In partnership with

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

Here’s what’s on the menu today:

• GPT-4.5: All size, no brains🤦‍♂️

  OpenAI rolled out GPT-4.5.

• Your AI finally understands context🔎

  Google released Gemini Embedding.

  New here? Subscribe! 😎

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

MAIN COURSE

OpenAI's latest disappointment🤦‍♂️

OpenAI rolled out GPT-4.5 to ChatGPT Plus subscribers. It’s OpenAI’s largest model yet.

What's new?

OpenAI's largest model ever with enhanced conversational abilities and deeper knowledge. Aims to reduce hallucinations and improve creative tasks.

So, it’s incredible?

Nah. It’s surprisingly weak at complex reasoning and lags behind Claude and DeepSeek. Altman admits it “isn't a reasoning model and won't crush benchmarks.”

How much is it?

It’s 15 times more expensive than GPT-4o. This inevitably raised questions about who can afford to use it at scale.

What is it even good for?

It excels at persuasive conversation. It even convinced GPT-4o to donate virtual money and reveal secret codewords.

Ready to level up your work with AI?

HubSpot’s free guide to using ChatGPT at work is your new cheat code to go from working hard to hardly working

HubSpot’s guide will teach you:

• How to prompt like a pro

• How to integrate AI in your personal workflow

• Over 100+ useful prompt ideas

All in order to help you unleash the power of AI for a more efficient, impactful professional life.

Get the free guide and level up your AI game today!

SIDE SALAD

Will this fix Google?🔎

Google released Gemini Embedding. It’s a new AI model for understanding text relationships across languages.

What are embeddings?

They convert text into numbers. This helps computers understand relationships between words and concepts. They power semantic search, AI recommendations, and more.

What's improved?

• It handles 8K tokens (double previous limit).

• It creates 3K-dimensional outputs for richer representations.

• It works in 100+ languages (twice previous coverage).

• It outscores competitors by 5.81 points on key benchmarks.

Who is this for?

It’s helpful if you’re building search, recommendation, or classification systems. It works across finance, science, and legal fields without special training.

Any special features?

It allows you to shrink embeddings to save storage while maintaining accuracy.

Why do I care?

Better embeddings mean more accurate search results and smarter AI that understands context, not just keywords.

When can I use it?

Available now through Gemini’s API in experimental form with limited capacity. The full release is expected in the coming months.

YOUR DAILY MUNCH

Cool Tool 🛠️

Your Cold Emails, Delivered

Take back control over your deliverability with cold email infrastructure built specifically for large-scale cold outbound.

No matter how many emails you want to send, Infraforge has you covered with automated set up of domains and mailboxes, warm up and seamless API for unlimited integrations.

See all the features

Think Piece 🧠

Coding expertise has never been more important. 25% of companies are using AI to generate 95% of their code bases. Humans are essential to ensure security/reliability.

Startup News 💰

Pika AI introduced new video-editing features. These allow you to control specific video regions, insert AI elements, smoothly interpolate frames, and blend images.

Tavus introduced Phoenix-3, Raven, and Sparrow. These models offer immersive, natural human-AI interactions suitable for real-time applications like video calls.

Research 👨‍🔬 

Babel — a multilingual model that supports 25 major global languages. These include underserved ones like Hindi and Bengali.

MultiAgentBench — introduces a comprehensive benchmark for evaluating LLM-based multi-agent systems across collaborative and competitive scenarios.

MEME FOR DESSERT

SPOT THE AI

3 of these are real paintings. 1 is fake. 👨‍🎨

Which one is AI-generated? 👇

Ideas? Comments? Complaints?

Respond to this email or hit me up on 𝕏.

Until next time 🤖😋🧠