Your organization isn’t making you more capable. It’s making you look ordinary. AI is about to make that obvious.
🔬 The Big Thing
Your organization isn’t making you more capable. It’s making you look ordinary. AI is about to make that obvious.
Nate’s Newsletter dropped an Executive Briefing yesterday that extends the argument from last week’s piece about expanding ambition when execution cost drops — but this time the lens is on the individual, not the company. And the framing hit harder.
The setup: a senior PM at a company Nate works with resigned after six years of “exceeds expectations” ratings. She was, by his account, the person you wanted in the room when something was broken. She was also spending 75% of her working hours on alignment meetings, cross-functional syncs, stakeholder management, and the slow process of translating her judgment into something a team of eight could execute across three time zones. She left. She picked up Claude Code and Cursor. In her first month solo, she shipped a product addressing a market gap her former employer was roadmapping for Q3. Not a prototype. A product, with paying customers.
Nate’s thesis: everyone is asking “how do we find extraordinary people?” The question they should be asking is “how did we spend the last thirty years building organizations that make extraordinary people look ordinary?” And now that AI lets those people route around organizational overhead entirely — what exactly is the plan?
The examples are stacking up. Maor Shlomo bootstrapped Base44 — a vibe coding platform — to 300,000 users and sold it to Wix for $80 million in cash, six months from start to exit. Ben Broca crossed $1M ARR with Polsia in his first month. Sarah Gwilliam, a grief coach who doesn’t “speak AI,” launched Solace through an AI-native incubator — no founding team, no technical co-founder. Dario Amodei puts 70-80% odds on a one-person billion-dollar company by 2026. Carta reports the share of solo-founded startups rose from 23.7% in 2019 to 36.3% by mid-2025.
Here’s why I keep pulling Nate’s frameworks into this newsletter: every one of these examples translates directly to the clinician-builder context, and the translation matters more than the original.
That PM who shipped solo in a month? She had deep domain knowledge of customer problems, years of pattern recognition about what works and what doesn’t, and the judgment to make a hundred small product decisions correctly without a committee. She didn’t become a better engineer. She routed around the organizational overhead that was suppressing her capability.
Now think about the pharmacist in your ED who’s been describing a specialty-specific med rec tool for months. She has the same ingredients: deep domain knowledge, years of clinical judgment, pattern recognition about what fails and why. What she’s been missing is the execution capability — and what Nate’s piece documents is that the gap between “expert who knows what should exist” and “person who ships it” is collapsing. Not shrinking. Collapsing.
Nate’s piece last week was about what organizations should do: when execution cost drops 10x, expand your ambition. This piece is about you, individually: the organizational structure you’re embedded in has been suppressing your output, and you now have the tools to route around it. Your clinical judgment — the specific, contextual, 2-AM-potassium kind — is the scarce input. The execution is increasingly commodity.
I’m honestly not sure what this means for health systems yet. Do clinician-builders stay inside and push for more building autonomy? Do they leave and build independently? Both? The answer is probably “yes, and it depends on the individual, the institution, and what they’re building.” But I think the era where a talented clinician’s building ambition gets absorbed into committee meetings and requirements-gathering cycles that take nine months to produce a spec — that era is ending. Faster than most CMIOs realize.
Nate’s Newsletter — Executive Briefing · Base44 / Wix $80M acquisition (TechCrunch)
📡 Builder’s Radar
Nature Medicine published a framework that makes every clinical LLM benchmark you’ve seen look like a toy.
A team from the Broad Institute and collaborators published the Clinical Environment Simulator (CES) in Nature Medicine last week — and it landed during the HIMSS chaos, so I’m flagging it now. CES evaluates clinical LLMs inside a simulated hospital where every decision dynamically alters future states. It has a “hospital engine” tracking bed availability, staff workloads, and equipment in real time, and a “patient engine” simulating disease progression and treatment responses based on the LLM’s interventions. The LLM has to execute decisions through realistic EHR interfaces while managing trade-offs between individual patient optimization and system-wide efficiency. This is what the validation gap conversation at HIMSS was pointing toward: static benchmarks don’t tell you if an LLM can handle a 2 AM admission surge with three beds left and a ventilator shortage. CES does. If you’re building clinical AI tools, this is the evaluation standard that governance committees are going to start demanding.
Codoxo launched deepfake detection for AI-generated medical records — and yes, that’s a thing now.
This one also landed during HIMSS week. Codoxo, backed by a recent $35M Series C led by CVS Health Ventures, shipped a tool that identifies AI-generated or manipulated medical documentation and diagnostic images submitted with insurance claims — before payment. The builder angle: generative AI made it trivially easy to fabricate convincing clinical narratives and medical records. Codoxo’s detection system analyzes documentation and images in seconds, flagging cloned records reused across patients and synthetic content that rules-based systems miss. This is the flip side of the vibe coding revolution — the same tools that let clinicians build faster also let bad actors fabricate faster. If you’re building anything that generates clinical documentation, the audit trail and provenance story just became a feature, not an afterthought.
What are you building this week? Reply and tell me — I read every one.
— Kevin