Hey, AI Enthusiast!

Welcome back to the #1 AI newsletter in the world!

Here’s what we have today:

Writing

Writing Isn’t Done Until AI Checks It

If you’ve ever had AI check something you wrote before sending it, you’re not alone.

People are running their cover letters, their work emails, and even their texts to friends past AI before they hit send.

For 1 in 3, confidence in their own writing has dropped since AI came along.

And 74% have held back something they already considered finished, so AI could check it first.

A survey of 705 US adults who write with AI asked what they check, and why.

Job applications get checked the most. 81% have run one past AI before sending it.

It goes all the way down to texting a close friend. 39% run those by AI as well.

They aren’t hunting for typos. 79% changed something after AI checked it, even when they thought it was fine the way it was.

But it’s not only young people. Gen Z leads at 85%, and two in three Baby Boomers do it as well.

The youngest take the biggest hit, though. 42% of Gen Z feel less confident writing on their own.

Most still trust their own writing. They check anyway.

The more routine the check gets, the less a first draft feels finished.

Each draft leaves the app you wrote it in. Copy it into a chatbot. Ask how it sounds. Paste the fix back.

When AI’s feedback changes a send, a third of the time people ask AI to fix the draft itself.

And the next message makes the same trip.

Our take: Checking with AI turned into part of writing, same as spellcheck. The difference is that spellcheck lives right on the page. The second opinion still lives in another tab.

Together with Wispr Flow

Fix It Where You Wrote It

That tab is where the copy-paste lives.

Wispr Flow’s Command Mode moves the fix onto the page, by voice.

Highlight what you wrote, hold a shortcut, and say what you want. It works on Mac and Windows, in the apps you already write in.

If a message sounds too casual. Say “make this more professional,” and the new version lands right there.

Turn rough notes into a bulleted list. Summarize a long reply. It writes new text as well.

Open an email, hold the shortcut, and say “reply and tell them Monday works.” The reply shows up without typing a word.

The draft never leaves the page, and the copy-paste never starts.

Switch it on in Settings, under Experimental.

Models

Gemini 4 Argon Drops

Google’s newest Gemini model, Argon, is skipping the usual launch line.

It’s going to cyber defenders first.

That’s because it’s seriously good at security work. It finds a software flaw, confirms it’s real, and patches it, all on its own.

Trusted defenders even get a version with the cyber guardrails switched off, so nothing slows them down.

And it’s already paying off.

Security firm Wiz used it to catch a critical flaw in hospital software that was exposing people’s personal info.

It also chews through much bigger jobs in one go.

Google’s also letting the government take an early look, through a voluntary program.

Everyone else gets it once the safeguards finish testing, starting with devs and Google AI Ultra subscribers.

Our take: Security teams usually play catch-up with attackers. Handing defenders the strongest model first flips that, and the fixes land before the public version does.

Investigation

The FTC Investigates Top AI Labs

The FTC wants a closer look at how the biggest AI labs keep people safe.

It’s opened an investigation into OpenAI, Anthropic, and other labs over the risks their tech poses to consumers.

In the coming weeks, the labs should get formal demands, a lot like subpoenas, to hand over internal information.

Even METR is on the list.

It’s the nonprofit that tests AI models for risks.

The timing’s hard to miss. The news broke a day after AI leaders signed a voluntary safety pledge at the White House.

The FTC isn’t trying to slow anyone down, though. An official said it won’t get in the way of U.S. leadership in AI, and the probe is still early.

Our take: A pledge is a promise. An investigation asks to see the receipts. Whatever the labs have said about safety, they now get to show their work.

Cuey In Focus

Stop Starting Over With Every AI

Weeks of chats taught your AI how you write, what you’re working on, and how you like your answers.

Then you open a different model to try it out.

It knows none of it. So you explain everything again, from the top.

Cuey makes your memory portable.

It pulls your chat history from ChatGPT, Claude, and Gemini into one memory, then carries it into whichever model you open next.

Your writing style, your projects, your favorite format. All of it comes along.

Switch models freely. Every one already knows you.

Coming in Hot

AI Tools of the Day

Executive Order

The Government Stops Saying “AI”

Federal agencies have a new word for AI. Super Intelligence.

A new executive order tells them to swap “Artificial Intelligence” and “AI” for “Super Intelligence” and “SI” in letters, websites, reports, and public statements.

And where the order applies, they’re not supposed to acknowledge the old terms at all.

The thinking goes like this.

Today’s systems do a lot more than what “Artificial Intelligence” described when the term was coined, 71 years ago.

On paper, nothing about the tech changes.

“Super Intelligence” means exactly what “Artificial Intelligence” already meant.

That part might shift. The White House science adviser has 60 days to propose a new legal definition.

Old rules, contracts, and grants keep their wording.

Our take: The tools on your screen work the same under a new name. Keep an eye on that 60-day definition, since whatever it says will shape what future federal rules cover.

Notable AIs

Notable AI Tools

🔌 Nango lets the AI agents in your product act on and pull data from 1,000+ APIs, with auth, tool calls, syncs, and triggers handled for you.

📈 CXL’s live AI cohort has marketers build AI workflows on their own customers and channels, with instructors on hand when they get stuck.

By TAAFT

AI vs. Cancer, Explained

An AI designs drug candidates on a computer before anyone touches a lab bench.

Another reads scans alongside doctors. A third helps scientists map what’s happening inside human cells.

In our latest video, we look at how AI is showing up across cancer research, and where it’s fallen short before.

Robots

Old Robots Jump Into Molten Steel

Figure had a problem.

Its older humanoid robots were retiring, and each one was packed with custom parts it wanted to keep private.

Taking them apart by hand would’ve held up its next robot.

Figure asked the internet what to do, and Arnold Schwarzenegger had the answer. Melt them.

Then he signed on to help.

Finding a furnace was the hard part. No foundry in the U.S. wanted robots with lithium-ion batteries leaping into its equipment. The only one willing was overseas.

The robots trained first.

They jumped off a second-floor building into airbags, then learned the real move in simulation, with stunt performers as the guide.

The target was a bucket of molten steel at a foundry they’d never seen. The team got 24 hours to pull it off.

Each melt gave them about 20 minutes before the steel crusted over.

The heat and electromagnetic fields knocked out cameras and electronics. The robots kept going.

The leftover metal is becoming a small run of collectibles for sale.

And Figure’s clear on one thing. In an age of AI video, this footage is real.

Our take: Under the stunt is the real story. A robot trained in a simulator nailed a move in a place it’d never been, somewhere hot enough to break the cameras filming it. And it held up.

Prompt of the Day

Open the Cold Case

Important Note: You’ll need to click the button to get the complete prompt.

That’s all for Wednesday!

Thanks for reading,

— There’s An AI For That