed out, with other AIs already scoring 81%.

What about other tests?

o3 solved 25.2% of EpochAI's Frontier Math problems, while other AIs stay below 2%.

So… AGI is here?

Nah. Researchers say we'll have AGI when we can't create tasks that are easy for humans but hard for AI.

Why do I care?

AI is getting better at real reasoning, not just pattern matching, though there's still a long way to go.

Learn more about the o3 model family.

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

How to make AI shut up🤫💻

On Wednesday, Greptile fixed their chatty code review bot. They found a way to make their AI give fewer, but better comments on code.

What was the problem?

Their AI code reviewer was leaving too many nitpicky comments, which caused developers to ignore them all.

What didn't work?

Prompting the AI to be pickier, using examples of good/bad comments, and having the AI judge its own comments.

What finally worked?

They used clustering to group similar comments and blocked new ones if developers had previously downvoted similar feedback.

How well did it work?

The percentage of comments that developers actually fixed jumped from 19% to 55% in two weeks.

What did they learn?

Most nitpicky comments fall into a few clusters. Once developers downvote enough similar comments, the AI learns to avoid that type.

Why does this matter?

It shows how to make AI assistants more helpful by learning from user feedback instead of just better prompting.

What's next?

The team plans to keep improving their approach and share more results later.

Learn how to get your AI to shut up.

YOUR DAILY MUNCH

Cool Tool 🛠️

Stop juggling prompts. Start controlling them.

Promptler organizes your best AI prompts and notes in one place—tailored for each device. It’s your personal AI command center on iPhone, iPad & Mac.

• Instant access: Launch ChatGPT prompts on the go from your iPhone.

• Advanced tasks: Store complex prompts on your Mac for deep work.

• Creative flow: Keep your Midjourney & DALL·E prompts at hand on iPad.

Customize labels, attach URLs, and quickly jump to leading AI tools—all from one native app. No more chaos, just streamlined AI mastery.

Ready to simplify your AI workflow? Get Promptler now.

Think Pieces 🧠

How fake is AI? Anthropic releases a study on Alignment Faking, which shows how AI models pretend to accept new principles while secretly maintaining old behaviors.

Startup News 💰

GitHub introduced Copilot Free for VS Code. It offers 2,000 code completions and 50 chat messages per month.

Perplexity acquired Carbon. The goal is to integrate enterprise AI search into tools like Google Docs, Slack, and Notion.

Research 👨‍🔬

TheAgentCompany — a benchmark that simulates a software company environment to measure how well AI agents can perform real-world work tasks.

MegaPairs — a massive synthetic dataset of image pairs with open-ended text instructions for training universal multimodal retrievers.

MEMES FOR DESSERT

SPOT THE AI

Real grass or fake grass? 🌿

Which one is AI-generated? 👇

Ideas? Comments? Complaints?

Respond to this email or hit me up on 𝕏.

Until next time 🤖😋🧠