Chrome adds a shortcut for AI workflows

Chrome starts saving how you use AI

Most people repeat the same prompts.

Google just added “Skills” to Chrome, letting users save AI prompts and run them across pages with a click. The feature builds on Gemini in Chrome and works by storing prompts from chat history, then triggering them with a slash command or menu.

This is Google trying to lock in behavior, not just usage, and it matters because habits are harder to switch than tools once they get baked into daily browsing. It also quietly shifts Chrome from a passive browser into something closer to a working layer that follows you across tabs and tasks.

Early uses like recipe tweaks, shopping comparisons, and macro tracking sound basic, but they show how quickly repetitive tasks turn into default workflows when the setup cost drops to one click. I can picture this late at night, multiple tabs open, reusing the same prompt without thinking.

If Chrome owns your saved workflows, switching browsers starts to feel heavier.

What happens when those workflows start to overlap with ads or recommendations?

Anthropic takes its most powerful model to Washington

They briefed the government before releasing anything.

Anthropic confirmed it showed its Mythos model to the Trump administration, even while suing the Department of Defense over access and being labeled a supply-chain risk. The model is withheld from public release because of its cybersecurity capabilities, and banks like JPMorgan and Goldman Sachs are already being encouraged to test it.

This tells you these models are being treated less like products and more like sensitive infrastructure, with early access going to governments and major institutions instead of developers. The tension is obvious, where companies want to cooperate on national security but draw a hard line on how far that access goes.

There is also a quieter signal in the background, with early signs of weaker job prospects for new graduates already being tracked inside Anthropic. I imagine a conference room with officials and executives, phones face down, discussing tools that are not yet public but already in use.

The rollout path for AI is starting to look more like defense tech than software.

How long before public access becomes the secondary concern?

AI labs are now competing on cyber defense

This turned into a race in one week.

OpenAI introduced GPT-5.4-Cyber, a version of its latest model tuned for defensive cybersecurity, days after Anthropic revealed Mythos and said it had already found thousands of vulnerabilities across software systems.

The speed matters more than the feature set here, because both companies are signaling that cybersecurity is now a primary battleground for advanced models, not a side use case. Limiting access to vetted researchers and expanding tiered programs shows they know how sensitive this is and how easily it could go wrong.

The early results are already concrete, with Anthropic claiming large-scale vulnerability discovery, which suggests these systems are moving beyond theory into active infrastructure testing. I picture security teams running scans late at night, watching logs update in real time as issues surface faster than before.

If this works, software security could shift from slow audits to continuous AI-driven probing.

Who decides how far these models are allowed to go?

Prohuman team

Free Guides

Explore our free guides and products to get into AI and master it.

All of them are free to access and would stay free for you.

Feeling generous?

You know someone who loves breakthroughs as much as you do.

Share The Prohuman it’s how smart people stay one update ahead.