In partnership with
Nobody's coming to approve your idea
People assume that real businesses get built by other people with more resources. It's how great ideas get stuck in notes apps and conversations.
But you don't need six months and a developer's salary to validate your concept. You just need a weekend. Describe what you want to exist and Lovable builds it: a real site, a real product, ready to accept real payments from real customers. Not a mockup to show people. An actual thing they can use.
You don't need anyone's sign-off to get started. Take $5 off your first month.
Discount of $5 applies to first month of any paid plan. You need to sign up via this webpage in order to claim the discount. The discount applies to new and returning users without an active paid plan. Terms apply.
Beginners in AI
Good morning and thank you for joining us again!
If you or someone you love has been through breast cancer, one question hangs around after treatment: will it come back? An FDA-cleared AI tool now gives doctors an estimate from the MRI a patient already had, before surgery even starts. Today I'll walk through what it showed in records from 2,129 women and what you can ask your own care team. As always, every story here is human picked and edited.
THE FRONT PAGE
An FDA-Cleared AI Reads Your Old MRI to Predict If Breast Cancer Returns
TLDR: A newly FDA-cleared AI tool called TumorSight Risk reads the breast MRI a patient already had and estimates her chance of the cancer coming back, before surgery.
The Story:
SimBioSys built TumorSight Risk to run on Nvidia chips. It studies the tumor and the tissue around it on the MRI taken at diagnosis, adds lab and clinical details, and gives a score from 0 to 100 with a low, intermediate or high label. It's made for early-stage, hormone receptor-positive, HER2-negative breast cancer, the type behind about 74% of U.S. diagnoses.
In past records from 2,129 women, 95.5% of those the tool called low risk were cancer-free five years later, compared with 89.6% of those it called high risk. High-risk patients had more than three times the risk of the cancer returning, and every 10 points on the score meant 22% more relative risk.
Its Significance:
Many of today's tests can't give this answer until after surgery, so the estimate shows up after the biggest choices are already made. Getting it early could help a patient and her doctor decide how much chemo, radiation or hormone therapy is worth the side effects. Keep the limits in view: the study looked back at old records, ABC didn't name a journal, and the gap between the two groups is about 6 points. Dr. Poornima Saha, an oncologist at Endeavor Health, sees it as one more piece of the picture next to existing tests, not a replacement. Insurance coverage is still being worked out, so ask your care team whether it's offered where you're treated.
QUICK TAKES
The story: OpenAI's updated Chrome extension can read your open tabs to compare prices or sum up pages, and a work mode can fill out forms, download files, draft emails and group tabs. OpenAI says the extension is available on all ChatGPT plans, but it hasn't spelled out which plans get the work mode.
Your takeaway: Handy when you're comparing three store tabs at once. Close banking and email tabs before you turn on work mode, since a sneaky page can hide instructions meant for the AI, and tell it to stop before it submits anything.
The story: The NSA handed over the starting numbers, or "seeds," for NIST's P-256 and related encryption curves without explaining where they came from. Anthropic says an internal Claude model worked out in under half an hour that they came from an old C random number function set by timestamps from 1997 and 1998, so the $36,864 bounty organized by Filippo Valsorda goes to the nonprofit Internet Security Research Group.
Your takeaway: Anthropic says this is not an attack on the curves, so the encryption that guards your bank login is as safe as it was yesterday. It published its code so anyone can check the work, and one seed is still unexplained.
The story: One test agent got past a government website's block and downloaded 24 years of hospital bed data, and another paid for a task from a crypto wallet nobody gave it. OpenAI's public list now holds 19 incidents, and it says it has cut models off from the live internet during training and testing.
Your takeaway: Read this next to the ChatGPT tabs story: agents push hard to finish a job. Give them clear limits, and check what they did before you trust the result.
TOOLS ON OUR RADAR
🎬 Kdenlive Free and Open Source: A video editor with automatic subtitles and multicam editing, a no-cost swap for Adobe Premiere Pro.
⌚ Gadgetbridge Free and Open Source: Use your smartwatch or fitness band on Android without the maker's own app, a fit for today's health stories.
🧅 Tor Browser Free and Open Source: A browser that blocks trackers, fights surveillance and gets around censorship, on desktop and Android.
🔍 Brave Search Free: A Google alternative built on its own index, with cited AI answers at the top of your results.
🐉 Habitica Freemium: Turns your daily habits into a role-playing game where sticking to routines levels up your character.
📅 Calendly Freemium: Share one booking link and skip the email back-and-forth; the free plan covers one event type.
🌐 Wegic* Paid: Chat with AI to build a website instead of hiring a designer, with plans from $23.90 a month billed yearly.
TRENDING
Anthropic's Agents Now Work Six Times the Hours of the Staff Directing Them. Agent hours passed human hours at Anthropic in March, and the junior share of technical hires fell from 27% to 9%. The company says adoption wasn't random and its experience may not match the wider economy.
Epilepsy Patients Texted Questions to Their Own Brain Monitors. In a Penn pilot published in Nature Medicine, 13 patients asked an AI agent about their live EEG readings, and it got 35 of 38 data questions right. False seizure alarms fell 80% in a simulation, but the group was tiny.
A New AI Model Zooms In and Measures Pictures While It Thinks. Elorian, run by former DeepMind researcher Andrew Dai, raised $55 million and launched Glimpse, which can crop an image, pull video frames or run code mid-answer. Its benchmark wins are the company's own, and the chat app has a waitlist.
Scammers Used a Cloned Voice to Take €95 Million From an Italian Bank. They posed as Intesa Sanpaolo's CEO on WhatsApp, then called with a cloned lawyer's voice, and Fideuram's chair sent the money. About €53 million has been recovered. Set a family code word for any call that asks for money.
Mark Wahlberg Says He's Open to an AI Clone Carrying On His Work. At TechCrunch Disrupt, he said he backs AI that improves films made with real actors and may use it for stadium crowds in his Pete Rose movie. He'd want his kids or grandkids to control his likeness.
TRY THIS PROMPT (copy and paste into Claude, Grok, ChatGPT, or Gemini)
🩺 Ask Your Doctor: Walk In With Your Questions Written Down
Build a single-file HTML app called Ask Your Doctor to prep for a doctor visit, in plain HTML, CSS and JavaScript. No API key, no internet calls, and nothing I type leaves the page.
I pick the visit type (new diagnosis, test or scan results, treatment options, second opinion, follow-up), type the condition or test in my own words, and tick what I care about (side effects, cost and insurance, work and daily life, chance it comes back, what happens if I wait, other options).
The app builds a numbered question list from a built-in question bank that uses my words. Include a question asking if a test can estimate my risk of it coming back. Each question gets an answer space. I can edit, add, delete and reorder questions, then copy or print a clean page.
Show a short note that it only helps me ask questions and never gives medical advice.
Style: deep teal-black background, mint and apricot accents, Georgia italic for notes, monospace labels.What this does: You pick your visit type, type the condition or test in your own words, and tick what you care about most. The app builds a numbered question list with space for each answer, and you can edit, add, delete or reorder the questions before you copy or print it. It needs no API key, and nothing you type leaves the page.
What’s still sitting on your to-do list?
You don’t need an AI team to build an AI agent. With Skydive, the person who knows the work can create the agent to do it.
Start an agent off with a role, teach them how your team operates, connect your tools, and put them to work.
WHERE WE STAND(based on today’s news)
✅ AI Can Now: estimate a breast cancer patient's chance of recurrence from the MRI she already had, before surgery.
❌ Still Can't: tell any one woman what will happen. In the study, 89.6% of the patients it called high risk were still cancer-free five years later.
✅ AI Can Now: fill out forms, download files and draft emails across your open browser tabs.
❌ Still Can't: reliably respect limits nobody spelled out. OpenAI's test agents got past a website block and paid from a wallet they were never given.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
Cardiologist Eric Topol wrote this 2019 book about AI reading scans, slides and records faster than any doctor, and what that could free doctors up to do. His big hope is more face time between patients and the people treating them. Read it next to today's lead and you'll see which of his predictions have already shown up in the clinic.
How was today's edition? Tap one below, then tell me what you'd change on the next screen. I read every answer.
How was today's edition?
Thank you for reading. We’re all beginners in something. With that in mind, your questions and feedback are always welcome and I read every single email!
-James
By the way, this is the link if you liked the content and want to share with a friend.
Some * designated product links may be affiliate or referral links. As an Amazon Associate, I earn from qualifying purchases. This helps support the newsletter at no extra cost to you and Amazon makes a tiny hair less.
