Conversation with Anthropic Growth Lead: How to Achieve an Exponential Leap to $19 Billion with Minimal Resources?

Smart growth intentionally creates precise friction to filter and truly understand your users.

It’s April 2026 now. Lenny Rachitsky sits in front of his podcast microphone, listening as his guest—Amol Avasare, the head of growth at Anthropic—recites the numbers across the table. He pauses, just for a beat.

Fourteen months ago, Anthropic’s annual recurring revenue (ARR) had only just broken $1 billion. That was early 2025. One year and two months later, today that number has skyrocketed to $19 billion. Atlassian and Palantir have been at it for almost twenty years, and their ARR is still stuck around $4–5 billion. Anthropic is basically conjuring a Palantir out of thin air every few months.

That kind of momentum makes everyone inside the company feel like it’s somehow not real.

When Amol was still fiddling as a regular user with Claude, he watched this smart AI that—while strong technically—was a bit clumsy when it came to the business loop, and he sighed. He was certain none of these nerds had an actual, serious growth team. Relying on the intuition he’d sharpened by founding companies before, he emailed Mike Krieger, who was then the product lead—yes, the same guy who casually co-founded Instagram.

The email didn’t include a single bit of pleasantry. The gist was: your product is impressive, but you have no growth strategy—you need to talk.

Mike replied. Amol became the only product manager who had made it in via cold email at a company that was growing the fastest, and later climbed all the way up to become the head of growth.

Before the story moves on, here are three unconventional insights Amol brought that go against the grain

• Smart growth intentionally creates precise friction to filter and truly understand your users.

• Traditional, micromanagement-style A/B testing is basically ineffective in the face of AI-level exponential surges. Either bet on the plates that will be able to multiply 1000x in the future, or do nothing at all.

• If your project development time is less than two weeks, for now you don’t need a product manager—have the engineers own the outcome themselves.

The success disasters that keep you trapped

Seventy percent of Amol’s previous growth experience from other companies basically doesn’t apply here. He spends more than half of every day dealing with a mess internal teams call “success disasters.” New models are too popular; traffic instantly punches through the servers, or directly crushes the existing user journeys.

On-screen metrics are all green. The numbers keep surging up and to the right—but the team’s nerves are stretched so thin they’re about to snap.

In this context, the traditional workplace “triangle” starts to fall apart. With tools like Claude Code, engineers’ output efficiency doubles or even triples. Back when one product manager coordinated five engineers, now those same five people’s code output could match what used to take fifteen.

PMs are suffocating under delivery volume—they can’t manage everything. Amol’s solution is so direct it almost feels rough: any engineering work that isn’t going to take more than two weeks—have the engineers act as the PM themselves. They align with stakeholders themselves, and they even go argue with Legal themselves.

If writing docs and running data can be replaced by machines, what is a human PM supposed to live on?

Amol’s strategy is to stop trying to optimize standardized skills and instead look for cross-fire points. Find the weird combination of experiences you personally have. Amol started businesses, worked in investment banking, and almost went into sales—then he blended all of that into commercial growth. In an era where large models can write your copy for you, only people with a bit of cross-domain, off-the-road experience won’t be so easily weeded out by the algorithm.

Growth hackers taken over by large models

Facing the push behind it brought by technology, humans’ reaction speed can’t keep up.

They built a project called CASH (Claude Accelerates Sustainable Hypergrowth). Put simply: letting Claude take over the growth experiments itself.

The system automatically identifies opportunities, formulates hypotheses, directly edits the copy and UI code, ships tests, and finally presents the data results on the table. Amol calculated that this system’s win rate is already no worse than a junior product manager who has been doing the job for three years. As long as you set the brand boundaries, it’s a tireless machine.

If machines can handle writing docs and running data, what are humans doing?

Amol dumps expense reports and meeting-room booking onto AI. Even more interestingly, every week he has Claude scan the public articles and internal Slack chat history of his direct supervisor, Ami Vora. Then he asks the AI: based on what you know about Ami, and on what I did this week—how do you think she’ll respond?

He’s using the large model to get ahead of his boss’s temperament. It sounds a little absurd, but it’s extremely effective.

The head that got smashed in one kick

Amol isn’t the kind of Silicon Valley elite whose life runs smoothly.

In 2022, during a routine mixed martial arts (MMA) training session, Amol’s head took a solid kick.

Life immediately had the brakes slammed on.

For nine straight months, he couldn’t work. In the first few months, besides showering and going to the bathroom, his wife handled everything for him. He’d feel nauseous and vomit after listening to just 20 seconds of music. He’d look at the screen and the world would spin. It took him half a year to learn how to walk like a normal person again. At the time, he even discussed with his wife what would happen “if my life is ruined” and how he should live afterward.

Rehabilitating from a brain injury is an incredibly exhausting process. But it was those days of probing the edge of despair that gave him an unusually steady calm.

Later, when he joined Anthropic and faced growth curves and work noise that could make people’s nerves frazzled, he found he didn’t get anxious as easily anymore. Someone who’s lost even the basic ability to survive won’t be frightened by a few spreadsheets of wildly jumping data.

Because he lost everything, he started craving constraints. Just like Anthropic was broke early on—no Meta cash, no OpenAI first-mover advantage—so they had to stake all their chips on B2B and code generation. The path they were forced into because they had no money or resources ended up letting them dodge the meaningless “burn money” death spiral that big companies get stuck in.

Your anti-derail recovery guide

Finally, here are a few wake-up suggestions for workplace people who are being dragged toward losing control in the AI era.

Learn to leave some money on the table.

Never try to squeeze out that last drop of profit. Sacrificing the user experience or breaking safety boundaries for a few points of conversion rate is extremely short-sighted. Giving up immediate monetization opportunities for safety principles—and holding back by pushing money outward—that restraint is actually Anthropic’s toughest survival wall today.

Cut the power on purpose.

Even if the company is pulling in tens of millions of dollars every day and the countdown to new model releases is everywhere, Amol still forcefully unplugs the power in both the morning and afternoon, and spends ten minutes alone in the office meditation area.

The world doesn’t lack your ten minutes to make a change—but you need those ten minutes to confirm that you’re still alive.

After all, no matter how wild the exponential growth is, it can’t withstand a fragile blood vessel inside your head suddenly rupturing.