Blog

See careers: /careers .

At any company there is always a need for secure networking for provisioning access to databases, giving engineers privileged access to internal networks, or just having internal services company wide. Traditionally, this would be done with a VPN or SSH tunnels to specific servers with privileged network access. This is not optimal, even once your team grows past 5 it becomes burdensome. Maintaining either solution can grow to a full time job itself, and punching holes in our office network and/or our virtual private cloud is very error-prone. We would also need to provision separate solutions for each type of access. Instead, we turned to Tailscale , which would allow us to very quickly provision access to key internal services with minimal operational work for engineers.

My first post in this series looked at the moral implications of centralized healthcare decision-making, dubbed “The Algorithm”. That post explained why an obsession with healthcare data leads to oppression and moral degradation. This post aims to show why that oppression isn’t even markets-wise effective or efficient, anyways.

Everyone in health insurance is talking about data. In the Year of our Lord 2025, AI is the big bright mysterious star on the horizon steering these conversations, but the conversation about data in health insurance is far from new. There is no doubt that good people want to use data for good purposes. They want to measure the quality of healthcare being delivered and the cost of various services. With this knowledge, they believe they will be able to create an omniscient, all-knowing god that shares their enlightened values. This god I shall call “The Algorithm”. Trillions of dollars have been invested in the divine undertaking of creating The Algorithm. Our nation’s Mission-Oriented-Product-Managers have spent countless hours on echo-y Zoom calls explaining The Algorithm to the gentiles, their sermons recorded and transcribed and subject to your organization’s Mutual NDA.

ICHRAs are a hot topic. Not only have many companies raised a lot of money recently to “reinvent health insurance” by administering ICHRAs, but early versions of the One Big Beautiful Bill (BBB) Act included a tax break for businesses that offered ICHRAs. This combination of events has made ICHRAs a hot topic on my LinkedIn feed. The general lack of substantive discussion about the mechanics, history, and system-level impacts of ICHRAs have frustrated me as someone who likes to dive into the details of how health insurance works. In this post, I aim to give you enough context to understand the ICHRA landscape from both the micro and the macro perspective. You will understand both how an individual ICHRA plan operates and the role that ICHRAs play within the broader topic of healthcare financing.

I’ve spent the last few weeks navigating the U.S pharmaceutical industry and my general instinct with topics that are either intensely confusing or interesting is to write about them. So this is my attempt at unraveling the tangled web of stakeholders in pharmacy and their dynamics with each other as concisely as possible. I’m hoping to particularly highlight the unintuitive flow of money between these stakeholders, and how it ultimately affects the consumer. It’s precisely the complexity of these economics that is driving waste in the way health plans are designed and run, and it starts with the manufacturers. The U.S is a global behemoth in pharmaceutical manufacturing—46% of global pharma sales in 2022 originated in the US

To a layperson (including almost all VCs) there is a belief that healthcare administration is ripe for AI disruption. Most people believe that this disruption will be good. Everyone knows that healthcare administration costs too much money, too much time, and prevents too much innovation. It's natural to think that if individual healthcare administrative jobs could be automated by AI, then the entire healthcare industry post AI utopia will spend less on admin and more on care.

Context: A good thread in the Out of Pocket Slack community prompted a discussion about different types of health insurance plans. The world of health benefits has gotten remarkably confusing in the last 10 years, and there are a lot of styles of health insurance plans in the market. In order to explain this complexity, it helps to actually think about the problem from first principles and understand the order in which various types of health plans actually developed. The right framework to have is this: healthcare should be simple and used to be simple. It only got complicated because of bad incentives (for all players involved) and the many "innovations" in healthcare that tried to reduce cost but ended up increasing complexity.

In a not-that-alternate alternate universe, I would be a math professor studying something like game theory. Game theory is the study of incentives and how incentives can shape how people and organizations interact with each other. Game theory is a popular approach to studying social and economic problems and designing how markets should work. The game theory approach to answering social problems is to model the incentive functions of the agents whose interaction you want to study and then solve for the optimization functions of these agents under various conditions. Game theory can reach some pretty interesting results about human behavior through very basic assumptions and 5th grade math.

Context: I'm writing this blog post to describe the role and interaction of the players that make a self-funded health plan work. The audience for this post is an intelligent business leader who is curious and wants to know a bit more detail than what the average broker can or will provide. At Yuzu, our #1 value is honesty. We never treat you like you "can't handle the truth". That said, it's not necessary you understand everything in this post in order to make good decisions about health insurance. But, we believe every business leader should understand a little more about the way they spend healthcare dollars than they do now.

The Central Limit Theorem is a result in statistics that says that the sum of a bunch of random variables will (under certain loose conditions) approximate the normal distribution. For simplicity, we will talk here only about the case where the variables being summed up are themselves independent, identically distributed, and with finite variance. Interestingly, the theorem holds in more general situations but with more stipulations that are beyond the scope of this blog post. I'll first explain the statement of this theorem (mathematically), then explain some of its intuition and lastly give a proof.

What does it mean when a provider says that they "accept insurance"? We see this all over the place, in signs outside of urgent care to advertisements in commercials for prescription drugs. Companies in the digital health community spend a lot of time fretting over whether their product will be "accepted by insurance". Physicians hoping to open their own practice will say that one of their greatest challenges is figuring out how to accept insurance. When patients go to get care, their top priority is usually going somewhere that "takes my insurance". People will suffer long phone wait times, horrible user interfaces, and limited selection of providers just to make sure that they go to a provider that "accepts insurance".

Franz Kafka was an esteemed healthcare claims administrator and insurance underwriter in Prague. He wrote stories on the side, at night, when he was tired. I am more interested in his insurance affairs. His second job out of college was working for Vienna. Specifically, Kafka worked for the Workmen's Accident Insurance Institute, which administered workers compensation claims in the Austro-Hungarian province of Bohemia. Workers compensation was a new social innovation in Kafka's time. It was invented to essentially mitigate strife when people got injured doing dangerous industrial jobs.

So many people in healthcare, including myself, just sit behind a computer and pontificate about how to make the healthcare system better. I wanted to escape this, even just for a day, so I asked my sister if she could show me around. My sister has BSN in nursing and two years of experience at Fairview Health as a acute care nurse.

I grew up thinking that actuarial science was one of the most advanced things you could do with math. This was before AI was really a thing and before anyone told me what software engineering salaries were like out in the wild. But...I didn't know any of this in high school so I took some actuary tests since they were a fun way of learning probability.

Medical Billing / Coding is a job that (high-level) involves turning procedure notes about diseases and interventions into ICD and CPT codes (respectively). The ICD-10 system was created by the WHO and modified by the CDC. The CPT code system was created by the AMA. HIPAA (1996) mandates the use of both code systems. There are a few certification routes to get certified in billing. The certification is not mandated, but expected for many jobs. Russell is pursuing certification. There are a few commonly accepted certifications, but the most recommended first step is CPC ("Certified Professional Coder"). Here are his steps:

About: this is meant to be a collection of links and external resources relating to TPA stuff, but also a sort of memo about what self-funded insurance is and how it works. Most people have no idea how self-funded insurance works. There are two categories of insurance, fully-insured and self-insured. These categories relate to health insurance, but also other types of insurance companies might buy for employees (including workers' comp, vision, dental, etc) There are four reasons to go self-insured: