Welcome to this week's edition of Overclocked!

In this week’s edition, we’ll discuss the three major model updates from OpenAI, Google, and DeepSeek. Plus, we’ll explore the causes of ‘AI fatigue’ and how to prevent it. Let’s dig into the updates ⬇️

In today’s newsletter ↓
🔥 OpenAI's 4o gets native image generation
💡 Google Gemini Pro 2.5 gets a sharp upgrade
🧐 DeepSeek V3 leads non-reasoning benchmarks
🏠 AI fatigue is real—and we’ve got ways to manage it
💎 Build your first ‘Gem’ with Gemini

🔥 3 Huge AI Breakthroughs You Might Have Missed

OpenAI just released an update that enables ChatGPT-4o to generate images natively for the first time. Previously, it relied on DALL·E. Now, 4o can create high-quality visuals in real time directly in the chat.

What makes it wild:

• You can edit parts of an image by clicking on it.

• The model remembers what it created, so you can iterate like a designer.

• It understands light, texture, and shape far better than older models.

But that’s not all. These new image capabilities are fully integrated into conversations, meaning you can ask follow-up questions about an image, request changes in plain language, or even start a design from scratch—all without switching tabs or apps. 

For example, you could say, “Make the background dusk instead of daylight,” or “Add a shadow under the tree,” and the model will make those changes on the fly. It’s like having a design assistant that understands your creative intent instantly.

This brings generative AI a big step closer to real creative workflows, empowering artists, marketers, and hobbyists to bring ideas to life with unprecedented speed and control.

💡 Google Gemini Pro 2.5 Gets Smarter (and More Transparent)

Google has fine-tuned Gemini Pro 2.5, making it significantly better at:

• Multi-step problem solving

• Internal reflection and self-correction

• Calling and using tools more effectively

• Image creation and editing

• Coding and math

Gemini now shows how it thinks, and its internal reasoning is more transparent to users. This boosts trust—and makes it easier to spot when it's off.

Bonus: It also reads long documents faster, and integrates better across Gmail, Docs, and Search.

🧐 DeepSeek V3 Crushes the Non-Reasoning Model Benchmark

DeepSeek's new open-source V3 model is blazing fast and incredibly capable. It doesn't specialize in step-by-step reasoning like Claude or GPT, but it excels in:

• Text generation

• Coding assistance

• Instruction following

Early testing shows it topping other non-reasoning open models in several benchmarks, including MT-Bench and MMLU.

🏠 What to Do if You’re Feeling Burned Out by Constant AI Updates

It’s not just your imagination. AI fatigue is real, especially for those overwhelmed by constant model releases, new tools, and digital demands. When everything feels like a must-know update, your brain starts to tap out. Here’s how to stay sharp and keep burnout at bay without falling behind.

Recognize the Signs

How to prevent & overcome AI fatigue

• Set digital boundaries (turn off non-essential alerts)

• Schedule non-AI time in your day—go analog for a bit

• Pick just 1-2 tools to master each month

• Unsubscribe or pause newsletters that overwhelm you (yes, even ours—we’ll still love you!)

• Curate your inputs: follow only the most helpful voices, not the loudest

Use AI as an ally, not a boss. You don’t have to chase every headline or tool drop. Instead, create your own personal AI toolkit and ignore the noise. Take a breath, set a sustainable pace, and stay curious without burning out.

[Read more tips →]

The Weekly Scoop 🍦

🌍 Google adds Gemini-powered travel planning to Search and Maps

📈 OpenAI set to close $40B funding deal with SoftBank

🤔 Study links heavy ChatGPT use with increased loneliness

🛳 MIT teams up with shipping industry for AI-powered nuclear tech

🚀 Nvidia unveils next-gen Vera Rubin & Blackwell Ultra chips

💥 North Korea's new "suicide drones" spark global concern

💎 AI Challenge: Build Your Own Gemini Gem

Google’s new “Gems” feature lets anyone create custom AI tools in seconds using Gemini. Want a travel planner, joke generator, study buddy, or startup pitch assistant? It’s all possible.

Why it’s cool: Gems let you create purpose-built versions of Gemini with specific instructions, tone, and functionality.

Here’s how to build one:

1. Go to Google Gemini

2. Click "Explore Gems" or “Manage Gems”

3. Hit "Create a Gem"

4. Describe what you want it to do (e.g., “Help me practice French vocabulary every morning”). You can make it as simple or complex as possible.

5. Give it a name and test it out.

It’s like building your own personal assistant, but it only takes a minute.

That’s it for this week’s Overclocked! Did OpenAI just make image tools obsolete? Is Gemini finally catching up to GPT? Until next time!

Zoe from Overclocked