Did you know that we have LinkedIn, Instagram and X accounts
that you can follow?
Hi {{first_name|everyone}}, Google thinks its smartest AI shouldn’t have to do all the work, so we’re looking at how Gemini 4 Hydrogen helps split the job between models. After that, we’ve got a local experiment that stress-tested a big AI claim, a few new tools worth bookmarking, one creation you can play with in your browser, and a quick word from our sponsor, The Flyover. It’s Friday, so the poll is up too and we want to hear where you stand, so let’s dig in.
🤿 DEEP DIVE
Image credit: Google (with edits and animation)
Google Is Splitting Up the AI Work
Gemini 4 Hydrogen brings a different approach to getting things done.
Apparently, Google doesn't think its smartest AI should have to do all the work.
And that makes sense when you think about it. Not every task needs an AI model that spends ages reasoning through a problem. Sometimes you just need something that follows instructions, gets the job done, and doesn't make you repeat yourself 5 times.
That's pretty much the idea behind Gemini 4 Hydrogen, Google's newest model.
It's designed to be fast and efficient, but Google says it's also surprisingly capable. On one intelligence benchmark, Hydrogen actually scored slightly higher than Gemini 4 Argon, one of Google's more advanced reasoning models.
But the more interesting part is what Google plans to do with it.
One AI thinks, another does the work
Google is introducing something called Orchestrator in its Antigravity coding platform.
Instead of relying on one AI model to handle an entire project, Orchestrator switches between different models depending on what needs to be done.
For example, Gemini 4 Carbon can handle the difficult decisions, like figuring out how a software system should be built. Hydrogen can then take care of writing code, running tests, and checking whether everything works.
The models can even work in parallel, with one coordinating the project while others handle different tasks.
In Google's tests, this approach achieved nearly the same task success rate as Carbon working alone, but used 24% less quota and finished 19% faster on average.
So the idea isn't necessarily to make every AI model smarter. It's to use the right model for the right part of the job.
And it's not just for developers
Hydrogen is also coming to the Gemini app and Google Search, which is where most people are more likely to encounter it.
In Gemini, Google says the model is better at following detailed instructions, remembering what you've asked throughout a conversation, and producing useful results without endless revisions.
That includes everyday tasks like writing emails, researching topics, comparing products, and working with documents.
It can also pull information from Gmail, Google Drive, and connected services like Airtable and Dropbox when those connections are available.
Google also says Hydrogen is much less likely to fabricate information than its previous Gemini 3.8 Flash model.
And in Search, the upgrade is supposed to make complicated questions easier to understand, even when they're messy, contain typos, or involve several things at once.
The part worth watching
Google is also giving Hydrogen 4 thinking levels, from low to xhigh. That means users and developers can adjust how much reasoning the model does depending on the task.
For developers, the model costs $1.50 per million input tokens and $7.50 per million output tokens.
Hydrogen is rolling out globally across the Gemini app, Search, Google AI Studio, and Google's business tools, although some integrations require eligible subscriptions.
Google isn't just trying to build one AI that does everything. It's building a way for different AIs to handle different parts of the same job.
The quick take
Gemini 4 Hydrogen is Google's new model designed for faster everyday tasks, coding, and AI workflows.
Google's Orchestrator can switch between models automatically, reducing quota usage by 24% and completion time by 19% in its reported benchmark.
Hydrogen is better at following instructions, using tools, and handling complex tasks than Gemini 3.8 Flash, according to Google's tests.
The model is rolling out across the Gemini app, Google Search, AI Studio, and business products, with some features limited by subscription.
🤝 FROM OUR FRIENDS
A quick share we thought you might find useful
Subscribe to The Flyover for FREE!
Tired of biased headlines, endless scrolling, and news that takes
forever to read?
Join nearly 3 million readers who trust The Flyover for quick, fact-focused coverage of politics, business, sports, technology, and more.
Our editorial team reviews hundreds of sources every day and delivers only the stories that matter—without the spin, fluff, or agenda.
Stay informed in just minutes each morning and start your day ahead of the curve.
By clicking below, you are subscribing to The Flyover's FREE daily newsletter and will begin receiving future editions in your inbox.
🗳️ FRIDAY POLL
Where the community stands on AI, tech, and productivity
What do you use AI for most often?
🔥 WHAT’S THE BUZZ
What people are building, sharing, and talking about
Independent developer and researcher Simon Willison tried reproducing a buzzy “weight-space programming” result on his own laptop using Qwen 3.8 27B. He fed the model raw neural-network weights and asked it to identify which game the network had been trained to play. It correctly identified Pac-Man, Pong and Breakout, though it missed Freeway.
This is a good example of an independent person stress-testing a headline AI claim instead of simply repeating it. The bigger takeaway is that some capabilities that look exclusive to giant frontier models may be reproducible locally, just slower.
⚒ TOOL SNAPSHOTS
Futuristic tools within AI, no-code, and productivity
🐛 Plurai Simulation
Create a safe sandbox that mirrors your agent setup.
Why it’s useful
It helps you catch an agent’s worst bugs using synthetic personas and documents plus mocked tools and databases. Run it from Claude Code, Codex, or GitHub Copilot without ever touching production.🎬 YouArt Desktop App
Make videos end to end using your files and apps.
Why it’s useful
It pulls local files plus context from Slack Gmail and GitHub onto one canvas and can operate After Effects CapCut and Blender. Use it for feature launch videos viral remakes or clips from long videos while routing shots to the cheapest provider with your own API keys.🎨 Adobe AI Assistant
Tell Creative Cloud what you want and get it done.
Why it’s useful
It handles multi step creative tasks across Photoshop, Premiere, Illustrator, InDesign, After Effects, and Firefly. Handy when you need to move from design to edit to motion without switching your workflow each time.
👀 ON OUR RADAR
One more thing we thought you'd like
Explore 500+ meals by culinary legends.
Done with the same leftovers? Add more variety to your week with CookUnity. Discover 500+ meals nationwide handcrafted by world-class chefs, featuring 60+ cuisines and diets. Delivered fresh and ready to heat in minutes, this is a great way to explore authentic global flavors, and the best of your local food scene at home.
🤖 AI MADE THIS
Interesting and inspiring creations made with AI
Aftertouch - An Interactive World You Can Grow and Rewind
Nicolas Chapuis worked with OpenAI Codex to create Aftertouch, an interactive world where drawing with your cursor grows colorful landscapes and produces music. You can even rewind time and explore a different creative path. The whole experience runs in your browser.
AI used
OpenAI Codex. The project's documentation credits Codex with designing, coding, scoring, rendering, and testing.
ℹ️ ABOUT US
The Intelligent Worker helps you to be more productive at work with AI, automation, no-code, and other technologies.
We like real, practical, and tangible use-cases and hate hand-wavy, theoretical, and abstract concepts that don’t drive real-world outcomes.
Our mission is to empower individuals, boost their productivity, and future-proof their careers.
We read all your comments - please provide your feedback!
Did you like today's email?
What more do you want to see in this newsletter?
