Hello, this is Hamamoto from TIMEWELL. Today I want to talk about something in the tech world that stopped me in my tracks.
A friend recently sent me a summary of a video — Y Combinator CEO Garry Tan speaking about a concept he called the "20x Company." The moment I read it, my back straightened. This is exactly what we at TIMEWELL are trying to do.
But summarizing a summary felt irresponsible. So I went back to the original video, interviews, and primary sources on the companies involved. What I found confirmed that the "20x Company" isn't a buzzword — it's a business strategy that's already being validated in the real world.
In this piece, I'll work through Garry Tan's original words, dig into three real startups, and explain why TIMEWELL is serious about automating everything.
What the "20x Company" Actually Is
On February 14, 2026, Garry Tan published a video on the official YC channel titled "The New Way To Build A Startup." Within 22 hours, it had over 60,000 views and sent ripples through the startup world.
In the video, he makes a direct claim:
"In the AI era, startups are not winning by hiring faster. They're winning by automating every function inside the company as much as possible."
A "20x Company" is one that uses AI to automate not just its product, but all of its internal operations — and beats companies twenty times its size with a small team. This isn't about automating one specific function like coding or customer support. It means deploying AI across every department: marketing, sales, recruiting, quality assurance, everything.
The term itself was coined by the founders of Giga (formerly Giga ML), a YC-backed startup. Four or five engineers beat competitors with a hundred times more headcount and landed enterprise customers like DoorDash. They could win against opponents twenty times their size — so they started calling themselves a "20x Company."
On a personal note: when I heard this, it took me back to my time at Panasonic. When you're inside a large organization, the idea of a team of ten beating an organization of a hundred simply doesn't surface. Adding headcount was the measure of progress; winning budget battles was the work. I wish I could send that video to my past self.
Looking for AI training and consulting?
Learn about WARP training programs and consulting services in our materials.
The Numbers Behind the New Reality
The "20x Company" isn't a gut feeling. In his October 2025 Mixergy interview, Garry Tan put concrete numbers on it:
"During the YC batch, the average weekly revenue growth used to be 2–4%. Now it's 10–20% on average. Before, if one or two companies hit 10% weekly growth, that was remarkable. Now that's the average across the whole batch."
Growth rates have jumped five-fold — and not as an outlier story but as the batch average.
Tan described the phenomenon as "something in the water." That something is AI. His explanation: software used to be a "nice to have" for customers. Adoption priority was low and sales cycles were long. That's changed.
"They see a demo and immediately say, 'This is incredible. I need this right now. When can we start?'"
Software has moved from "nice to have" to "need it right now." That shift is fundamentally changing how fast startups can grow.
He used the HVAC industry to illustrate the structure. HVAC companies spend roughly 1% of revenue on operational software like ServiceTitan, and 5–6% on labor for tasks like phone handling and scheduling. When AI replaces that labor component, the addressable market expands by five or six times overnight. YC-backed startup Avoca is already growing rapidly in exactly this space.
AI penetration in the development process itself is equally striking. In a March 2025 post on X, Tan revealed that for 25% of the YC Winter 2025 batch, 95% of code lines were generated by LLMs. Ninety-five percent. He added that he wanted to make clear this wasn't a typo.
And he noted:
"We regularly see 10-to-20-person YC companies reaching $10M–$20M in annual revenue within 10 to 20 months. This has never happened before in the history of software."
| Metric | Pre-AI | Post-AI |
|---|---|---|
| YC batch weekly revenue growth | 2–4% | 10–20% |
| 10% weekly growth | 1–2 companies, exceptional | The batch average |
| LLM-generated code share | Minimal | 95% for top quartile |
| $10M–$20M ARR timeline (10–20 people) | Typically years | 10–20 months |
Looking at these numbers, I became convinced: this isn't a temporary trend. The structure of startups is changing at a foundational level.
Case Study 1: Giga — AI Teammates That Amplify Humans
Giga (formerly Giga ML), the company that coined "20x Company," is a YC Summer 2023 alumni. Founders Varun Vummadi and Esha Dinne, graduates of IIT Kharagpur, provide voice AI agents that automate enterprise customer support.
Their internal AI agent, called Atlas, is particularly interesting. Atlas isn't a chatbot. It operates browsers, edits internal policies, and writes code — an "AI teammate" that autonomously completes complex tasks that human engineers used to handle.
With Atlas handling routine work, human engineers can focus on creative challenges. Each person's effective scope expanded by two to three times, they report. In a pilot with DoorDash, a single human managed between 500,000 and 1 million calls per day alongside Atlas.
When I first saw that number, I didn't believe it. A million calls handled by one person? But if AI agents process the vast majority of initial interactions and humans intervene only for exceptions, the logic works. A traditional call center operator handles perhaps 50–80 calls a day. One million calls is four orders of magnitude more.
Giga's path is also instructive. They initially developed an on-premise LLM (a model called X1 Large), then pivoted to voice AI agent-based customer support automation. The pivot worked: in November 2025, they raised a $61M Series A led by Redpoint Ventures. According to YC's profile, the team is now 30 people — raising that kind of capital within two years of founding.
| Item | Details |
|---|---|
| Company | Giga (formerly Giga ML) |
| YC Batch | Summer 2023 |
| Founders | Varun Vummadi, Esha Dinne (IIT Kharagpur) |
| Business | Voice AI agents for enterprise customer support |
| Internal AI agent | Atlas (autonomous browser operation, code writing, policy editing) |
| Key customer | DoorDash |
| Funding | Series A $61M (led by Redpoint Ventures) |
| Team size | 30 |
Case Study 2: Legion Health — Data Integration as the Foundation for a Lean Operation
Legion Health, a YC Summer 2021 alumni, is an AI-powered remote psychiatry clinic. It was co-founded by three Princeton roommates — Arthur MacWaters, Danny, and Yash — and operates in Texas.
Their approach centers on consolidating all internal data into a single, AI-accessible source of truth: patient history, scheduling, insurance codes, past communications. Everything that had been scattered across disconnected systems was unified into a custom interface built for the care operations team.
The result: revenue and patient count quadrupled over the past year — without adding a single person to operations. They support thousands of patients and multiple care providers with a team of three: one clinical lead, one patient support person, and one billing specialist.
According to a detailed report from Healthcare AI Guy, Legion Health has an internal rule: "Whatever a human can do, an LLM should eventually be able to read and do." This principle is enforced rigorously. Their AI applications are organized around four pillars: AI patient support, AI CFO for revenue cycle management, clinical co-pilot, and patient personalization.
Automating billing alone improved their profit margins by roughly 12%. The average patient has 5.3 sessions, well above the industry average of 2.9–3 — which means their AI-driven efficiency is also directly improving care quality. AI is making it easier for patients to stay in treatment.
80% of patients use insurance and pay less than $30 out of pocket. The founders describe their approach as the "Tesla model": start with humans at the center, then automate progressively — the same way Tesla started with human-driven cars and moved toward autonomy.
What strikes me most about Legion Health is that they didn't "deploy AI tools." They redesigned their operations from the ground up with AI as the assumption. Layering AI onto existing workflows and designing workflows around AI from the start produce entirely different outcomes. The first gets you 10% improvement. The second gets you 10x transformation.
| Item | Details |
|---|---|
| Company | Legion Health |
| YC Batch | Summer 2021 |
| Business | AI-native remote psychiatry clinic |
| Founders | Arthur MacWaters, Danny, Yash (Princeton classmates) |
| Four AI pillars | AI patient support, AI CFO/RCM, clinical co-pilot, patient personalization |
| Results | 4x revenue and patient growth in one year; zero operations headcount added; ~12% profit margin improvement from billing automation |
| Funding | Seed $6.3M |
Case Study 3: Fazeshift — A Culture of Eliminating Manual Work One Task at a Time
Fazeshift (referred to as "Phase Shift" in Garry Tan's video), a YC Summer 2024 alumni, provides an AI agent for accounts receivable automation. Founders Caitlin Leksana and Timmy Galvin met at Harvard Business School and experienced the inefficiencies of AR firsthand during their previous startup.
Fazeshift's method is deliberately unglamorous. They have employees document their manual tasks. Engineers read those documents and build "quick AI agents" to automate each one — a methodical process of eliminating tedious work, one piece at a time.
This culture has developed in an interesting direction. Fazeshift has no designers. Instead, engineers handle front-end design using a tool called Magic Patterns. Rather than building departments with specialist hires, they extend existing members' capabilities through AI tools — effectively avoiding the need to hire entire functions at all.
Garry Tan wrote on LinkedIn:
"Fazeshift took what required a 12-person accounts receivable team day-to-day and made it so 1 person running software could do the same thing. Their finance team can now focus on higher-value work."
Over 500 billion invoices are sent worldwide each year, and half are still managed in spreadsheets. Unpaid invoices cost billions of dollars annually. A team of twelve was taking on that enormous inefficiency — despite facing competitors with hundreds of employees who had been operating since 2006.
In January 2025, Fazeshift raised a $4M seed round led by Google's AI-focused fund, Gradient Ventures, and was adopted as a Harvard Business School case study. Following the YC program, they've grown rapidly — over 30 customers and monthly ARR growth of 20–30%.
What I find most instructive about Fazeshift is this: "Automation is a culture problem, not a technology problem." Employees honestly document their manual work. Engineers automate it. Whether an organization can build and sustain that cycle is the dividing line between becoming a 20x Company and not.
| Item | Details |
|---|---|
| Company | Fazeshift (called "Phase Shift" in the video) |
| YC Batch | Summer 2024 |
| Founders | Caitlin Leksana, Timmy Galvin (Harvard Business School) |
| Business | AI agents for accounts receivable automation |
| Approach | Employees document manual tasks; engineers build custom agents for each one |
| Results | 12-person AR team's work handled by 1 person; 30+ customers; 20–30% monthly ARR growth |
| Funding | Seed $4M (led by Gradient Ventures) |
Why TIMEWELL Is Committing to Full Operational Automation
We've looked at three cases. Giga extended human capability with its internal AI agent Atlas. Legion Health supports thousands of patients with three people through data integration. Fazeshift automated its way out of needing entire departments.
The approaches differ, but the underlying stance is the same: not "using AI in parts," but "designing the organization with AI as the assumption."
TIMEWELL's vision is to "create the world's #1 infrastructure for human ambition." As a company that helps businesses use AI, we'd have no credibility if we weren't pushing AI to its limits ourselves. Telling clients "let's transform your operations with AI" while doing our own expense reporting by hand would be a contradiction we can't afford.
Honestly, TIMEWELL still has inefficiencies. Inquiry handling, document creation, internal reporting, recruiting — individually small, but cumulatively they consume enormous time.
So we are committing to becoming a 20x Company. Here's how we're approaching it.
First, following Fazeshift's lead, we'll map every manual task across the organization. Who does what, in what steps, taking how long. This inventory becomes the blueprint for automation. It's unglamorous work, but skipping it leads to automating the wrong things.
Second, following Legion Health's approach, we'll unify our internal data — customer records, deal histories, product data, internal documents — into a single foundation that AI can access securely. Even the most capable AI agent is limited by fragmented data.
From there, we'll start automating cross-departmental shared work: expense processing, internal inquiries. Getting every team member to experience working alongside an AI agent early is the first step toward shifting the organizational mindset.
Ultimately, we're aiming for AI teammates — like Giga's Atlas — that can autonomously complete work within defined domains. Humans focus on goal-setting, strategic judgment, and handling exceptions; daily operations are handled by AI. That's the organization we want to build.
Engineers freed from infrastructure maintenance and routine bug fixes can focus on building new features. Sales can delegate prospect list building and appointment scheduling to AI and concentrate on conversations with clients. Marketing can step away from micro-managing ad adjustments and report generation and spend time finding new market insights. This isn't just efficiency — it's a structural transformation that fundamentally unlocks organizational creativity.
The North Star: "10 People, $100 Billion Company"
Beyond the "20x Company," a larger vision is emerging. In summer 2025, YC announced it wants to invest in "the first 10-person, $100 billion company" — an organization generating $10 billion in value per employee.
Does that sound absurd? I thought so too at first. But recall the numbers Garry Tan shared: YC startup growth rates have accelerated fivefold, 95% of code is written by AI, and 10-to-20-person teams are hitting $20M ARR in 20 months. "10 people, $100 billion" is the logical extension of that trajectory.
Imagine a team of ten, each leading a hundred AI agents — those agents running customer support, marketing, and product development around the world, 24 hours a day, 365 days a year. A founder articulates a vision; AI produces a prototype in minutes and deploys it globally.
The correlation between company value and employee count is already breaking down.
TIMEWELL is pursuing this north star not just to maximize valuation, but because realizing our vision of "creating the world's #1 infrastructure for human ambition" requires us to be on the side that breaks convention. AI is ushering in an era where individuals and teams can take on challenges at scales that were previously unimaginable. The company building the infrastructure for that era cannot remain structured like the old world.
I left Panasonic and founded TIMEWELL in 2022. Twelve years in a large organization taught me a great deal, but it also gave me an inescapable front-row view of how organizations slow down as they grow. AI removes that constraint. Small teams can generate impact beyond what large companies can. Proving that is our work.
To borrow Garry Tan's framing: this is the "new way to build a startup." And the organizations that understand this earliest and act on it will be the ones that win.
TIMEWELL intends to be among them.
References
- Y Combinator. (2026, February 14). The New Way To Build A Startup. YouTube.
- Mixergy. (2025, October 17). Garry Tan: Y Combinator Startups Growing 5X Faster - Here's What Changed.
- Tan, G. (2025, March 19). For 25% of the Winter 2025 batch, 95% of lines of code are LLM generated. X.
- Y Combinator. Giga: AI Support Agent for Enterprises.
- Nolan, B. (2025, November 5). Giga raises $61 million to expand enterprise voice AI. Fortune.
- Y Combinator. Legion Health: AI-native psychiatry, built for scale.
- Healthcare AI Guy. (2025, November 20). Company Deep Dive: Legion Health.
- Y Combinator. Fazeshift: AI agent for Accounts Receivable.
- Tan, G. (2024, July 16). Fazeshift reduced the day-to-day need for a 12 person AR team. LinkedIn.
- Business Wire. (2025, January 7). Fazeshift Secures $4MM Seed Round Led by Gradient.
- Inc. (2025, August 1). Here's What Y Combinator Is Looking For in AI Startups Right Now.