Sponsored by
Good morning.
This issue starts with an AI catching something doctors missed for a hundred years, and carries one idea through to the end: write down what "good" actually means before you ask AI for anything, whether that's a first draft, a back-to-school list, or your own calendar.
AI INSIGHT
AI Could Help Doctors Match Breast Cancer Patients to the Right Treatment
Researchers at the University of Southampton built an AI tool that studied more than 330,000 tiny structures inside breast cancer cells, called centrosomes, and found something doctors had missed for a hundred years.
The tool, called CenSegNet, uncovered two separate kinds of centrosome damage. One is cells collecting too many of these structures. The other is the structures growing abnormally large. Doctors had always lumped the two together. The AI showed they behave independently, showing up in different parts of the same tumor.
The research, published in Nature Communications, is open source and free for scientists to use. The same method already works on kidney, colon, and appendix tissue. No pathologist could realistically count 330,000 of anything by hand. That's the real shift here: AI didn't discover a new fact so much as it finally let researchers see an old one clearly, at a scale no human eye could sustain.
Researchers hope it eventually helps match patients to treatments based on which specific defect their tumor carries, not just a general diagnosis. For a disease that behaves differently from patient to patient, that kind of precision could change which treatment a doctor reaches for first.
This is the pattern that keeps showing up with AI in medicine: not a replacement for a doctor's judgment, but a way to catch what a human eye alone never could. Full details are in the original research writeup.
The catch: This is lab research, not a clinic test. No doctor can order this for a patient yet.
FROM OUR PARTNER
Stop googling AI tools at midnight.
There are over 3,000 of them in The Shift’s vault, already vetted, ready to explore.
Subscribe and get instant access to the tool vault, a 1000+ prompt library, and free AI courses built for people with actual work to do.
Plus, the daily newsletter that keeps you sharp on everything moving in AI in under 5 minutes a day.
They’re also giving away a free 1-year Claude Pro subscription to 3 subscribers.
Subscribe for free to enter. All free. All in one place.
AI TUTORIAL
Teach AI Your Standards, Not Just Your Preferences
I've taught ChatGPT my wife's and my travel preferences, and my own music taste, once. For music, I gave it ten or twelve of my favorite artists and let it build a real profile from there. For travel, I told it how many steps a day we like to walk, what we like for breakfast, lunch, and dinner, and the kind of tours we actually enjoy, usually something on the water, and local live music if we can find it. I don't retype any of it. ChatGPT still applies it months later. The thing that made that work wasn't telling it what I like in the moment. It was writing down what counts as good, once, and putting it somewhere AI would actually keep it.
1. Pick one recurring decision you hand to AI: travel, music, gifts, a first draft, a purchase you keep researching.
2. List three things that make the result genuinely good, and three specific things that make it a miss.
3. Turn that list into one paragraph, written like a rule you'd hand someone else, not a mood.
4. Teach it to AI once, in ChatGPT's memory or a Claude Project, not a note you plan to paste back in.
5. Test it on a real request a few days later, with no reminder, and correct whatever it still gets wrong.
Try This With AI What it does: Hands AI a written standard and asks it to remember the standard itself, not just the request you happen to be making today. "Act as an assistant who remembers this for every future conversation, not just this one. Here's what counts as good for [task, e.g. music recommendations, gift ideas, a first draft]: [list 3-5 standards]. Save this and apply it automatically from now on, and check with me before you assume an exception." Customize it: In ChatGPT, this saves to memory directly. In Claude, add it to a Project's instructions instead of a one-off chat so it holds across sessions. |
If it does not work: ChatGPT will sometimes say it saved something and still not apply it a few chats later, memory doesn't catch everything on the first try. Open Settings, then Personalization, then Memory, and check it actually stuck. If it didn't, shorten it a bit and try again, that alone fixes most misses.
What this doesn't fix: A written standard still needs your judgment on the close calls, the ones where two good options both technically fit the rule. It gives AI something to check against instead of a guess, not a replacement for deciding what actually matters to you.
FROM OUR PARTNER
Short on cash? Get up to $750, no credit check.
Access up to $750* from your upcoming paycheck within minutes for a small fee, or wait 3 days and get it for free. No credit check. No late fees.
*Not all users will qualify. Advances range from $25–$750; average advance is $170. Express transfer fees may apply.
NEWS YOU CAN USE
This Matters at Home
Short, practical summaries to help you stay updated on recent AI news—and what it means for everyday life.
📺 Alexa+ Just Went Free On Every Fire TV. No subscription needed now. Ask for a highly rated thriller instead of a title, or pull up your Ring camera feed on the big screen.
📢 Your Free ChatGPT Is About To Start Selling You Things. Ads land in ChatGPT across 31 European countries today, six months after the US pilot. Free and cheap Go plans get them. Paid tiers stay clean.
🎓 Google Is Giving Students A Free Year Of Gemini. College students get 12 months of Google AI Pro, normally $240. A card is required up front and it auto renews at full price. Set a reminder.
🔧 The Voice Booking Your Oil Change Is Probably A Bot. That friendly voice may be a bot trained to keep offering alternatives every time you say no. Stick to the schedule in your owner's manual.
REAL WORLD
How To Use AI This Week
Real people, real uses—here’s how AI is quietly making life better, one small win at a time.
Before You Buy School Supplies
🎒 Set Your Standard Before You Shop. "Act as a practical shopping assistant who asks what actually matters before recommending anything. Help me define what "good enough" looks like for this year's back-to-school list, covering clothes, school supplies, and any tech, so I stop second-guessing items once they're already in the cart. Then build the full list against that standard: [ages or grades of your kids], [total budget], [must-haves vs nice-to-haves], [anything already on hand from last year]." ◆ Try in ChatGPT ◆ Try in Claude
READER POLL
Southampton researchers used AI to catch a pattern doctors missed for a century. How do you feel about AI being used this way in medicine?
TOOLS FOR YOU
📅 Tool Spotlight: Reclaim.ai
Reclaim.ai is an AI calendar tool that learns your actual habits, not just your stated preferences, and defends time for the things you keep meaning to do. Tell it once that mornings are for deep work or Tuesdays are for family calls, and it protects those blocks automatically instead of letting meetings creep in. It's the calendar version of today's tutorial: teach it your standard once, and it holds the line for you. Start here.
PASS IT ON
Know someone who would like this newsletter? Send them here.
Missed an issue? Read the archive.
Advertising Disclosure: We evaluate all recommendations of products and services independently. Clicking on links provided on this page may result in AI for Daily Living earning compensation, which supports independent publishers like us.
