去鱼多的地方钓鱼：Andrew Wilkinson 谈无聊生意、AI 代理，以及钱为什么没让他更快乐

Wilkinson says he has been a primary contributor to about 75 projects or businesses, which he admits is not something to brag about, since he has been "an inch deep and a mile wide". The upside is that he has seen almost every business model. His first piece of advice is to pick something you will be interested in, but to avoid the trap of picking what everybody is interested in, like a cool cafe or restaurant.

Those dream businesses are miserable to operate, he says, and brutally competitive because millions of people wake up wanting to start a cafe. By contrast almost nobody wants to start a funeral home, a pest control business or software that helps people fill out government forms. He cites Charlie Munger's advice to "fish where the fish are" and to skip the crowded pond for a small fishing hole with lots of fish and few fishermen.

Lenny Rachitsky questions why small markets are the place to start. Wilkinson says it need not stay small forever, but competition equals lower margins, because the more competitors there are, the lower your prices have to be. Finding a niche where you can make real money is, in his view, probably the most important thing in business.

Wilkinson compares starting a company to the gym: you do not deadlift 300 pounds on day one. A first-time founder announcing the next great AI company or a new bank is attempting something rigorous, regulated and highly competitive, and should take the baby weights instead. His own web design agency, which became Metalab, worked immediately because he only had to know how to build websites and talk to customers.

That early win built a narrative that he was good at business, and then he went wrong by spending the money on a pizzeria, a designer cat furniture business, an online DJ school and a skin cream company, losing it almost immediately. He argues that the first win is what kept him going, so people should choose a business that gives them one.

Lenny raises the fear of building a beast you hate, like a friend who started a coffee shop and ended up just buying milk. Wilkinson says it is a question of scale: a cafe that never grows is just a job, but if a pressure washing business can drive enough leads to hire staff, the founder never has to touch a pressure washer. He calls the mindset he recommends lazy leadership, getting away from the things you hate as fast as possible, or being "Teflon for tasks".

For digital natives deciding between software and physical businesses, Wilkinson says the question is your unfair advantage. His own superpower was sales, which applies anywhere you need to win customers. He gives an example of a friend whose beautiful restaurant barely made money, while the vendors servicing it, such as grease trap and exhaust vent cleaners, were making a killing.

He applies the same approach to his own passions. He loves movies, looked into funding films, and found that about 90% of the time you lose all your money. But the research led him to Letterboxd, a social network for film reviewers with a network-effect moat that he could buy at a fair price, and his past as a barista led to the AeroPress coffee maker company. His rule is to follow your passions, learn industries over time, then find the profitable niche within.

Lenny adds the example of Brian Armstrong, whose mix of computer science, cryptography and economics suited Coinbase. Wilkinson offers another case: a marketing student managing social media for small restaurants could earn about a thousand dollars a month, but pivoting to realtors or wealth managers could justify around $5,000 a month. He did something similar early on by finding a San Francisco job board where startups would pay five times as much for five times less work.

How do you tell whether there are fish but not too many fishermen? Wilkinson quotes Buffett that he is a better businessman because he is an investor and a better investor because he is a businessman. At 20 he thought people would pay a lot to solve his cat furniture problem, but he did not understand how little they were willing to pay. He knows a realtor can make $20,000 to $50,000 on one house, so a lead worth $5,000 is plausible.

Asked for a heuristic, he admits it is mostly gut, built on a latticework of mental models, like an AI model trained on what works. For newcomers he says there is now no excuse: you can ask Claude or ChatGPT to break down the numbers, the regulatory moat and the payroll for, say, a Botox clinic, which he never had.

He then explains why boring is good. His early product Flow was essentially Asana before Asana, in a category everyone wants to build and where people love jumping between tools, and he lost $10 million of his own money competing with venture-backed rivals. He contrasts that with a business he heard of making $30 million a year just filling out forms for government assistance.

Asked what ideas always fail, Wilkinson says his biggest mistakes were entering business models where others had repeatedly failed and thinking he could do it better. A bar he and a friend tried to start taught him that tech people know nothing about business compared with a pizzeria owner, because there are a hundred failure points from the baker's 3 a.m. dough to deliveries. A software company by contrast needs only nerds, computers and internet connections. He also bought a local newspaper in Vancouver and learned that a brilliant management team cannot fix a bad business model.

Lenny summarizes that if there are many dead bodies in a space, something keeps killing them. Wilkinson cites Instacart as an exception, since technology changed after Webvan, but says it is far easier to start an enterprise SaaS company than Instacart, Amazon or Coupang. He returns to his gym analogy: only after years of reps should you try the heavy lift.

On lifestyle versus venture-scale businesses, he rejects the idea that bootstrapped companies must stay small, pointing to Tiny's almost $300 million in revenue. In his view the only difference from a VC is tolerance for burning money, and businesses with brands or network effects grow "like balloons" without it. He praises the task manager Things as a success from the founder's point of view and says its tiny team has not been attacked by VC-backed rivals, because a market worth $5 to $25 million is too small to excite them.

Wilkinson explains how he moved from starting businesses to buying them after realizing the failure rate was stressful and reading about Warren Buffett's quiet, calm life. He looks for a business he cannot mess up, and Tiny generally leaves acquired companies alone, changing only the CEO if the founder wants to leave. What he wants is a moat: a brand with pricing power, or a network effect such as Letterboxd's community, where nobody wants to leave because their friends are there.

Lenny uses Elon Musk's takeover of X as a test of network effects. Wilkinson mentions a third moat, high switching costs like Salesforce, but says he does not love it. Tiny's latest purchase is Serato, the largest DJ software company in the world, which he understood because he used to DJ and which had a passionate user base and deep hardware integration.

On people, his partner Chris says there are no problems, only people problems, and Wilkinson spends about 20% of his time filtering for people he enjoys. He says he can never change someone, that if he wonders even once whether to fire someone he should do it, and that you should hire for what you need rather than potential. With CEO candidates he looks for someone whose plan he already agrees with, because he compares them to an elephant with the owner as the rider.

His main tool is Lindy, where he builds workflows and agents himself. His email setup is a chain of agents: the first archives anything he doesn't need to see, which cuts email by about 20%, another flags what is time-sensitive, and another turns simple decisions into multiple-choice questions, like answering a lunch invitation, then writes the reply as him. He says this replaced a full-time assistant, and Lenny discloses he is a small angel investor in Lindy.

Other agents add emojis to calendar events, record the city each contact lives in so that he is prompted with people to meet when he travels, and thirty minutes before any meeting send him a Perplexity-researched brief on the person. He says these are things his assistant could never do consistently, and calls it "the world's most reliable employee who costs $200 a month".

He also uses Replit to build web apps and websites that previously would have needed a team of five, and wears a Limitless pendant that records his day so that he can query it as an LLM, for example for promises he made or arguments with his girlfriend. For models he uses Gemini for large inputs like his medical records, Claude for writing and ChatGPT for everything else, including photographing his medicine cabinet so that it can warn him about interactions.

Citing William Gibson's line that the future is already here but unevenly distributed, Wilkinson says we are in the Palm Treo phase: people like him can build agents, but it is not accessible to everyone yet. He expects an iPhone moment within about five years, when you could tell ChatGPT you run a business and have it set up a sales agent that works as a digital employee, as in the movie *Her*.

Even without further progress he sees serious displacement for translators, researchers and some admin and assistant roles, and potentially all knowledge work if the models scale as promised. He notes Dario Amodei said that by 2027 models would be smarter than all PhDs, and while he takes that with a grain of salt, he perked up. Lenny adds that Mike Krieger told him Amodei's predictions have so far been right.

His advice is to get very good with the tools and use them to build wealth, perhaps diversifying into compute and energy. He argues people overestimate short-term and underestimate long-term change, so robotics will lag and there will be a long window to start new businesses. He guesses that outcomes will be either totally fine or terrible, and says that for his own kids, aged five and eight, he just focuses on their being polite and socialized and following their passions.

Wilkinson says that whatever anxiety loop is in your brain does not disappear with a bigger house, more money or a move to Bali. He kept moving the goalposts, from $60,000 a year to a million to being worth over a billion, and remains as anxious as ever. Ten years ago he stressed about the same things at $15 million in revenue as he does now at almost $300 million, and he found that even multibillionaires keep comparing themselves to their peers.

He made a few changes. He reframed money as something that does good, after a wealthy friend told him he gives over 90% to his foundation, and he cut spending because the more stuff and staff he had, the more he felt owned by it. The biggest change, he says, was medication: in 2020 he realized he had watched 20 minutes of a movie while looping on emails, tried a tiny piece of an SSRI, and within days felt that the volume had been turned down on the nasty voice inside him.

He says treating his ADHD moved his brain from Times Square to a quiet library, and argues that people accept Tylenol or antihistamines but fear drugs for conditions that ruin their lives. To manage side-effect worries he used a DNA test to find an SSRI he metabolizes well, and he recommends starting with your doctor.

Wilkinson recounts a baseline cognitive test his doctor requested in his thirties. The neurologist found his long-term crystallized intelligence above average but his working memory in the 10th percentile, and suggested an ADHD evaluation. He resisted, since he was organized and had built a business, but a five-day process confirmed it.

He says about 5% of the general population has ADHD versus about 30% of entrepreneurs, who love new things and describe themselves as unemployable. At work he can delegate and build systems, but at home it shows up in forgetting to take out the garbage three times, which his partner reads as not caring. The diagnosis gave him empathy for himself and made him feel "not broken".

He encourages entrepreneurs to consider getting checked, even by asking ChatGPT to interview them first, and says it is always better to know, since even without medication there are lifestyle changes and ways of explaining your needs to a partner. In the closing lightning round he names *The Laws of Human Nature* and *How to Get Rich* as favorite books and says the motto "Easy choices, hard life. Hard choices, easy life" guides him.