- AI Fire
- Posts
- 🙄 Grok 4 Costs Like It’s GPT-6
Read time: 5 minutes
Google Chrome is getting bodied… Nobody’s talking about this enough, but it might be the most strategic move in the entire AI browser war.
What are on FIRE 🔥
LEARNING PARTNER AIRCAMPUS
Create your own AI Agents 👨🏭👩🏭
You don’t need another productivity hack. You need a system that works without you.
While most people juggle tabs and chase tasks, a few are handing it all off — to AI agents that don’t assist, they act. Imagine:
Your email replies are done. 📬
Instagram posted. 📸
Calendar managed. 🗓️
40+ apps humming in the background. ⚙️🤖
All without lifting a finger. 🧘♂️
On Saturday, 12th July at 10 AM EST, we're pulling back the curtain.
It’s a Masterclass showing you how to architect AI agents that actually run your work — across 40+ tools — so you can focus on the parts only you can do.
This is your invite.
First 50 get in for free. Then the doors close.
🎟️ Save Your Seat – Before It’s Gone
AI INSIGHTS
While OpenAI gears up for GPT-5 and Google pushes Gemini 2.5 Pro, xAI just dropped Grok 4 - Elon Musk’s most powerful (and controversial) model to date. And it’s a philosophy shift.
The livestream was messy (started an hour late) and the timing couldn’t be more chaotic - xAI’s chief scientist quit the same day. But Grok 4 still launched with one goal: to be the anti-ChatGPT.
It supports:
Multimodal inputs: text, images, and soon maybe video
Grok 4 Code: a dev-focused version to write and debug code
Voice: new natural, interruption-free voice experience
Real-time web access: thanks to DeepSearch, it pulls live info from the web (especially X), right in the chat
Internet culture fluency: memes, slang, sarcasm
It will be accessible through a new $300-per-month “Pro” subscription tier. Quite expensive, for me, it's like you’re paying for a next-generation model that doesn’t even exist yet (like GPT-6).
Where OpenAI builds for reliability, and Anthropic for alignment, Grok is betting on chaos, humor, and dev power.
Why It Matters: xAI is going all-in on the idea that people want less-filtered AI, faster models with personality, sarcasm, and freedom from the “woke” defaults of ChatGPT or Gemini. It’s risky. It’s unpredictable. But it also might work.
PRESENTED BY JIRA
56% of workers say scheduling a meeting is the only way to get information. With Jira, use AI to automatically add work from Slack, create subtasks, or attach relevant resources. So instead of scheduling a meeting, check the status in Jira. Easy.
TODAY IN AI
AI HIGHLIGHTS
🔥 Here's "Viral AI-Vlogs with Veo 3" course for you to easily build ultra-realistic Bigfoot-style vlogs & grow 100K+ followers. You’ll get the exact prompts, video templates, tools... Take it here.
🎬 A new AI filmmaker is trained on fully licensed footage, you can change the object by dragging it, instead of writing a full prompt. Try it here or watch an overview here.
🌐 Perplexity finally dropped Comet, its new AI web browser, to chip away at Google’s grip on the internet. It’s now limited to Max users & early waitlisters. Take a look here.
👉 Then OpenAI is reportedly dropping its own AI-powered browser and it’s coming in just weeks. You might not even need to leave ChatGPT to browse the web anymore!
🛑 AI voice clones just impersonated U.S. Senator Marco Rubio and used his fake voice to call high-ranking officials. 5 officials got fooled by a fake Rubio in 15 seconds.
💰 AI Daily Fundraising: Microsoft will invest over $4B in AI tools, training, and cloud services to help 20M people earn AI certificates, part of a global push to bring AI into schools and careers.
AI SOURCES FROM AI FIRE
AI CHEET SHEET
NEW EMPOWERED AI TOOLS
📂 fileAI gives you structured, clean and zero-shot data from any files.
🎞️ Magic Animator animates your Figma designs in seconds with AI.
🔍 Instagram Checker detects fake followers on your account.
📝 Lazy is one shortcut to capture, chat with your notes, everywhere.
AI QUICK HITS
🗣️ GPT prompt tips to tell it straight, not just agree with anything you say.
💡 Box CEO’s tips to find your next big business idea by tapping AI agents.
📱 GPT hallucinated about an app so often, its CEO made the lie come true.
📚 Claude can now connect to learning apps like Canvas, Panopto & Wiley.
🚨 Elon Musk’s X’s CEO is stepping down, after AI goes on pro-nazi rant.
AI CHART
Google Research just dropped two new models in its MedGemma family, MedGemma 27B Multimodal and MedSigLIP, built for serious medical AI use. They’re open, multimodal, lightweight, and powerful enough to run even on a mobile.
Here’s how it works:
MedGemma 4B and 27B: Gemma 3-based models that understand text + images → for X-ray reports or EHR summaries
MedSigLIP: A lean 400M param image-text encoder trained on X-rays, histopathology, derm scans, and more
What can you actually build?
Text-to-report tools: X-ray analysis and generation that rivals task-specific SOTA models
Semantic medical search engines
Progress note summarizers with real clinical grounding
Zero-shot derm and pathology classifiers that don’t need giant datasets
Oh, and MedGemma 4B scores 64.4% on MedQA, and 27B scores 87.7%, just a few points behind DeepSeek R1, but at 1/10th the inference cost. That’s wild.
And yep, MedGemma 4B’s chest X-ray reports were rated clinically usable 81% of the time by certified radiologists. Quite production-level accuracy.
→ This is Google’s Firebase moment for health AI. They even work on mobile. Maybe, somedays, you can build Health AI without needing enterprise-level cloud infrastructure or signing a BAA!
We read your emails, comments, and poll replies daily
How would you rate today’s newsletter?Your feedback helps us create the best newsletter possible |
Hit reply and say Hello – we'd love to hear from you!
Like what you're reading? Forward it to friends, and they can sign up here.
Cheers,
The AI Fire Team
Reply