Instagram | Sponsor this newsletter

You can automate almost anything in-browser. Today we dive into how to do just that so you can start powering through your tasks.

We also have two very different AI power plays landed on the same day. One puts $5 billion behind the idea that AI can save American science. The other pits a JAMA study against the American Medical Association over who actually gets to diagnose you. Let's get into all of it.

In today's edition:

Most people know ChatGPT's desktop app has a built-in browser. Almost nobody is using it right. Here's how to actually put it to work.

💡 Pro tip: Use the annotation tool (the little pencil icon) to click and comment on specific page elements, hold Shift to select multiple at once, then Cmd+Enter to send the note without breaking your flow. It's the fastest way to give ChatGPT visual feedback on a page instead of describing it in words.

Watch the full video here →

AI agents are doing the work of full GTM teams. Pipeline, content, customer service. The startups using them aren't waiting for headcount approvals.

Get the free Practical Guide to Agentic GTM for Startups and see exactly where to start.

Unlock the Playbook

The Trump administration wants AI to power a new "golden age of science." At the center of it is a $5 billion initiative called the Genesis Mission, and this week the administration's chief science and technology adviser, Michael Kratsios, walked Congress through why.

Here's what's actually in it:

Here's the asterisk. Kratsios framed this as a revival of Bush's vision. But historian Patrick McCray at UC Santa Barbara points out a structural gap between the two. Bush's golden age was expansive: it built new institutions and flooded university research with funding.

This one is narrower, funneled through a single AI platform, and it's landing in the middle of an administration that has terminated thousands of research grants and pushed for cuts to science agency budgets.

Yale synthetic biologist Farren Isaacs is a case in point. He was told in January, with no warning, that a federal grant he was renewing simply wasn't going forward. Weeks later, the same administration is testifying to Congress about a golden age of AI-powered discovery.

Both things are true at once, and that's the tension nobody in the hearing room addressed directly.

🔮 Prediction: The Genesis Mission won't function as broad-based funding for scientists. It will function as a plumbing layer. Expect the real winners to be compute and infrastructure providers (think Nvidia, Cerebras, and the national labs themselves) who get plugged directly into the DOE's new AI platform, not individual researchers applying for grants the old-fashioned way. The administration's own "Science: A New Golden Age" report leans into this by proposing "X-labs" that chase one hard problem and dissolve, plus "golden ticket" reviewers who can fund a single unconventional idea a panel would've killed. That's a bet on concentrated, fast-moving bets over broad distributed funding, which is the opposite of what built the postwar research boom Kratsios is invoking.

Watch for a widening gap between headline numbers like $5 billion and $200 billion, and what actually reaches working scientists like Isaacs. If grant terminations keep outpacing Genesis Mission disbursements, the "golden age" framing becomes a much harder sell in the next round of congressional hearings, and I'd expect Democratic lawmakers to start using exactly that contrast as their opening line.

What do you think: does routing science funding through an AI platform make it faster, or just easier to control politically? Hit reply. I want to hear from anyone who's actually applied for a federal grant this year.

Read the full NPR story here →

A new AMA framework, provided first to Axios, insists physicians stay central to patient care as AI tools get better and better. It landed days after a JAMA opinion piece argued the opposite: that AI is already good enough to make "human in the loop" a liability, not a safeguard.

The two sides, laid out plainly:

Here's the uncomfortable part. Whyte's counterargument isn't really about accuracy. It's about definition. "Is simply performing a diagnostic test the practice of medicine? I don't think it is," he told Axios. That's a semantic hedge, and both sides seem to know it. Meanwhile, the actual regulatory groundwork nobody's arguing about is barely there. The FDA only this week put out a discussion paper on how to even assess generative AI medical devices.

Nobody has settled who's liable when an AI gets it wrong, or how insurers reimburse AI-provided care. And in Utah, autonomous AI drug prescriptions are already rolling out live, ahead of any of these guardrails.

🔮 Prediction: This turns into a reimbursement and liability fight over the next 12 to 24 months, not a capability fight. The AMA's framework reads less like a scientific position and more like a defensive claim staked on the fact that today's entire billing infrastructure, CPT codes, malpractice insurance, hospital credentialing, is built around a physician's signature.

If AI genuinely rivals doctors on five core cognitive tasks, payers have a massive financial incentive to test AI-first pathways, especially in places with thin physician supply like rural ERs, urgent care, and telehealth. Utah's autonomous prescribing pilot becomes the test case everyone points to, for better or worse.

Expect state medical boards and malpractice insurers to start writing rules faster than Congress does, which means the rules will diverge wildly state to state before anything national gets settled. The FDA's new discussion paper is the document to watch: whether they choose to regulate generative AI devices like a piece of hardware or like a licensed clinician sets the liability default for the next decade of AI in medicine.

If an AI outperformed your doctor on diagnosis, but a human still had to sign off, would that reassure you or annoy you? Reply and tell me honestly; I'm genuinely curious where the line is for you.

Read the full Axios story here →

Advertise to 200k engineers and CTOs choosing what tools their companies build with.

👉️ Learn how to sponsor this newsletter

🧬 A startup called Vivodyne opened the "world's largest human data center," growing real human tissue to fix the data problem AI drug discovery still can't solve, even Dario Amodei says curing cancer claims have become more cliché than credible.

⚖️ Lawyers defending State Farm in a fire-damage lawsuit admitted their filings were full of AI hallucinations, citing cases, quotes, and holdings that simply don't exist.

🇨🇳 A White House official says Chinese AI firm Moonshot accessed Nvidia's advanced chips through a facility in Thailand, exposing a loophole in export controls that lets Chinese companies rent compute they legally can't own.

⚡ Cerebras unveiled the CS-4, claiming up to 30x faster inference than GPU systems and over 1,000 tokens per second on models north of 10 trillion parameters.

Would love to hear your thoughts! Send me your thoughts by replying to this email (yes, I read them all :)

Until our next AI rendezvous,

Anthony | Founder of Uncover AI