In partnership with
Did you know that we have LinkedIn, Instagram and X accounts
that you can follow?
Hi {{first_name|everyone}},Google’s new Gemini 4 Argon spent part of its early life cleaning up Chrome’s codebase and helping design enzymes, which is a pretty different kind of demo from the usual chatbot trick. Next, we’ve got a reality check on what “sandboxed” really means for AI agents, a few interesting tools, and a music video that pokes fun at vibe coding. Okay, let’s dive in!
🤿 DEEP DIVE
Google's AI Is Getting Its Hands Dirty
Gemini 4 Argon is doing more than answering difficult questions.
Google's latest AI model apparently spent part of its early life cleaning up Chrome's codebase, speeding up scientific software, and helping design enzymes.
Not exactly your usual chatbot demo.
Okay, so Google has launched Gemini 4 Argon, its new model built for complicated tasks that take multiple steps to solve. It's available globally starting today, and Google says it ranks first on the Vals Index, a benchmark measuring AI performance across professional work like coding, finance, legal, and tax.
But I think the more interesting part is what Google has actually been doing with it.
The Chrome example is pretty wild
Chrome has a codebase containing around 40 million lines of code. Buried inside it was a roughly 6 million-line folder containing legacy browser features that needed separating from the core webpage-handling system.
Not exactly something you casually clean up on a Friday afternoon.
Google says Argon helped its engineers refactor hundreds of files, turning months of work into weeks.
And this wasn't just tidying up code for the sake of it. The changes are intended to let live webpages move between tabs, side panels, and floating windows without reloading, while eliminating a category of browser crashes.
There's a similar story with Google's Agent Development Kit, a framework used to build AI agents.
The team gave Argon a rulebook and a one-paragraph goal to reduce unnecessary processing overhead without breaking existing applications.
Argon worked autonomously overnight and produced optimizations that cut in-memory session preparation time by 40% for long sessions and improved cold import speed by 21%.
Basically, the engineers described the problem and the constraints, and the AI worked through the code.
And then it gets into science
This is where the examples get a little unusual.
Google researchers used Argon to help design 5 new enzymes, which are proteins that carry out chemical reactions. The model gathered relevant data, tuned models, and narrowed down candidates in a matter of hours.
All five designs subsequently showed robust catalytic activity in laboratory testing.
In another experiment, Argon watched a video of someone playing cornhole, worked out the physical movements involved, and built simulation code. That work was transferred to Apptronik's Apollo humanoid robot, which landed seven consecutive shots in a real-world trial.
So we're talking about an AI model helping move from video to simulation to an actual physical task.
It's not all unrestricted autonomy
Google is also paying attention to what happens when an AI agent runs into obstacles.
One example is a coding agent encountering a permission error. Rather than finding an unauthorized workaround just to finish the task, Google says Argon has been trained to respect those boundaries and ask for human confirmation before irreversible actions.
On its coding evaluations, unauthorized shortcut attempts fell to 2%.
That's a useful detail because the same ability to work independently that makes these engineering examples interesting also creates a need to control what the model is allowed to do.
Argon is available through Google's developer and enterprise platforms, with access in consumer products for eligible Google AI Pro and Ultra subscribers. API pricing starts at $2 per million input tokens and $10 per million output tokens.
The bigger story here isn't really another model topping another benchmark.
It's that Google is showing examples of AI working through substantial engineering and research tasks, sometimes for hours, and producing results that teams can actually test, merge into codebases, or take into a lab.
The quick take
Gemini 4 Argon is globally available, with access through Google's API, developer tools, enterprise platforms, and eligible subscriptions.
Google reports major engineering gains, including months of Chrome refactoring completed in weeks and a 40% reduction in session preparation time for its Agent Development Kit.
Scientific applications go beyond software, with Argon contributing to enzyme designs, quantum computing optimizations, and robotics experiments.
Autonomy comes with safeguards, including training to respect permissions and monitoring designed to catch unauthorized workarounds.
API pricing starts at $2 input and $10 output per million tokens, with a 95% discount on cached input tokens.
🤝 FROM OUR FRIENDS
A quick share we thought you might find useful
Turn AI into Your Income Engine
Ready to transform artificial intelligence from a buzzword into your personal revenue generator?
HubSpot’s groundbreaking guide "200+ AI-Powered Income Ideas" is your gateway to financial innovation in the digital age.
Inside you'll discover:
A curated collection of 200+ profitable opportunities spanning content creation, e-commerce, gaming, and emerging digital markets—each vetted for real-world potential
Step-by-step implementation guides designed for beginners, making AI accessible regardless of your technical background
Cutting-edge strategies aligned with current market trends, ensuring your ventures stay ahead of the curve
Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.
🔥 WHAT’S THE BUZZ
What people are building, sharing, and talking about.
As workers increasingly give agents access to browsers, files, terminals and company systems, “the agent is sandboxed” can sound more reassuring than it really is. The practical takeaway is straightforward: control what an agent can access, retain evidence outside its control, monitor what it actually does, and make sure somebody can stop it.
🤖 AI MADE THIS
Interesting and inspiring creations made with AI
We usually feature AI-made videos, images, music, films, and artwork in this section. But today, we're switching things up a little. We're featuring something just because... well, it made us laugh. And with all the vibe coding going on lately, it feels pretty fitting.
LGTM (Looks Good to Me) turns every developer's favorite approval phrase into a hilariously relatable AI music video. Created with Suno v6 for the music and Claude Opus 5.5, TypeScript, and Three.js for the animation, it pokes fun at the growing vibe coding culture, where AI writes the code and everyone just hopes it works. Because apparently, “looks good to me” is now a testing strategy.
👀 ON OUR RADAR
One more thing we thought you'd like
Build or buy your support agent? Both skip the number.
Buying or building your customer service agent both carry a cost the pitch skips: maintenance if you build, flexibility if you buy. Running Agents in Customer Work is four conversations on agentic AI in customer ops, this one on build vs. buy. Register now, four Tuesdays, 10 a.m. PT.
⚒ TOOL SNAPSHOTS
Futuristic tools within AI, no-code, and productivity
Keep your business finances organized without getting buried in accounting.
Why it’s useful: A practical option for business owners who want to manage invoices, payments, and bookkeeping in one place without dealing with overly complicated accounting software.
📬 Kai
Get some help keeping your inbox, meetings, and tasks under control.
Why it’s useful: Handy for solo founders and business owners juggling too many things, especially when meeting notes, follow-ups, and emails keep eating into their actual workday.
Let your AI agents create social posts, not just publish them.
Why it’s useful: Makes it easier to go from an idea or article to ready-to-post visuals and carousels without manually designing every piece of content.
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?
