GritAI Newsletter: OpenAI's New Image Model Is Insane, Google's Gemini 2.5 Dominates, and OpenAI Adopts MCP


Hey 👋

This week, OpenAI practically broke the internet with its new image generation capabilities—turning simple prompts into insane photos, illustrations, and pixel-perfect designs. But what really blew people away was the model’s near-perfect text output (see examples).

Another major headline: the launch of Google Gemini 2.5 Pro, now widely considered the best large language model—period. It tops most benchmarks and is being called "an absolute beast" by those who've put it to work. The AI race is heating up—again!

And finally, it looks like MCP is here to stay. OpenAI has now adopted Anthropic’s standard for connecting AI models to data, signaling a broader industry shift.

But that was not the only AI news this week. Let’s jump in!

AI Pulse 🚨

A bite-sized curation of this week's most important AI news.

🚀Google dropped Gemini 2.5 Pro, its smartest model yet—ranking #1 on LMArena with top-tier coding and reasoning scores. It introduces a new “thinking model” structure that boosts problem-solving accuracy, and supports a massive 1M-token context window—doubling to 2M soon. 🤯

🖼️ OpenAI replaces DALL·E with native image generation in GPT-4o, delivering visuals that are sharper, more prompt-aware, and better at rendering text. Multi-turn editing is now built-in, and it's live for all ChatGPT users—including the free tier.

💰OpenAI expects revenue to triple to $12.7B this year (up from $3.7B in 2024), fueled by 2M paid corporate users. Some advanced products may cost thousands per month, but profitability isn’t expected until 2029.

🧠 DeepSeek released its upgraded V3-0324 model with major improvements in reasoning, code generation, and Chinese language tasks.

🔄 OpenAI’s Agent SDK now supports MCP (Model Context Protocol). This unlocks a wide range of tools for agents, with support for the OpenAI API and ChatGPT desktop app on the way.

🔍 Perplexity adds structured answers and verticals for travel, shopping, jobs, images, and videos. The new answer modes come with built-in commerce capabilities and a more visual, card-style experience.

🤖 Qwen (Alibaba) launches Qwen2.5-Omni, a multimodal model with a “Thinker-Talker” architecture—enabling seamless voice and video chat. A strong contender in the multimodal race.

🔊 OpenAI rolls out next-gen audio models, including Voice Engine—able to mimic realistic voices from just a short sample, perfect for apps that need lifelike speech generation.

💻 Zapier adds MCP integration, letting users create AI agents that plug directly into Slack, Google Calendar, Jira, and more.

🔐 Microsoft expands its Security Copilot suite with 11 new AI agents, aimed at easing cybersecurity workloads—tackling phishing detection, incident response, and regulatory notice drafting.

📉 College graduate unemployment is up 30% since 2022, with degree-based hiring slowing faster than other sectors—a signal that knowledge work might be entering a new era.

How to Vibe Code a Voice AI Agent Dashboard

"Can we add realistic voice to our AI police dashboard without writing a single line of code ourselves?”

Click to watch the full tutorial 👇

Worth Reading and Listening to

The Top 100 GenAI Products, Ranked and Explained 📱

This A16Z podcast dives deep into the 2025 edition of the GenAI 100 list—an annual ranking of the most widely used and highest-grossing generative AI products across web and mobile.

"The app store is going to be chaos. We're going to need an AI just to solve the AI app management problem.”

Why Every Nation Needs Its Own AI Strategy 🌍

NVIDIA CEO Jensen Huang and Mistral AI CEO Arthur Mensch make the argument for Sovereign AI: that every nation must develop its own AI infrastructure to protect cultural identity, economic autonomy, and technological competitiveness.

“Your country’s digital intelligence is not likely something you want to outsource.”

Trending on Product Hunt 📱

Here are the hottest AI tools trending on Product Hunt this week

  • 🛠️ Base44 2.0 → Idea to app in minutes similar to Lovable and Bolt
  • 🤖 SimplAI → Build and deploy Agentic AI
  • 🎥 SuperCut → Record your screen, share instantly, look like a PRO
  • ✍️ Effie → Effortless AI Writing, Endless Inspiration
  • 🔌 Zapier MCP → Connect your AI to any app with Zapier MCP
  • 🔗 ActionKit → 1000+ integration actions to your AI agent
  • 🎨 Scene 2.0 → Modular AI Web Builder
  • 💻 DeepSeek-V3-0324 → Big update to DeepSeek's V3 model
  • 🎬 FLORA → Cool AI Canvas for Generative Image and Video generation
  • 🧠 Gemini 2.5 → Google’s most intelligent AI model
  • 🖼️ OpenAI 4o Image Generation → OpenAI most advanced image generator

More GritAI videos 🔥

  • 📹 These ChatGPT Features Save Me HOURS (link)
  • 📹 How to Vibe Code a Conversational AI Dashboard (link)
  • 📹 I Built a Real-Time AI Police Tracker with Cursor AI (link)
  • 📹 Getting Started with Claude (link)
  • 📹 10x Your Research with Perplexity AI (link)
  • 📹 5 Powerful AI Tools for Beginners: Your 2025 Guide (link)
  • 📹 Write AI Prompts Like a Pro - From Beginner to Expert in 5 Minutes (link)
  • 📹 5 AI Predictions for 2025 (link)
  • 📹 Google AI Studio Will Change How You See The World (link)
  • 📹 This Anthropic Console Trick Saves Me 3 Hours Every Day (link)

Did someone forward this to you?

Click here to sign up to our newsletter​

Have questions? Hit reply to this email and we'll help out!

113 Cherry St #92768, Seattle, WA 98104-2205
Unsubscribe · Preferences

GritAI Newsletter

Weekly AI news and content on how to make AI work for you 💪

Read more from GritAI Newsletter

Hey 👋 This week, the "vibe coding" hype showed no signs of slowing down. Google rolled out an updated version of Gemini 2.5 Pro—its most capable model yet for coding apps—ranking first on the WebDev Arena Leaderboard. Rumors also surfaced that OpenAI is acquiring Windsurf, a leading AI coding tool, hinting at deeper integration of native code capabilities inside ChatGPT. At the same time, Cursor, now arguably the most popular AI coding tool, announced a massive funding round—reaching a $9B...

Hey 👋 This week Meta released a new AI assistant powered by Llama 4 that uses your Facebook, Instagram, and WhatsApp data to give more personalized responses. With nearly 1 billion people already using Meta AI each month across its apps, Meta is a strong contender in the AI consumer market competing with ChatGPT and Google Gemini — that is, if users are willing to share their social media data with AI. OpenAI had to remove their latest GPT-4o update because users found the AI was being too...

Hey 👋 This week, OpenAI made its new image model available for developers “gpt-image-1”, unlocking thousands of exciting startup use cases. Big names like Adobe, Figma, and others have already integrated it into their offerings, and we’re expecting a wave of new apps to build on top of these image generation capabilities. Elon Musk’s xAI also made moves: the Grok 3 family of models is now available via API, priced competitively compared to other reasoning models. On top of that, the Grok...