In partnership with
Did you know that we have LinkedIn, Instagram and X accounts
that you can follow?
Hi {{first_name|everyone}},Microsoft built an AI doctor that works through a case one question at a time instead of guessing, and the way it does that is worth a look. Also in the mix this time, a debate over what programming languages are even for now, a few new tools worth bookmarking, and an AI mirror that’s a little too flattering. Alright, let’s jump in!
🤿 DEEP DIVE
Microsoft's AI Doesn't Guess Diagnoses
It works through the problem, one question at a time.
Apparently, giving AI all your symptoms and asking what's wrong isn't necessarily the best way to make it a good doctor.
Microsoft researchers have been testing a different approach, and the results are pretty striking. Their AI system was more than twice as accurate as general physicians at diagnosing over 300 unusually difficult medical cases.
But I think the more interesting part is how it got there.
It's not just looking for an answer
Most medical AI systems work by taking everything available about a patient, like symptoms, blood tests and scans, and trying to produce a diagnosis from all that information at once.
Microsoft's system does something different.
Instead of getting the whole picture upfront, it has to figure out what information it actually needs. Much like a doctor would when someone first walks into the room.
So it starts by deciding what questions to ask about the patient's history and what to look for during a physical exam. Then, based on those answers, it decides which tests to order.
And it doesn't stop there.
As the results come in, it updates what it thinks might be wrong, decides whether more testing is needed, and keeps working through the possibilities.
Basically, it's not just trying to get the diagnosis right. It's trying to figure out the right way to reach that diagnosis.
There's a whole AI medical team behind it
The researchers call the coordinating system an "orchestrator."
It's essentially the part that manages the diagnostic process, using a simulated panel of clinical experts that collaborate on the case.
And that seems to be the key distinction here. Rather than having one AI produce an answer from a pile of information, the system coordinates a series of decisions about what to investigate next.
The researchers also found that this approach produced diagnostic work-ups at a lower cost than those performed by human doctors.
So the improvement wasn't just about getting more answers right. It was also about being more selective with the questions and tests needed to get there.
Getting the answer right isn't the whole test
There's another detail I liked from the research.
Microsoft's Dr. Dominic King pointed out that AI has become extremely good at medical multiple-choice questions. But real patients don't arrive with a neatly written summary of everything that's wrong with them.
That's exactly the gap this research is trying to address.
And while the results are impressive, the next step is a more practical test. Microsoft is collaborating with Beth Israel Deaconess Medical Center in Boston to evaluate whether the system can help triage patients waiting in emergency departments.
For now, the reported results come from diagnostic evaluations, not that upcoming hospital test.
Still, the distinction is worth remembering. The system isn't simply being tested on whether it knows the correct answer. It's being tested on whether it can work out which questions to ask in the first place.
The quick take
Microsoft's AI outperformed general physicians on over 300 unusually complex diagnostic cases.
Instead of analyzing everything at once, it asks questions, orders tests and updates its diagnosis step by step.
An orchestrator manages a simulated panel of clinical experts working through each case.
Researchers found that the AI-guided process was less costly than physician-led work-ups.
Microsoft and Beth Israel Deaconess Medical Center will evaluate its ability to triage emergency department patients.
🤝 FROM OUR FRIENDS
A quick share we thought you might find useful
Turn Your Achievements Into a U.S. Immigration Pathway
Publications, patents, leadership, critical roles, compensation, awards, and business impact can carry different weight in O-1, EB-1, and EB-2 NIW cases. That makes it hard to know which path fits your background or how to use the evidence you already have.
At Manifest Law, our attorneys help engineers and founders organize the full picture, identify questions that need legal review, and understand practical next steps. Take the intake quiz to share your background and begin a personalized assessment.
Attorney Advertising
🔥 WHAT’S THE BUZZ
What people are building, sharing, and talking about
DHH had AI agents rewrite the same app in Rust, Go and Elixir, and now he’s questioning what programming languages are actually for.
AI coding may change more than how quickly software gets written. If humans spend less time directly manipulating source code, some long-standing reasons for choosing one programming language over another could start shifting toward things agents care about: compiler feedback, execution speed, testability and how reliably the model can work in that ecosystem.
90 AI agents went looking for new magnets. One promising candidate was hiding in a 1999 research paper.
This is a practical example of AI agents doing more than automating routine tasks. They can help researchers explore possibilities, run calculations and revisit existing knowledge to uncover things people may have overlooked. The human researchers still need to verify the results.
⚒ TOOL SNAPSHOTS
Futuristic tools within AI, no-code, and productivity
📊 DeepCell
Build reports, forecasts, and presentations without juggling separate files.
Why it’s useful: Handy for teams working on budgets, sales reviews, or monthly reports where the numbers, explanations, and slides all need to stay in sync.
Turn speech into text, or create voices that sound natural.
Why it’s useful: Worth a look for developers building multilingual transcription tools or voice agents that need fast responses and more control over how speech sounds.
Take your ads from initial ideas to running campaigns.
Why it’s useful: Useful for small marketing teams that want to create ads, run campaigns, and use performance data to improve the next round without managing every step manually.
👀 ON OUR RADAR
One more thing we thought you'd like
Notifications were allegedly timed for maximum excitement and families are filing claims over alleged addictive design. You may qualify if you're a parent of a harmed minor, use began under 18, and there was a diagnosed eating disorder, severe anxiety, depression, or self-harm.
🤖 AI MADE THIS
Interesting and inspiring creations made with AI
An AI Mirror That Turns You Into an Uncomfortably Perfect Version of Yourself
What happens when an AI mirror decides what your perfect face should look like? Matthias Oostrik's SMILE turns visitors into eerily idealized versions of themselves using Stable Diffusion, making AI's beauty biases impossible to ignore. The installation is returning to Amsterdam Dance Event..
AI used:
Stable Diffusion The artist's technical specifications also name YOLOv8, DeepFace and DepthAI. Midjourney and DALL·E are discussed as related technologies, but the artist does not list them in the installation's confirmed software stack.
ℹ️ ABOUT US
The Intelligent Worker helps you to be more productive at work with AI, automation, no-code, and other technologies.
We like real, practical, and tangible use-cases and hate hand-wavy, theoretical, and abstract concepts that don’t drive real-world outcomes.
Our mission is to empower individuals, boost their productivity, and future-proof their careers.
We read all your comments - please provide your feedback!
Did you like today's email?
What more do you want to see in this newsletter?
