创业新路径

如果你上个月还没尝试过Claude Code，是时候再给它一次机会了。如果你已经试过，你就明白我在说什么。感觉就像AGI已经到来。Anthropic的一位工程师写道："Claude创作了《共同工作》。我们人类亲自见面协作。"

Us humans meet in person to discuss foundational architecture and discuss foundational architecture and discuss foundational architecture and product decisions. But all of us devs product decisions. But all of us devs product decisions. But all of us devs manage anywhere between three and eight manage anywhere between three and eight manage anywhere between three and eight clawed instances, implementing features, clawed instances, implementing features, clawed instances, implementing features, fixing bugs, or researching potential fixing bugs, or researching potential fixing bugs, or researching potential solutions. Think about what that means. solutions. Think about what that means. solutions. Think about what that means.

The team developing one of the most The team developing one of the most The team developing one of the most sophisticated AI products in the world, sophisticated AI products in the world, sophisticated AI products in the world, something many of you probably use every something many of you probably use every something many of you probably use every day, is using this AI internally to day, is using this AI internally to day, is using this AI internally to improve their product. I think this improve their product. I think this improve their product. I think this points to a fundamental shift in how points to a fundamental shift in how points to a fundamental shift in how startups operate. Right now, the best startups operate. Right now, the best startups operate.

Right now, the best teams aren't automating one or two teams aren't automating one or two teams aren't automating one or two internal functions. They're automating internal functions. They're automating internal functions. They're automating all of them. Often they're tiny teams all of them. Often they're tiny teams all of them. Often they're tiny teams able to beat huge incumbents thanks to able to beat huge incumbents thanks to able to beat huge incumbents thanks to internal automation. Their leanness is internal automation. Their leanness is internal automation. Their leanness is their superpower. I've been calling their superpower. I've been calling their superpower. I've been calling these startups 20x companies.

Several years ago, my friend Parker Several years ago, my friend Parker Several years ago, my friend Parker Conrad, founder of Ripling and Zenits, Conrad, founder of Ripling and Zenits, Conrad, founder of Ripling and Zenits, coined the term compound startup to coined the term compound startup to coined the term compound startup to describe companies that build multiple describe companies that build multiple describe companies that build multiple integrated products in parallel rather integrated products in parallel rather integrated products in parallel rather than focusing narrowly on one thing. than focusing narrowly on one thing. than focusing narrowly on one thing.

>> The theory of like the compound software >> The theory of like the compound software >> The theory of like the compound software business is that there's this island of business is that there's this island of business is that there's this island of product market fit that's kind of over product market fit that's kind of over product market fit that's kind of over the edge of the horizon line that's sort the edge of the horizon line that's sort the edge of the horizon line that's sort of harder to get to. But if you can of harder to get to. But if you can of harder to get to.

But if you can build, you know, multiple parallel build, you know, multiple parallel build, you know, multiple parallel applications at once, you can get there applications at once, you can get there applications at once, you can get there and and it actually ends up being a much and and it actually ends up being a much and and it actually ends up being a much more powerful type of product market fit more powerful type of product market fit more powerful type of product market fit that's much harder to displace at that that's much harder to displace at that that's much harder to displace at that point. point. point.

>> The 20X company could be an evolution of >> The 20X company could be an evolution of >> The 20X company could be an evolution of Parker's idea, but applied to internal Parker's idea, but applied to internal Parker's idea, but applied to internal automation. Instead of just narrowly automation. Instead of just narrowly automation.

Instead of just narrowly automating a few things like writing automating a few things like writing automating a few things like writing code or handling customer support, code or handling customer support, code or handling customer support, [music] 20X companies build automations [music] 20X companies build automations [music] 20X companies build automations across all internal features, code, across all internal features, code, across all internal features, code, support, marketing, sales, hiring, QA, support, marketing, sales, hiring, QA, support, marketing, sales, hiring, QA, and more. [music] and more. [music] and more.

[music] This makes each of their employees This makes each of their employees This makes each of their employees orders of magnitude more powerful than orders of magnitude more powerful than orders of magnitude more powerful than they would be otherwise. It also allows they would be otherwise. It also allows they would be otherwise. It also allows them to postpone hiring additional sales them to postpone hiring additional sales them to postpone hiring additional sales and ops staff for much longer, keeping and ops staff for much longer, keeping and ops staff for much longer, keeping payroll down and culture from drifting. payroll down and culture from drifting. payroll down and culture from drifting.

The phrase 20x company was actually The phrase 20x company was actually The phrase 20x company was actually coined by the founders of Giga ML, which coined by the founders of Giga ML, which coined by the founders of Giga ML, which builds voice-based [music] builds voice-based [music] builds voice-based [music] customer service agents for enterprise customer service agents for enterprise customer service agents for enterprise to describe how they managed to close to describe how they managed to close to describe how they managed to close Door Dash as a customer going up against Door Dash as a customer going up against Door Dash as a customer going up against incumbents that were literally 20x as incumbents that were literally 20x as incumbents that were literally 20x as large.

When we got Doash as a customer, large. When we got Doash as a customer, large. When we got Doash as a customer, we were approximately like four to five we were approximately like four to five we were approximately like four to five engineers going against players who had engineers going against players who had engineers going against players who had like 100x engineers. So we kind of like like 100x engineers. So we kind of like like 100x engineers.

So we kind of like coined the term like hey we are a 20x coined the term like hey we are a 20x coined the term like hey we are a 20x company because we are able to beat company because we are able to beat company because we are able to beat these much bigger players who are like these much bigger players who are like these much bigger players who are like 20x us by having a better product and 20x us by having a better product and 20x us by having a better product and better numbers. better numbers. better numbers.

>> Giga was able to close Door Dash and >> Giga was able to close Door Dash and >> Giga was able to close Door Dash and several other Fortune 500 companies as several other Fortune 500 companies as several other Fortune 500 companies as customers because of a powerful internal customers because of a powerful internal customers because of a powerful internal agent they call Atlas. So Atlas can agent they call Atlas. So Atlas can agent they call Atlas. So Atlas can basically do anything within the product basically do anything within the product basically do anything within the product which you want to do. So it can use which you want to do. So it can use which you want to do.

So it can use browsers, it can edit the policies, it browsers, it can edit the policies, it browsers, it can edit the policies, it can write code, it can do anything can write code, it can do anything can write code, it can do anything within the product. within the product. within the product. >> Atlas dramatically expands the range of >> Atlas dramatically expands the range of >> Atlas dramatically expands the range of what each engineer can take on. what each engineer can take on. what each engineer can take on.

>> So let's say before Atlas, every >> So let's say before Atlas, every >> So let's say before Atlas, every engineer can probably work on four to engineer can probably work on four to engineer can probably work on four to five problems at once because they are five problems at once because they are five problems at once because they are bottlenecked by all the boilerplate bottlenecked by all the boilerplate bottlenecked by all the boilerplate stuff they have to do for the customers, stuff they have to do for the customers, stuff they have to do for the customers, right? Customers have integrations, they right? Customers have integrations, they right? Customers have integrations, they would have to probably work on that. Now would have to probably work on that.

Now would have to probably work on that. Now with AFD taking care of all the with AFD taking care of all the with AFD taking care of all the boilerplate stuff, each engineer scope boilerplate stuff, each engineer scope boilerplate stuff, each engineer scope is basically doubled or tripled because is basically doubled or tripled because is basically doubled or tripled because they don't need to work on the boiler they don't need to work on the boiler they don't need to work on the boiler plate code. plate code. plate code. >> But Atlas doesn't just accelerate Giga's >> But Atlas doesn't just accelerate Giga's >> But Atlas doesn't just accelerate Giga's engineers. It also acts as a full-time engineers. It also acts as a full-time engineers.

It also acts as a full-time AI employee that works in tandem with a AI employee that works in tandem with a AI employee that works in tandem with a human FTE to service dozens of accounts. human FTE to service dozens of accounts. human FTE to service dozens of accounts. >> Right now, we have only a single human >> Right now, we have only a single human >> Right now, we have only a single human FTE within the company. as hard as it's FTE within the company. as hard as it's FTE within the company. as hard as it's to believe because we have companies to believe because we have companies to believe because we have companies like Door Dash using us. We are in like Door Dash using us. We are in like Door Dash using us.

We are in pilots with multiple Fortune 500s, 10 pilots with multiple Fortune 500s, 10 pilots with multiple Fortune 500s, 10 plus Fortune 500s where each of these plus Fortune 500s where each of these plus Fortune 500s where each of these companies probably have volumes over companies probably have volumes over companies probably have volumes over like 500,000 or a million calls a day. like 500,000 or a million calls a day. like 500,000 or a million calls a day.

It's only been possible because like we It's only been possible because like we It's only been possible because like we have Atlas and this person can primarily have Atlas and this person can primarily have Atlas and this person can primarily focus on just the customer focus on just the customer focus on just the customer relationships, the ask by the customers, relationships, the ask by the customers, relationships, the ask by the customers, taking customer requests and turning taking customer requests and turning taking customer requests and turning them into feature requests and them into feature requests and them into feature requests and everything. everything. everything. >> Building an AI teammate is one approach. >> Building an AI teammate is one approach.

>> Building an AI teammate is one approach. Another is to build an AI integrated Another is to build an AI integrated Another is to build an AI integrated source of truth that gives employees source of truth that gives employees source of truth that gives employees instant context [music] across your instant context [music] across your instant context [music] across your entire system. Legion Health, which is entire system. Legion Health, which is entire system. Legion Health, which is building an AI native psychiatry building an AI native psychiatry building an AI native psychiatry network, is one example of how to do network, is one example of how to do network, is one example of how to do this. Legion built a custom internal this. Legion built a custom internal this.

Legion built a custom internal interface for their care operations team interface for their care operations team interface for their care operations team that lets them pull patient history, that lets them pull patient history, that lets them pull patient history, scheduling availability, insurance scheduling availability, insurance scheduling availability, insurance codes, and a lot more. What we're codes, and a lot more. What we're codes, and a lot more.

What we're showing you right now is an interface showing you right now is an interface showing you right now is an interface that a vast majority of our care that a vast majority of our care that a vast majority of our care operations team uses in their day-to-day operations team uses in their day-to-day operations team uses in their day-to-day work for anything that actually has not work for anything that actually has not work for anything that actually has not been yet automated. And this includes been yet automated. And this includes been yet automated. And this includes everything from, as Arthur's kind of everything from, as Arthur's kind of everything from, as Arthur's kind of showing on his screen, digging into a showing on his screen, digging into a showing on his screen,

digging into a particular patient or many patients particular patient or many patients particular patient or many patients backgrounds, trying to understand where backgrounds, trying to understand where backgrounds, trying to understand where they're at in their journey, if they they're at in their journey, if they they're at in their journey, if they need a new appointment to be need a new appointment to be need a new appointment to be rescheduled, if they're having a rescheduled, if they're having a rescheduled, if they're having a prescription issue, if they've sent us a prescription issue, if they've sent us a prescription issue,

if they've sent us a message that in traditional healthcare message that in traditional healthcare message that in traditional healthcare might have otherwise gotten lost in the might have otherwise gotten lost in the might have otherwise gotten lost in the sea of different communications that go sea of different communications that go sea of different communications that go back and forth between so many different back and forth between so many different back and forth between so many different people. All of that is at a fingertips people. All of that is at a fingertips people. All of that is at a fingertips reach for every single member of our KO reach for every single member of our KO reach for every single member of our KO ops. ops. ops.

>> This single source of truth interface >> This single source of truth interface >> This single source of truth interface has let Legion keep its ops headcount has let Legion keep its ops headcount has let Legion keep its ops headcount [music] flat even as it's dramatically [music] flat even as it's dramatically [music] flat even as it's dramatically scaled revenue. scaled revenue. scaled revenue. >> So we've grown 4x in the past year, but >> So we've grown 4x in the past year, but >> So we've grown 4x in the past year, but we haven't hired uh a single net new we haven't hired uh a single net new we haven't hired uh a single net new person. We've been able to 4x the number person. We've been able to 4x the number person. We've been able to 4x the number of patients.

We're seeing thousands of of patients. We're seeing thousands of of patients. We're seeing thousands of patients a month. We have dozens of patients a month. We have dozens of patients a month. We have dozens of providers, but we have one clinical providers, but we have one clinical providers, but we have one clinical lead. We [music] have one patient lead. We [music] have one patient lead. We [music] have one patient support person and we have one billing support person and we have one billing support person and we have one billing person. And in a typical healthcare person. And in a typical healthcare person. And in a typical healthcare company, those are all departments. You company, those are all departments. You company, those are all departments. You know, those are call centers.

[music] know, those are call centers. [music] know, those are call centers. [music] Those are groups of people sitting Those are groups of people sitting Those are groups of people sitting around desks doing a ton of things around desks doing a ton of things around desks doing a ton of things manually. A third approach is actually manually. A third approach is actually manually. A third approach is actually build custom agents for each employee build custom agents for each employee build custom agents for each employee depending on their workflow and depending on their workflow and depending on their workflow and preferences. Phase shift, which is preferences. Phase shift, which is preferences.

Phase shift, which is building agents to automate accounts building agents to automate accounts building agents to automate accounts receivable, took this approach. receivable, took this approach. receivable, took this approach. >> So phase shift right now is a 12person >> So phase shift right now is a 12person >> So phase shift right now is a 12person team and we're going up against team and we're going up against team and we're going up against companies that have been around since companies that have been around since companies that have been around since 2006 that have hundreds of employees. 2006 that have hundreds of employees. 2006 that have hundreds of employees.

The key to us as a 12 person team moving The key to us as a 12 person team moving The key to us as a 12 person team moving so fast is we bring AI into every so fast is we bring AI into every so fast is we bring AI into every process that is manual and try to process that is manual and try to process that is manual and try to automate as much as possible [music] automate as much as possible [music] automate as much as possible [music] with AI agents. with AI agents. with AI agents.

>> One way phase shift does this is by >> One way phase shift does this is by >> One way phase shift does this is by literally asking its employees to literally asking its employees to literally asking its employees to document the manual tasks they do document the manual tasks they do document the manual tasks they do [music] and then building custom agents [music] and then building custom agents [music] and then building custom agents for them. So what we do is essentially for them. So what we do is essentially for them.

So what we do is essentially say what do you spend your time doing say what do you spend your time doing say what do you spend your time doing throughout the day and we make them throughout the day and we make them throughout the day and we make them document [music] that and then we build document [music] that and then we build document [music] that and then we build quick AI agents quick AI agents quick AI agents >> and this culture of relentless >> and this culture of relentless >> and this culture of relentless automation has let phase shift delay automation has let phase shift delay automation has let phase shift delay hiring for entire functions. hiring for entire functions. hiring for entire functions.

实际上，我们公司至今都避免雇佣专门的音乐设计人员，目前我们大约是一个12人的团队，仅通过利用现有的音乐模式和模板，我们的工程团队就能以此构建所有的前端设计。

这些方法并非互斥。你可以构建AI队友、一个统一的真相来源，并为团队中的每位成员定制专属智能体。采取这些举措的公司正以精简的姿态，创下前所未有的高增长率。

这是新的构建方式，率先领悟这一点的初创企业必将胜出。