🍭 Grok 4.1 is HERE! 🤖

Good morning. The intern stayed up all night to crunch baseball stats using Grok 4.1.

Coffee needs to be extra strong this morning!

Let’s dig in. 👇

🍭 What’s Cookin’:

• Grok 4.1 hits #1, but real tests say “eh.”

• Google’s SIMA 2 learns new worlds fast.

• AI safety fight draws big PAC money.

• Investors rethink Nvidia’s rocketship.

• ToolBox™

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🧠 Grok 4.1
Your new go-to conversational model with sharper reasoning, better writing, and far fewer hallucinations than previous Grok releases.

📊 GoMarble AI for Meta Ads
An AI agent that audits your Meta Ads, explains what changed, why it happened, and gives you instant recommendations and ready-to-send reports.

📨 Markopolo AI
Personalized outreach at scale using real behavioral signals to generate adaptive follow-ups across email, SMS, text, and voice.

🧩 InsForge
Backend infrastructure built for AI-assisted development, giving developers APIs, auth, storage, and deployment without manual setup.

🎤 Typeless
AI-powered voice dictation that formats your speech into clean, polished writing that sounds like you actually typed it.

xAI
🤖 Grok 4.1 Tries Its Best

The Bite: Grok 4.1 just dropped, and depending on who you ask, it’s either a promising underdog or the kid in class who swears he studied but still flunks the quiz.

The big headline is LM Arena says it’s climbing the rankings.
But the real-world tests tell a different story.

Coding is slow, correctness is shaky, and creative writing lands like a group project intro nobody revised.

But then Grok flexes something no other model has:
It’s trained on the social firehose of X.
Live sentiment, live context, actual tweets, all built into the brain.

Snacks:

• Benchmarks on LM Arena… don’t quite match hands-on testing

• Coding is slow and often breaks simple tasks

• Creative writing falls behind GPT and Claude

• Grok uniquely pulls real tweets and fresh X chatter

• Only model with native access to live social sentiment

Why it Bites: Leaderboards are nice, but real usage exposes the gaps — and Grok’s still learning to walk while everyone’s sprinting.

But the X-data advantage?
That’s real.

No other model has a direct pipeline into the internet’s collective brainstem.

If it gets faster, cleaner, and stronger, Grok won’t just be another GPT alternative. It’ll be the only model that actually knows what the world is talking about.

Steal This Prompt
🗒️ Powerful Script for 30 Second Short Video

This prompt helps you craft a tight, high-engagement script for a 30-second video.

Drop in your topic, tone, audience, and preferences and the prompt will output a shoot-ready short that's built to capture attention and convert.

1. Click this Prompt

2. Paste into GPT

3. Replace the #’s with your details

4. Watch it generate a nice script

   
   🍭 Try the prompt on Snack Prompt 🍭

_{Everything Else}
🧠 You Need to Know

🗳️ a16z-Backed Super PAC Targets NY Lawmaker

→ A super PAC funded by Andreessen Horowitz, OpenAI, and other tech execs is attacking Assemblymember Alex Bores over his AI safety bill.

📉 Elite Investors Cool on Nvidia

→ Three high-profile investors have dumped Nvidia stock despite the chipmaker’s massive valuation and dominant position in AI hardware.

📈 Peec AI Raises $21M for Search Shift

→ Peec AI landed $21M as brands scramble to adapt to users swapping Google searches for ChatGPT queries.

🛡️ Meta Launches Reels Protection Tool

→ Meta rolled out a mobile tool that detects when creators’ Facebook or Instagram reels are reposted without permission.

🤖 Stack Overflow Pivots to AI Data Provider

→ Stack Overflow is transforming its Q&A platform into an AI training data service built around its MCP framework.

_Google
🎮 Google’s SIMA 2 Actually Learns the Game (Literally)

The Bite: Google dropped SIMA 2, a next-gen agent that can jump into unfamiliar 3D worlds, figure out what’s going on, and then get better the more it plays.

What is it?
A self-improving system that blends Gemini’s reasoning with hands-on control.

It’s the closest thing we’ve seen to an AI that can pick up a controller and not immediately run into a wall, which is more than some humans can say…

But AGI?
No.

It’s still doing chores in virtual playgrounds, not filing your taxes.

Snacks:

• Learns from never-seen-before environments

• Operates across multiple 3D games, not just one

• Uses Gemini for planning and decision-making

• Improves its own performance through repetition

• Built as a stepping stone toward more capable embodied agents

Why it bites: SIMA 2 isn’t intelligence, it’s progress; the slow, necessary kind.

If agents are ever going to work outside demos, they need to generalize, adapt, and stop acting like goldfish in a maze.

Google’s finally nudging in that direction.

Playful AGI talk aside, the real win is simple.
Agents that learn by doing tend to outperform agents that memorize the brochure.