AI Outpaces Doctors in Real ER Cases

Welcome back to The Shift. Let’s get straight to what matters in AI today…

Today we have:

🏥 A Harvard Study Just Showed AI Outdiagnosing Doctors in the ER

✨How to Keep Every AI-Generated Deck On-Brand With Templafy

🛡️ The Pentagon Just Added Eight AI Companies to Classified Networks and Left Anthropic Out

🔨Tools and Shifts you cannot miss

🏥 A Harvard Study Just Showed AI Outdiagnosing Doctors in the ER

A Harvard study published in Science this week put OpenAI's o1-preview through 76 real emergency room cases, and the AI diagnosed patients more accurately than two attending physicians using nothing but raw health record text.

The Decode:

The Numbers Are Hard to Ignore - At initial ER triage, o1-preview gave the correct diagnosis 67.1% of the time compared to 55.3% and 50.0% for the two attending physicians across the same 76 real cases.

Physician Reviewers Could Not Tell the Difference - Two separate doctors tasked with scoring the diagnoses could not identify which came from the AI and which came from human physicians, the outputs were indistinguishable in quality and presentation.

It Caught What Doctors Missed - In one case the AI flagged a rare flesh-eating infection in a transplant patient roughly 12 to 24 hours before the treating doctor identified it, a gap that can be the difference between survival and death.

This Was Not a Specialized Medical Model - The study used o1-preview, a general reasoning model released in 2024, not a purpose-built medical AI. It worked from raw electronic health record text alone with no additional data, imaging, or clinical context beyond what was in the file.

The study compared the model across three decision stages of patient care not just diagnosis. and it held up throughout. AI diagnostic tools are not replacing doctors yet but studies like this are making that conversation increasingly difficult to dismiss.

Together with Viktor

The ops hire that onboards in 30 seconds.

Viktor is an AI coworker that lives in Slack, right where your team already works.

Message Viktor like a teammate: "pull last quarter's revenue by channel," or "build a dashboard for our board meeting."

Viktor connects to your tools, does the work, and delivers the actual report, spreadsheet, or dashboard. Not a summary. The real thing.

There’s no new software to adopt and no one to train.

Most teams start with one task. Within a week, Viktor is handling half of their ops.

Add Viktor to Slack for free.

✨How to Keep Every AI-Generated Deck On-Brand With Templafy

Your team is prompting presentations. Your brand has no say in what comes out. Here's how to fix that once.

Step 1: Upload your brand assets and rules Set your approved fonts, colors, templates, and content guardrails inside Templafy. This happens once, everything after runs automatically.

Step 2: Connect to PowerPoint Templafy works directly inside PowerPoint. No new tool for your team to learn or adopt.

Step 3: Team prompts as normal Describe the deck in plain language. Templafy generates it using your brand rules in the background, your team never has to think about compliance.

Step 4: Every output is client-ready by default No rogue fonts. No off-brand slides. No policing decks before they go out. Brand compliance is built into every generation automatically.

BDO saved $1.65M in a year. Adobe teams recovered 72% of their time. 4M+ users at Adobe, KPMG, and Pandora are already running on it. You can try Templafy free here.

🛡️ The Pentagon Just Added Eight AI Companies to Classified Networks and Left Anthropic Out

The Department of Defense added SpaceX, OpenAI, Google, Nvidia, Microsoft, AWS, Oracle, and Reflection to its classified networks this week while keeping Anthropic out despite the new contracts carrying the same limits that got Anthropic blacklisted.

The Shift:

Who Got In and What It Means - SpaceX, OpenAI, Google, Nvidia, Reflection, Microsoft, AWS, and Oracle are now inside Pentagon classified networks. The DoD said the deals will accelerate its transformation into an AI-first fighting force.

Anthropic Is Still on the Outside - DoD CTO Emil Michael confirmed Anthropic's supply-chain risk label still stands. The same autonomous weapons and surveillance restrictions that led to Anthropic's exclusion are present in the new contracts, raising obvious questions.

Mythos Is Being Treated Differently - Emil Michael called Anthropic's Mythos model a separate national security moment, suggesting it may be evaluated outside the standard contract framework. What that means in practice has not been explained.

The White House Timing Is Notable - Anthropic's exclusion came days after the White House came out against a broader Mythos rollout over compute concerns affecting its own access. The two events in the same week are not a coincidence.

The Pentagon has not explained why Anthropic remains excluded while admitted companies operate under the same contractual limits. With Mythos flagged as a separate national security consideration, Anthropic's path into classified government work may still be open, just through a different door.

Together with Morning Brew

Read less. Know more.

Morning Brew delivers the biggest stories in business, finance, and tech in about 5 minutes — with just enough personality to keep things interesting.

Join 4,000,000+ professionals who start their mornings a little smarter.

Check it out

🔨AI Tools for the Shift

🤖 AskFred – Answers your customers 24/7, even when you are offline, with a chatbot you can set up in just 5 minutes.

🌐 WebZum – Instantly analyzes your market and launches a ready-to-use website in minutes without manual setup.

📸 Profyle – Generates studio-quality professional headshots in hours without needing a photoshoot.

⚡ GetTheGists – Reads any page and delivers the key insights, why it matters, and only the parts worth your time.

🪨 Rock Identifier – Identifies rocks, crystals, and minerals from a photo with details like type, hardness, and estimated value.

🚀Quick Shifts

🎬 SAG-AFTRA reaches a four-year deal with studios, including higher streaming residuals, pension contributions, and new AI protections for actors, following recent industry-wide labor negotiations.

🎬 The Oscars now require human-performed acting and human-written scripts, making AI-generated performances and screenplays ineligible, with the Academy enforcing stricter rules on authorship and AI usage.

🤖 Meta acquires robotics startup ARI to advance humanoid AI, adding expertise in robot learning, behavior prediction, and physical task automation as part of its long-term artificial intelligence and robotics strategy.

⚖️ Microsoft is launching a Legal Agent in Word that reviews contracts, tracks edits, and analyzes risks using structured workflows, designed for legal teams handling complex documents and negotiations.

⚖️ A Chinese court rules that replacing a worker with AI is not valid grounds for termination, ordering a tech firm to compensate the employee for wrongful dismissal.

🧩 Prompt of the Day

Brand Collaboration Outreach That Gets Replies

Most outreach emails get ignored because they are generic. Strong collaboration pitches are clear, relevant, and show mutual value immediately.

Make your outreach feel like an opportunity, not a cold ask.

Paste the prompt: Drop this into ChatGPT, then fill in your brand and partner details.

Prompt to paste

Write a brand collaboration outreach email to [Insert brand name]. Start with a brief introduction of your brand and why there is alignment. Clearly explain the collaboration idea and how it benefits both sides. Include a strong call to action with next steps such as a quick call or proposal. Keep the tone concise, personalized, and focused on mutual value.

🤳AI Nugget of the Day

That’s all for today’s edition. See you tomorrow as we track down and get you all that matters in the daily AI Shift!

If you loved this edition, let us know how much:

Forward it to your pal to give them a daily dose of the shift so they can 👇