Systems thinking provides a framework for understanding complexity in multivariable, nonlinear environments. These systems, like the weather or stock markets, are notoriously difficult to predict because they often remain stable for long periods before a single shifting variable triggers a significant change in behavior. Relying on linear models or single variables is risky because the consequences are often layered across first, second, and third derivatives.

Complex systems are multivariable, nonlinear systems. And multivariable nonlinear systems are very hard to predict. They can behave one way for a long time, and then one variable can switch and they can behave another way. There's consequences that can be first, second, third derivative. And you can't just think with a linear model or just think one variable because things can go way off the path.

Ignoring the broader system can lead to unintended outcomes. For example, a dating site once lengthened user profiles to increase engagement. While the metric improved initially, they discovered months later that the change actually hurt overall conversions. Bill notes that focusing too heavily on a single metric can blind a person to the long-term health of the entire system.

You just got to be really conscious of the consequence and not get too deterministic about a single metric or a single variable and know what's important and what's on top.

Bill started his career on Wall Street and built a foundation by reading the works of Peter Lynch, Burton Malkiel, Warren Buffett, and Howard Marks. While these authors often focus on traditional stocks, their principles translate to venture capital. Bill learned from investor Bill Miller that value simply means an asset is underpriced relative to its future worth. This perspective allowed Miller to be a major shareholder in Amazon while still identifying as a value investor. Understanding finance is vital for venture capitalists because Wall Street is the eventual buyer of the companies they build.

Value just means that the asset is underpriced relative to what you think it will be worth in the future. I have always thought of Wall Street as the buyer of the product that venture capitalists create. Even if we are starting at a very early place, you are still thinking about when this thing grows up, is it going to be something they are excited about?

Mastering the history of a field is a powerful differentiator. Bill points to John Lasseter of Pixar, who studied classic cartoons to understand the craft of animation. Similarly, world chess champion Magnus Carlsen knows the deep history of chess games. Even Picasso was a master realist painter at age 14 before he explored cubism. Most people only skim the surface of their industry today. Those who study the masters of their field stand out in interviews and demonstrate a genuine passion for their work.

I think more people would benefit by studying the history of whatever field they are in. If it is tedious to learn that, it is probably not the right passion. You are not in the right lane.

Obsessive learning is a defining trait of successful entrepreneurs. They stay on the technological edge because that is where disruption happens. Bill notes that while understanding history provides a valuable frame of mind, mastering the moving edge is what allows individuals to take market share. This was true during the mobile phone wave and remains true with AI today.

Entrepreneurs are going home at night and reading everything they possibly can because the edge is moving. They need to be a top 1 percentile person that understands this new thing that is happening.

Many people underestimate the capabilities of AI tools. Instead of just asking for a list of items, users can ask the model to analyze, rank, and provide pros and cons in a single prompt. Bill uses a variety of models for different purposes. He uses Gemini for restaurant data because of its access to reviews. He prefers ChatGPT for its memory features and project structure. Other specialists prefer Claude for coding or Perplexity for finance research.

Regulation can sometimes act as a protective barrier for large companies. Some industry players ask for regulation because it creates an expensive hurdle for new competitors. In contrast, the open-source movement in China is creating a highly dynamic environment. Bill uses an analogy of two agricultural societies to explain this. In one, farmers only trade goods. In the other, they are forced to share best practices. The society that shares information will always evolve faster.

The competitive dynamic in China is more intense. Because it is more intense, everyone has chosen to go open source. That creates a system that is capable of innovating far faster than the competitive system we have here.

The debate over whether AI startups are mere wrappers or sustainable businesses is still ongoing. Bill suggests that startups in fields like law create value through specific workflows and data moats. They ingest vast amounts of case law and build proprietary databases that general models might struggle to replicate. This mirrors the history of Microsoft. Microsoft started with an operating system and gradually moved up the stack to absorb applications like word processors. If general models eventually become near-sentient, specialized vertical models may become unnecessary. For now, specialized workflows provide a defense.

I just don't know that you then switch that to ChatGPT as they climb up the stack. Microsoft eventually moved up the stack from the OS. That could happen here. We are going to see how it goes.

A significant challenge for the future of AI is the potential scarcity of training data. Bill describes the current state as painting in the corners. This means most available information has already been used. To improve, companies are hiring experts for thousands of dollars an hour to fine-tune models on complex questions. This approach may eventually hit a limit at the edge of human knowledge. Yann LeCun argues that LLMs have inherent limits because language cannot capture everything. This explains why they often struggle with mathematics.

The next version of AI is not LLMs. It is broader than that. We are going to run into an asymptote with these because they are language-based and there is just a limit to what you can capture with language.

Models like AlphaGo have demonstrated the ability to innovate. They make moves humans never imagined. However, these successes occur within constrained environments. In a game, a computer can search every possible path. The real world is not a constrained system. It contains an infinite number of paths and random variables. This complexity makes reaching full autonomy in self-driving cars difficult. Corner cases and human behavior present challenges that are impossible to fully anticipate. Pedestrians might even test the sensors by jumping in front of the car because they know it will stop.

A non-consensus viewpoint involves the current tendency to vilify China. Having spent significant time there over two decades, it is difficult to adopt the mindset of vilification common in Washington and Silicon Valley. The United States represents only a small fraction of the global population. When people speak of American exceptionalism, they rarely consider how the other 95% of the planet perceives that claim.

The scale of current investment in technology is staggering. Five years ago, it would have been hard to believe that the largest tech companies would take massive amounts of free cash flow and spend it almost entirely on capital expenditures. This shift happens because the investment community now strongly believes in power laws and increasing returns. When startups prove they can grow based on their existing footprint or user base, they often end up being worth far more than anyone predicted. This belief makes investors more willing to take extreme risks.

The venture capital community as a whole is getting more risk-seeking and taking on more risk because of their knowledge of how things have played out in the past.

We see a trend where the losses required to reach profitability are scaling up. Amazon lost a few billion dollars before becoming cash flow positive. Uber lost around $15 billion. The current generation of AI companies will likely see much larger numbers. This environment is further complicated by circular deals. In these arrangements, a cloud service provider gives billions of dollars to an AI company. That company then spends that same money back on the provider's services. If the money were not given, it would not be spent. This creates inflated growth that might not exist in a more constrained environment.

If you were in a more constrained environment where you didn't do that, things wouldn't be growing as fast. You inflate what's happening.

High burn rates are a direct measure of risk. A decade ago, burning $1 million a month was considered very risky. Today, some companies burn over $100 million a month. When a company is this aggressive financially, it becomes very difficult to determine the true unit economics of the business. Success often leads to preemptive funding offers, which encourages even higher spending and further complicates the financial picture.

The role of retail investors in funding startups is often discussed, but access to capital is rarely the primary constraint for high-growth companies. Instead, the focus is on how retail participation influences pricing and market dynamics. In the public markets, retail interest can drive valuations to levels that institutional investors find difficult to understand. Tokenizing private assets could introduce similar dynamics, but it carries significant risks. Without the regulation and financial disclosure required in public markets, tokenization often leads to extreme speculation and manipulation.

One of the reasons they are staying private is so you don't have that dynamic. When they do liquidity events for their employees, they sit down with a handful of investors they trust and they negotiate a price. It is done on a one-off basis.

Private companies often choose to remain private to shield themselves and their employees from the chaos of fluctuating market caps. While the underlying value of a private company likely shifts constantly, the lack of a public ticker means these changes are not recorded or visible to the workforce. This lack of visibility is a strategic benefit for operators. When stock prices move erratically, it creates anxiety for employees who are also owners. Attempts by platforms like Robinhood to open private shares to retail investors have met legal resistance from companies seeking to maintain control over their valuation and investor base.

The current IPO process is unfairly structured because bankers manually select prices and shareholders. This system functions as a controlled oligopoly that ignores the efficient logic of supply and demand. If we designed the system today, it would look more like an anonymous auction or an initial coin offering. Bill explains that tokenization could eventually disrupt this power grab by making share allocation more transparent.

If you took a freshman computer science student and a freshman finance student and said imagine how a company should go public, they would match supply and demand anonymously like you would in any auction. With tokenization, no one would invent this thing where you cherry-pick your best customers and give them this sweetheart price. No one would do that.

The American banking system suffers from regulatory capture. This prevents the government from implementing instant bank-to-bank transfers, which are already common in countries like the UK, Australia, and Argentina. While other nations use systems like Pix for immediate payments, the US relies on credit cards with high fees and slow settlement processes. Stablecoins provide a digital alternative to these traditional rails. A stablecoin like USDC is backed by US Treasuries but operates on crypto networks. This allows for nearly instant transfers with minimal fees.

In America, if I want to send you 50 dollars digitally, I have to go through ACH, which is 3-day settlement. This is part of the regulatory capture. I can wire to you same day, but it costs me 25 dollars and I have to fill out a page of forms. There is no reason that it should be this way.

Because the regulatory environment moves slowly, stablecoins may become the primary way to bypass these inefficiencies before the government can provide a modern solution.

Visa and Mastercard maintain some of the highest operating margins in business history, often reaching 60 percent. These companies function as a duopoly created by the banking industry, which remains invested in the current system despite its inefficiencies. There is no fundamental reason for payment processing to cost two or three percent. This legacy structure is increasingly vulnerable to disruption.

There is zero reason why it should cost 2 or 3 percent, just zero. And it will change.

The landscape in China offers a glimpse of a more efficient alternative. Because the Chinese government prioritized easy money transfers, Alibaba and Tencent built digital wallets that allow for instant transactions via QR codes. Whether buying a hat from a street vendor or a car from a showroom, consumers use WeChat Pay or Alipay for everything, bypassing the three day settlement periods common in the United States. In the US, the slow rollout of projects like FedNow has created an opening for crypto and stablecoins to challenge the status quo.

Established institutions like Moody's also face questions about their competitive moats in the age of AI. While AI might eventually provide debt analysis that equals or surpasses human capability, firms like Moody's derive their primary power from being the industry standard. They act as a trusted watermark for the market. However, the same pressure applies to companies like ISS that advise shareholders on voting. Bill notes that the rapid advancement of technology means that almost every established business model is currently up for grabs.

Proxy advisory services have become problematic in the United States. This is largely due to the rise of index funds. Large funds like BlackRock do not have the resources to evaluate every shareholder vote. They rely on advisory services to tell them how to vote. However, these services often operate with a conflict of interest. They use secret methods to score companies. Then, they charge those same companies fees to learn how to improve their scores. This system functions more like a heist than a service for shareholders.

These advisory services often focus on risk mitigation and strict rules instead of shareholder value. Bill points to the compensation package for Elon Musk at Tesla as a key example. The package required the stock price to increase significantly before the CEO received any payout. Bill views this as an ideal arrangement for shareholders. If the stock goes up, the CEO makes an obscene amount of money, but the shareholders also win. Most advisory services disagree. They see the large headline numbers as a negative and vote against them.

The Tesla case is a great example. I would agree to that type of package for every company I've ever worked with. It basically says you don't make money unless the stock goes way up. If your stock goes way up, you make an obscene amount of money. I would do that deal over and over again.

The dominance of passive investing creates other challenges. Because index funds own such a large percentage of shares, their votes carry immense weight. Bill suggests it might be better if these funds simply did not vote. This would allow active shareholders to have more influence. Some believe the decline of active investors makes it easier to find an edge in the market. Still, beating the S&P 500 remains incredibly difficult. Many venture funds have even failed to outperform the Nasdaq 100.

Successful founders often share three key traits. These are storytelling, product instinct, and absolute determination. Storytelling is a critical skill because founders must constantly recruit employees, raise money, and close deals. Leaders like Jeff and Tobi are excellent at describing their mission in a way that makes people want to follow them. Writing is also a powerful way to refine one's thinking. Jeff uses six-page memos at Amazon because writing a cohesive argument forces a person to find and fix any gaps in logic. For investors, sharing these insights publicly acts as a magnet for new opportunities. Bill learned the craft of writing by studying a style known as the New Journalism. He studied writers like Malcolm Gladwell and Michael Lewis to see how they used long-form non-fiction to impact readers.

I just find it super powerful that someone can maybe put together 20 pages that really impacts you in a certain way. If you have to write it out and make it stand alone and be cogent, you will think through more of the problems. It will be more cohesive, and you will figure out the loose ends and tie them up.

Product instinct is another essential quality. It is very hard to teach. If a person does not have a product-first mindset, they rarely develop it later. Great founders also have intense determination. When Jeff evaluates entrepreneurs, he looks for people who will pursue their vision no matter what happens. This level of conviction is a hallmark of great leadership.

Scaling Uber revealed challenges that standard business school cases do not cover. The ride-sharing market had winner-take-all dynamics. This led to a battle of capital. Companies had to spend money aggressively to stay in the game. Burn rates were higher than anything seen in public markets before. There were no mentors to call because this level of spending was unprecedented. This mega burn strategy has now moved into the AI sector. The spending there is even larger.

There was no HBS case study. You could take the board members from the top ten companies and they would have never been in that situation before. There was no one to call. There was no mentor to find. It was harrowing to recognize that.

Benchmark's success was driven by its radical equal partnership structure. Unlike traditional hierarchical firms where senior partners take more money and credit, Benchmark gave all five partners equal economics and power. This structure eliminates political overhead and annual compensation reviews. It allows the team to focus entirely on the success of their companies rather than internal competition. Bill notes that this model encourages senior members to support new arrivals, as everyone shares equally in each other's success.

The founders decided at Benchmark that they were just going to make it equal, an equal partnership, and there's no lead partner, there's no king, there's no president, there's just 5 equal partners. It really encourages development of the new people that come in because I'm going to take an equal part of their success when they start delivering.

An equal partnership lacks a central CEO, which can make it difficult to scale or manage operational tasks like maintaining a website. Bill explains how this led the firm to eventually adopt a simple one-page website, which reflects a unique sense of confidence. Beyond structure, venture capital success often relies on network effects. A proven track record acts as a stamp of approval that attracts more high-quality founders. Younger venture capitalists also have a natural advantage because they can often understand new technologies better than established generalists.

It'd be very easy for a young venture capitalist to know more about what it takes to be successful on YouTube than John Doerr or Mike Moritz or me. You could just go spend 100% of your time on that. The whole industry bends towards youth because it's a hustle business.

As Bill looks toward the future, his definition of success has evolved. After retiring from his dream job in venture capital, he is focused on using the skills he honed, such as synthesizing information and blogging, to address broader societal issues. He hopes to apply these techniques to dent the universe in a new way, moving from professional achievement to contributing to the public good.