“People should stop training radiologists now,” once warned Geoffrey Hinton, the “Godfather of AI,” predicting that artificial intelligence would soon make the profession obsolete.

Ten years later, radiology residency programs, vacancy rates, and salaries are at all-time highs.

Today, 83 million Americans lack access to primary care, and the U.S. faces a shortage of 124,000 physicians by 2034. As AI reduces barriers to access, it will unleash demand for doctors and specialists far beyond what our current system can support.

So stop worrying about AI replacing doctors. Worry more about how many more we’ll need. This is Jevons’ Paradox in healthcare, a dynamic that cuts across learning, health, and work — all of our core investment areas. Jomayra Herrera dives into the looming workforce crisis and where we’re investing to strengthen the talent infrastructure.

The school tech backlash is real. But the debate deserves more precision. Too often the loudest voices coalesce around binary positions — ban it or embrace it fully — when the reality is more nuanced and consequential. 

Over the past few weeks, we've been writing a series to bring more depth to the conversation, drawing on developmental neuroscience, research, and what we’ve seen as former educators and active investors in this space:

Part 1. Banning Edtech Is the Wrong Response to the Right Questions: The instinct to protect kids from technology is right. Blanket bans are not. There is a meaningful difference between restricting YouTube on Chromebooks and removing the reading tool that catches a child's dyslexia. Policies that fail to make that distinction will cause real harm.

Part 2. Not All Edtech Is Equal: Schools run on three distinct layers of technology — operational infrastructure, teacher support platforms, and student-facing tools. Collapsing them into a single policy conversation risks breaking far more than it fixes.

Part 3. The Science Behind Smarter Screen Time: Whether a screen helps or harms a child’s learning comes down to three things — and not one of them is whether the screen is in the room. Developmental neuroscience offers a more specific and durable guide than any ban, grade by grade.

Congratulations to Carl Madi, Tressia Hobeika, Edoardo Serra and the Stepful team on raising their $55 million Series C, led by Oak HC/FT!

When we first invested in 2021, the company had a simple belief: a high school graduate should have a clear, affordable path to a thriving healthcare career. Since then, more than 32,000 learners have completed Stepful programs, including a Medical Assistant pathway that is 10x more affordable and 4x faster than traditional alternatives. Today, 35 health systems — including most recently Mount Sinai — recruit directly from its talent network.

Large language models excel at parsing text and code, but struggle with complex scientific data such as satellite imagery and seismic surveys. We’re thrilled to back Lium, which is making such complex data sets accessible to researchers through a conversational AI platform. We’re doubly excited to welcome CEO and co-founder Josh Knutson back into the Reach portfolio, following a successful exit with his previous company, Rhithm. Read Silicon Angle’s coverage of its seed round.

1-2-3-4 🎵 Schools need more artists in the door. We’re excited to co-lead, along with Yamaha Music Innovations, a pre-seed investment in CommonTime Pathways, which helps higher-ed institutions attract, recruit, and retain musicians, dancers, and other “AI” (artistic!) talent in much the same way that has been done for athletes.

A college hackathon organizer. Fujifilm and Japanese city pop enthusiast. A former Googler who’s shipped AI products at billion-user scale. And our newest team member. 

Say hello to Harry Luo, our Senior Associate! He will focus on sourcing AI deals, leading technical diligence, and supporting the technical founders already in our portfolio and the others headed there. If you’re a founder building something AI-native in learning, health or work, Harry wants to hear from you!

The spring, D’s and F’s for several CS classes at UC Berkeley skyrocketed 3x-5x from previous years. Nearly half failed CS 10: The Joy and Beauty of Computing.

One professor told The Daily Cal that students were “leaning a little too hard on LLMs to do their work for them, and then at exam time just really aren’t ready.” Another said many lacked the prerequisites, such as linear algebra, vector calculus and mathematical proofs.

Or are CS students perhaps caring less about grades and credentials now, especially in a field that has been severely impacted by AI? What do you think?

The 2026 Reach Conference Calendar is the best way to see all the major industry events across learning, health, and work (filterable by sector, date, and location). If you’re hosting a conference or know of one we missed, reply to the newsletter.

A new path to higher education: accelerating the talent of tomorrow and reimagine what college can (and should) be

The marketplace with 400+ clinicians delivering in-home & insurance-covered pediatric therapy

The largest microschool network in the country, helping educators start and run microschools

The agentic software creation platform that enables anyone to build apps and businesses

The multimodal platform getting students to speak, draw, and write to make their reasoning visible