All podcasts / Lenny's Podcast / Summary

Head of Growth (Anthropic): "Claude is growing itself at this point" | Amol Avasare

2026-04-05 - 113 min - source - Read full transcript
Lenny Rachitsky (host)Amol Avasare

Key insights

Anthropic went from $1B to over $19B in ARR in 14 months, a scaling rate with no historical precedent.
Amol lays out the quarter-by-quarter trajectory (roughly $1B, then $4B, then $9B, then $19B+ by the recording date) and notes public companies like Atlassian, Palantir, and Snowflake, each 15-20 years old, generate $4.5-6B in ARR total - Anthropic adds that much every few months. He says the number is already stale by the time the episode airs.
ai-native-growth-strategy
For AI-first products, growth teams should shift the mix toward large bets and away from the traditional heavy weighting on small optimizations.
Amol says a typical growth team spends roughly 70% of effort on small-to-medium experiments and 20-30% on big swings; Anthropic runs closer to 50/50 or reversed, because product value from AI is rising by an order of magnitude every couple of years rather than the 30-50% gains a normal SaaS roadmap delivers. He's explicit this only applies when AI is the core value driver, not a bolt-on feature.
ai-native-growth-strategy
Adding the right friction, not removing all of it, is one of the most consistent levers Amol has seen across every growth job he's held.
He cites Mercury's complex but heavily tested onboarding, MasterClass's purchase-flow quiz, and Calm's copycat quiz flow (built by former Anthropic PMs) as examples where friction that helps the product understand and address the user's specific situation drives higher completion and downstream lifecycle value, versus friction that adds no value and should be cut.
ai-native-growth-strategy
Anthropic's growth org is piloting an internal system, nicknamed CASH, to have Claude run the full growth experimentation loop with decreasing human review.
Claude is scored across four stages: identifying opportunities, building the feature, testing it against quality/brand bars, and analyzing shipped results. Amol says this only became viable after Opus 4.5, and Opus 4.6 pushed it further; today it handles mostly copy changes and minor UI tweaks at a win rate he compares to a junior PM's, not a senior one's.
growth-automation
Cross-functional stakeholder alignment is the one part of the growth-experiment loop Amol does not expect AI to replace soon.
He frames it as the hardest problem even hypothetically post-AGI, quoting his head of design's line that 'we will have AGI and it will still be impossible to get six people in a room to align.' Human review is expected to shrink for the other three stages (ideation, building, testing) as models better internalize brand and quality guidelines via skills.
growth-automation
As engineers gain outsized leverage from tools like Claude Code, PM and design bandwidth becomes the constraint, not headcount.
Amol estimates Claude Code makes a five-engineer team function like two to three times that size, while PM and design leverage hasn't scaled proportionally yet - so the same nominal team now effectively manages 15-20 engineers' worth of output with 1-2 PMs and designers, straining both functions company-wide.
pm-role-evolution
Anthropic's response to PM scarcity is a formal two-week ownership rule: engineers act as the de facto PM on anything under two engineering-weeks.
Below that threshold the engineer runs point on cross-functional coordination (legal, security, stakeholders) with the PM only advising if things go off track; above it, the PM stays accountable even while delegating execution. Amol expects more companies to adopt this deputized-engineer model as AI leverage compounds unevenly across roles.
pm-role-evolution
In PM-scarce, engineer-rich teams, the highest-leverage use of a PM's time shifts from shipping features personally to improving the judgment of engineers acting as mini-PMs.
Amol argues that if a company has 20 engineers and 1-2 PMs, the marginal PM feature ship matters less than raising the quality of the product 'why' and 'what' that engineers work from - a single well-placed PM insight compounds across every engineer downstream of it.
pm-role-evolution
Anthropic ships roughly 70-80% of its work with no written PRD at all.
Amol says he is personally averse to documentation and defaults to Slack threads, quick prototypes, or a short cross-functional kickoff meeting for smaller projects; only the 20-30% of work that's high-stakes or genuinely complex gets a real written PRD, built from a personal 'skill' plus a project of prior PRDs.
pm-role-evolution
Anthropic's deep, narrow bet on coding (documented internally as early as 2021, years before the market existed) was driven by a compounding research feedback loop, not just market size.
The logic: better coding models accelerate Anthropic's own researchers, which produces better models faster, which further accelerates coding - a flywheel distinct from simply picking a large addressable market. Amol says this view has since become mainstream across AI labs.
anthropic-mission-and-safety-tradeoffs
Anthropic structured itself as a Public Benefit Corporation and has repeatedly taken commercial hits for safety, including declining to launch an early chatbot before ChatGPT existed.
Amol says the PBC structure removes the fiduciary duty to maximize shareholder value as the sole goal, and cites the pre-ChatGPT chatbot as a case where Anthropic chose not to trigger what it saw as a premature AI arms race - a decision that arguably cost it the first-mover position in consumer AI that OpenAI later captured.
anthropic-mission-and-safety-tradeoffs
A near-fatal-feeling traumatic brain injury from martial arts sparring taught Amol 'freedom through constraints,' a philosophy he says now governs both his growth decisions and his daily routine.
A sparring injury led to nine months off work, an initial inability to tolerate music or screens, and a second re-injury a month into his next job at Mercury. He credits the forced acceptance of hard limits - and a daily meditation practice he maintains even on the busiest days at Anthropic - with making him more effective, and frames the same logic (accept the constraint, don't fight it) as applicable to product and business decisions.
constraints-and-resilience

Books referenced

Media referenced

Companies

Techniques and frameworks

Summary

Amol Avasare, Head of Growth at Anthropic, walks through what it's like running growth inside the fastest-scaling company in history - from $1B to over $19B in ARR in 14 months. He's candid that the growth team can't take primary credit for that trajectory (Anthropic is a model company first), and that roughly 70% of his time goes to firefighting "success disasters" - things breaking because they worked too well - rather than proactive optimization. He landed the job by cold-emailing CPO Mike Krieger with no open growth role posted, a story that sets up a recurring theme: Anthropic operates with far less process and structure than its valuation would suggest, shipping 70-80% of work with no written PRD and relying on Slack threads and quick kickoffs instead.

The core strategic argument is that AI-first products should invert the usual growth playbook. Where a typical team spends 70% of effort on small, high-conviction optimizations and 20-30% on big swings, Anthropic runs closer to 50/50, because the exponential growth in model capability means the product's total value two years out could be 100-1000x what it is today - a differential no amount of micro-optimization can capture. Alongside that, Amol repeatedly returns to the idea that friction is not inherently bad: onboarding steps that help the product understand a user (tested at Mercury, MasterClass, Calm, and Anthropic alike) reliably outperform friction-free flows, provided the friction is tested and tied to genuine personalization rather than arbitrary process.

A significant chunk of the conversation covers Anthropic's internal effort - nicknamed CASH, for Claude Accelerates Sustainable Hyper Growth - to have Claude run the growth experimentation loop itself: spotting opportunities, building features, testing them against quality and brand bars, and analyzing results, with shrinking human review. Amol is precise about where this currently sits (Opus 4.5 first made it viable; results today resemble a junior PM's win rate) and where he expects it to stall: cross-functional stakeholder alignment, which he doesn't expect AI to solve soon. That capability gap is reshaping the PM role generally - as Claude Code gives engineers 2-3x leverage without a matching leverage boost for PMs and designers, Anthropic has adopted a formal rule where projects under two engineering-weeks default to the engineer acting as the PM, freeing actual PMs to focus on judgment and alignment rather than shipping features themselves.

The episode also digs into Anthropic's mission structure: it's a Public Benefit Corporation, not a standard Delaware C-corp, explicitly freeing it from a sole fiduciary duty to shareholder value. Amol traces the company's narrow historical focus on coding and B2B to a 2021 internal document arguing that better coding models create a research-acceleration flywheel, not just a large market - and recounts that Anthropic had a working chatbot before ChatGPT existed but chose not to launch it over safety concerns, a decision that arguably ceded first-mover advantage in consumer AI. He frames growth trade-offs as either safety-adjacent (off-limits regardless of expected results) or merely uncomfortable (worth testing if conviction is high), and argues that Anthropic's willingness to leave money on the table for brand and safety reasons is itself becoming a long-term competitive advantage.

The conversation closes on a personal note: Amol describes a traumatic brain injury from martial-arts sparring that took nine months to recover from, including a re-injury a month into his next job at Mercury, and the "freedom through constraints" philosophy it left him with - accepting hard limits rather than fighting them, maintained today through a daily meditation practice even during Anthropic's busiest stretches. He connects this directly to how he thinks about growth trade-offs: comfort with leaving upside on the table in service of a longer-term principle.

Notable Quotes

"It's a complete miracle that we've gotten to the stage that we have." - Amol Avasare

"The product value that we will deliver in two years time is probably like a thousand X what it is today." - Amol Avasare

"We are very comfortable foregoing metric impact in order to prioritize safety, in order to protect our brand." - Amol Avasare

"One of the meditation teachers said the true freedom in life is learning how to be content when you don't get what you want." - Amol Avasare

"We will have AGI and it will still be impossible to get six people in a room to align." - Amol Avasare, quoting Anthropic's head of design