😺 ChatGPT MD?!

PLUS: OpenAI's new device revealed??

Welcome, humans.

OpenAI CEO Sam Altman dropped by YC Startup School to share war stories from OpenAI's journey, including that time Elon Musk sent him a brutal email saying they had “zero% chance of success” after seeing GPT-1 (ouch).

He teased that OpenAI wants to eventually send free humanoid robots with premium ChatGPT subscriptions, discussed how reasoning models like o3 are opening up massive opportunities for startups, and shared that Johnny Ive’s mysterious new AI device will make our current interfaces feel as outdated as flip phones.

Fun fact: OpenAI also just removed the Jony Ive partnership materials from its website over a trademark lawsuit; apparently the deal is still on, but the company was sued by a startup called “iyO” that develops custom-molded earpieces that connect to other devices. Soooo, an earpiece could very well be what the device is!

The most relatable moment from Sam’s talk? When he admitted to buying NVIDIA stock in 2018, and selling it less than a year later… proving that even the guy “building AGI” can't time the market.

Here’s what you need to know about AI today:

People are increasingly using AI chatbots for medical advice (including doctors).
Ex-OpenAI CTO Mira Murati raised $2B for her new AI startup.
Perplexity launched AI video generation on X with sound and dialogue.
Goldman Sachs launched a genAI assistant used by 10K employees.

Advertise in The Neuron here.

ChatGPT MD? ChatGPT isn't replacing doctors, and that’s a problem… (here’s why).

A Reddit user just credited ChatGPT with potentially saving his wife's life. It happened when she was feeling feverish after a cyst removal. She wanted to “wait it out”, but he plucked the details into ChatGPT (as he often does), and to their surprise, ChatGPT urged them to get to the ER, ASAP.

And guess what? It was right. She had developed sepsis.

Naturally, the comments exploded with similar stories: AI catching blood clots, diagnosing gallbladder issues, even spotting rare conditions that had stumped 17 different doctors.

While these anecdotal wins grab headlines… a major new Oxford study with ~1,300 participants revealed a vital truth about AI medical advice: people using AI chatbots performed worse than those just using Google.

Here’s what they found:

When AIs worked alone, they scored 90-99% accuracy identifying conditions.
But when paired with humans, users could only identify relevant conditions 34.5% of the time; significantly worse than the 47% achieved by people just using Google.
The breakdown? Users provide incomplete information, can't distinguish good from bad AI suggestions, and even ignore correct recommendations even when the AI gets it right.

On the flip side, doctors themselves apparently have amazing AI tools.

One startup OpenEvidence—an AI diagnosis engine used by ~25% of all US physicians—is trained exclusively on peer-reviewed medical literature and partners with journals like NEJM and JAMA (here’s a cool demo).

These doctors apparently access it an average of 10 times a day to help with complex cases, especially in oncology.

Some hospitals, like Johns Hopkins, even use AI themselves to predict sepsis risk…the exact thing ChatGPT told the guy to go check for. And of course you’ve seen how

Now, here’s the deal: the data on ChatGPT's medical accuracy is all over the map—from 49% accuracy in one study to outperforming emergency department physicians in another. But the pattern is clear: when AI is designed for medical professionals with proper oversight, it works remarkably well.

Partnered with professionals is key. It makes you wonder: who is going to make the first doctor approved patient-facing chat portal? Because the anecdotes are telling us that’s what the people really want. Are we just too afraid of being sued to release it?

Read our full deep-dive analysis with all the research details and implications here.

FROM OUR PARTNERS

💡 The easiest way to bring AI to your company

Sana Agents is where teams go to work with AI. Instantly co-create flawless reports, decks, and dashboards—built from your live company data.

One unified workspace: Bring together enterprise search, meeting notes, workflows, and agents in a single, secure surface.

Delegate the heavy lifting: Agents generate end-to-end deliverables—from presentations to SQL insights—in minutes, not hours.

Enterprise-grade intelligence: Get precise, permission-aware answers grounded in your data.

Drive enterprise-wide innovation today with Sana Agents.

Prompt Tip of the Day

If you want help getting started using AI, there’s probably no better source from a beginners perspective than Ethan Mollick. He just put together an excellent guide on on which AI systems to use and how to actually use them effectively.

Source: Ethan Mollick.

The TL;DR is that most people are using AI like Google—quick questions, no context, default settings—when they should be treating it more like a smart assistant.

Here's his key advice:

Pick one of the big three: ChatGPT, Claude, or Gemini ($20/month for full features).
Switch to the powerful models for serious work (not the default fast ones).
Try Deep Research for comprehensive reports and analysis.
Use voice mode with your camera/screen sharing for real-time help.
Give the AI context with documents and clear instructions.
Ask for lots of options (50 ideas instead of 10; AI be lazy if you aren’t specific).
Use branching to explore different conversation paths.
Test three things right away: a complex work challenge, Deep Research, and voice mode.

Our favorite insight: The difference between casual and power users is not actually prompting skill… it's knowing all these features exist and actually using them on real work!

Treats To Try.

*Anything marked with asterisks is sponsored content. Advertise in The Neuron here.

RefineAI can.

*RefineAI uses domain knowledge and the open-source Refine framework (31K Stars) to generate production-grade internal React apps. Start Building for Free.
MidJourney’s new videos are now part of the explore page; it hasn’t been ranked by Artificial Analysis yet, so we’re still waiting to see where it lands.
11.ai is ElevenLabs’ new experimental voice-first assistant that helps you automatically handle tasks like adding meetings to Notion, researching customers through Perplexity, and creating Linear issues.
Keyword shows you where your brand appears when people ask ChatGPT, Google AI, or Perplexity questions—free trial, then $24.5/month.
Martin acts as your personal assistant through text, voice, or Slack—tell him “text Jon my arrival time” and he'll check your calendar, calculate travel time, and send the message, or say “remind me to meal prep at 6PM" and he'll set it up while remembering your preferences (unlike Siri).
Databutton turns “build a customer feedback tracker” into a complete working app with database and login in 60 minutes—free first month
Human vs LLM shows you how your website looks to an AI; it’s free to try, but to see the whole page you need to sign up—here’s a great article on why understanding how AI sees your sites is really important for SEO going forward.

See our top 51 AI Tools for Business here!

Around the Horn.

* ~ raises hand~ * We are!!

Ex-OpenAI CTO Mira Murati raised $2B for her new startup Thinking Machines Labs, and will use the funds for what she’s calling “RL for businesses”, or using reinforcement learning to create custom models tailored to a specific business’ key performance metrics.
Google’s Chromebook Plus devices will now come with AI tools like NotebookLM, a Google Lens style screen-selector tool, and a “Simplify” tool to explain or summarize complicated stuff.
G7 nations launched a joint AI innovation challenge called the “GovAI Grand Challenge” to help small businesses adopt AI and address energy demands from data centers.
Economists say the current white collar hiring slowdown is not caused by AI replacing jobs, but actually the two-year decline in professional services hiring is due to more structural economic issues.
Perplexity launched 8-second video generation on X, so users can now tag @AskPerplexity with prompts and receive AI-generated videos with sound and dialogue, though high demand is causing delays.

FROM OUR PARTNERS

CodeRabbit’s Free AI Code Reviews in your IDE

Get senior engineer level code reviews in VS Code, Cursor, and Windsurf with CodeRabbit. Instantly catches bugs, suggests refactorings, and delivers context aware feedback. Works with all major languages; trusted on 10M+ PRs across 1M repos.

Install the extension and start vibe checking your code today!

AI in the Real World

An actual AI marketing success!! Here’s the dentist’s channel. Apparently his practice is “totally full” now. Shout out to Dan Taylor Watt, who shared this; his blog on where AI and media collide is great!

A wild 30-second Kalshi commercial created entirely with AI for the NBA Finals is raising questions about the future of advertising while costing a fraction of traditional production.
Teachers in 60 Philippine schools are using Microsoft's Reading Progress tool to help students improve reading skills by providing instant feedback and cutting assessment time from 2 days to minutes.
Goldman Sachs launched a genAI assistant to draft content and analyze data that’s already used by 10K employees, and will now roll out to all 46K+ employees (even in high-stakes banking roles).
Deezer found that 18% of uploaded tracks are fully AI-generated and is using new detection tools to filter AI music from recommendations.