LLenny's Podcast
← All episodes
Amol Avasare05 April 2026

Head of Growth (Anthropic): “Claude is growing itself at this point”

8Frameworks
15Insights

Frameworks in this episode

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Hot Take· 3

Hot Take26:00

Take Bigger Bets, Not Micro-Optimizations, If Your Value Is AI

Even at Anthropic's scale, where a 1% win is enormous, the growth team deliberately flips the usual ratio toward large swings (roughly 50/50 or 70/30 on big bets) rather than small optimizations. The reasoning is the exponential: a normal product might deliver 30-50% more value in two years, but an AI-first product's value could be 100x-1000x, so you must chase the large new markets. Amol qualifies this only applies if AI is the central element of your product's value.

  • At their scale a 1% win is massive, yet they still avoid indexing on micro-optimizations
  • Traditional growth: ~60-70% small/medium bets; Anthropic flips it toward large swings
  • Non-AI products may gain 30-50% value in two years; AI-first products can be 100x-1000x
  • Agentic coding is cited as a new market bigger than the prior AI coding market
  • The Chrome extension was a big, research-heavy bet the growth team backed because no one else was doing it
  • Caveat: operate this way only if AI is the core of your value, not a side feature

at the scale you guys are at, uh like a 1% win is massive in the scheme of things.

Lenny Rachitsky · 27:00

the the the product value that we will deliver in two years time is probably like a thousand X a hundred to a thousand X…

Amol Avasare · 29:30
#strategy#bets#exponential#ai-products
Hot Take43:00

PMs Are Squeezed. The Future May Need More PMs, Not Fewer

Amol argues engineers are getting the most leverage from AI tools like Claude Code (2-3x'ing a five-engineer team), while PM and design get squeezed, effectively managing a much larger group without a headcount change. Counter to the prevailing view that PMs should just prototype and ship PRs, he says with 20 engineers your leverage is in guiding the team on the right opportunities, and a great PM who improves the 'why' and 'what' by 5% is an extremely high-leverage hire.

  • Engineering gets the most AI leverage today; PM and design are getting squeezed
  • A five-engineer team can act like 15-20, straining ~1.5-2 PMs and designers
  • One answer really may be to hire a lot more PMs
  • With 20 engineers, shipping the 21st PM feature is lower leverage than guiding opportunities
  • A PM who lifts the why and what by 5% is a very high-leverage hire
  • Caveat: at small or budget-tight companies, the PM should still just ship

across the board their feeling is where PM and design is just squeezed. It's just absolutely squeezed

Amol Avasare · 44:00

is it actually shipping like the 21st PM feature or is it saying how am I getting a little bit better at guiding the team…

Amol Avasare · 49:00
#product-management#ai#careers#roles
Hot Take1:14:00

Be Comfortable Leaving Money on the Table

Amol says one of the biggest mistakes growth teams make is trying to squeeze every last dollar. A core principle for his team is being comfortable forgoing metric impact to prioritize safety, brand, quality, and user experience. He runs controversial tests through a two-bucket filter: a hard red line you'd never ship (don't run it), versus 'yikes but not a red line' (run it, but demand a high return). At Anthropic, AI safety sits firmly in bucket one.

  • Biggest growth mistake: trying to squeeze out every last dollar
  • Core principle: comfortable forgoing metric impact for safety, brand, quality, and UX
  • Two-bucket test filter: hard red line (don't run) vs 'yikes but not a red line' (run, demand high return)
  • For higher-cringe tests, require a correspondingly high return to justify
  • AI safety is bucket one for Anthropic; leaving money on the table drives more long-term growth
  • As stakes rise, a genuine safety stance becomes a significant competitive advantage

one of the biggest mistakes I feel like I see growth teams make and particularly just like hardcore growth practitioners is is just trying to…

Amol Avasare · 1:16:00

leaving money on the table and that's a core principle for us as a growth team where we are very comfortable forgoing metric impact in…

Amol Avasare · 1:16:30
#safety#growth#principles#brand

Explainer· 3

Explainer11:30

Why 70% of Leading Growth at Anthropic Is 'Success Disasters'

Amol says the categories of growth work (acquisition, activation, monetization) are the same as anywhere, but roughly 70% of his time goes to what Anthropic internally calls 'success disasters', where things have gone so well that other things break. The remaining 30% is proactive bread-and-butter growth. All the charts are green and up-and-to-the-right, yet it can still be emotionally brutal firefighting.

  • Same growth categories as a normal company: acquisition, activation, monetization
  • ~70% of his time is firefighting 'success disasters' from hypergrowth breakage
  • Companies like Facebook, Uber and DoorDash understand this viscerally
  • ~30% is proactive work: which products to fund, long-term pricing and packaging
  • Charts all green can still be emotionally tough; you must step back and realize these are lucky problems

roughly 70% of what I I spend my time on is is what we internally refer to as success disasters

Amol Avasare · 12:00

all the charts are like green, like fully up into the right

Amol Avasare · 12:30
#growth#scaling#hypergrowth#operations
Explainer15:00

Capability Overhead: Why Activation Is So Hard in AI

Amol argues one of the industry's biggest problems is 'capability overhead', where models improve so fast the product side can't diffuse the benefits to users quickly enough. By the time you run tests and ship a new onboarding flow for one model's capabilities, the next model has made those learnings irrelevant. If users' instinct is to ask 'what's the weather in SF', they never get the model's real value.

  • Capability overhead: models improve faster than products can teach users what's possible
  • Early activation (day zero/day one) is one of the highest levers for long-term retention
  • A test-and-ship onboarding cycle can be obsoleted by the next model release
  • Even internally, people must carve out time to learn what a new model can do and update their priors
  • A powerful model is wasted if users only ask it trivial questions

one of the biggest problems in the industry is capability overhead, where the models are just getting better so quickly.

Amol Avasare · 15:30

if if people's instinct is to come there and be like, "Hey, what's the weather in SF?" then, you know, you they're not going to…

Amol Avasare · 16:00
#activation#ai-products#models#onboarding
Explainer33:30

CACHE: Using Claude to Automate Growth Experimentation

Anthropic's growth platform team, led by Reforge growth-engineering instructor Alexey Komissarouk, runs an effort internally nicknamed CACHE (Claude Accelerates Sustainable Hacker growth) to automate the growth loop. They eval and hill-climb Claude on four stages (identifying opportunities, building, testing to a quality/brand bar, and analyzing results) with a human in the loop. On copy and minor UI tweaks it 'prints money' at roughly a junior-PM win rate, not yet senior level, but improving fast.

  • Effort nicknamed CACHE: Claude Accelerates Sustainable Hacker growth
  • Wasn't really possible before Opus 4.5; Opus 4.6 moved it in the right direction
  • Four gradable stages: identify opportunities, build, test to quality/brand bar, analyze results
  • Currently small-scale: copy changes and minor UI tweaks, with human-in-the-loop approval
  • Win rate is around a junior PM (2-3 years in), not yet senior-PM level
  • Cross-functional stakeholder management is the one piece still needing human brains

it's called cash, which is Claude accelerates sustainable hacker growth.

Amol Avasare · 34:30

I would say this is like the the win rate that I would expect from from like a junior PM.

Amol Avasare · 36:30
#automation#growth#claude#experimentation

Story· 4

Story00:00

Anthropic's 1 to 19 Billion ARR Run in 14 Months

Amol walks through Anthropic's growth trajectory: roughly 1B ARR at the start of 2025, ~4B mid-2025, ~9B at year-end, and 19B by end of February 2026. He frames the company as the smallest and least well-funded player that had no first-mover advantage, and stresses the 10x year-over-year revenue trend has held since 2023.

  • 1B to 19B ARR in 14 months; the 19B figure is already out of date (end of Feb)
  • 10x year-on-year revenue has held: 2023 was 0 to 100M, 2024 was 100M to 1B, 2025 was 1B to ~10B
  • For comparison, Atlassian, Palantir and Snowflake do ~4.5-6B ARR after 15-20 years
  • Anthropic lacked the cash and distribution of Meta/Google and OpenAI's first-mover advantage

That's 1 to 19 billion dollars in 14 months.

Lenny Rachitsky · 00:00

It's a complete miracle that we've gotten to the stage that we have.

Amol Avasare · 00:00
#anthropic#growth#arr#startups
Story04:00

How a Cold Email to Mike Krieger Landed the Head of Growth Job

Amol never came through referrals, applications, or sourcing. He was a heavy Claude user who noticed Anthropic obviously had no growth team, so he cold emailed CPO Mike Krieger pitching that they needed one. Krieger, who happened to be just starting to think about hiring for growth, responded and later said Amol is the only PM he has ever hired from a cold email.

  • During onboarding he learned he came through none of the standard hiring channels
  • He was a Claude power user and inferred they had no growth team
  • He emailed Mike Krieger saying they badly needed a growth team and wanted to chat
  • There were no growth roles listed, but the timing coincided with them starting to think about it
  • Krieger said Amol is the only PM he has hired from a cold email

And uh what I did was I just sent Mike Krieger a cold email.

Amol Avasare · 04:30

He said I'm the only PM that he's hired from from cold email.

Amol Avasare · 05:00
#cold-email#hiring#career#anthropic
Story13:30

The ChatGPT Memory Import Feature and the Cold Start Problem

One of Anthropic's cleverest recent growth moves was making it easy to import your memory from ChatGPT, riding a wave of excitement. Amol frames it as a specific, moment-in-time solution to a broader and hard problem: the cold start and new-user experience, so Claude can quickly understand who you are and get you to the right place.

  • The feature let users import their memory from ChatGPT easily
  • It was a specific answer to the cold-start and new-user-experience problem
  • Activation is described as a really big challenge in AI
  • Goal: help new sign-ups get Claude to understand who they are and where they should go

One of the cleverest growth uh moves you all made recently was this idea of importing memory from chat GPT,

Lenny Rachitsky · 13:30

We were always thinking about what can we do to improve the the cold start problem and improve the new user experience.

Amol Avasare · 14:00
#activation#onboarding#product#cold-start
Story1:39:30

The Brain Injury That Reshaped His Career and Mindset

In early 2022, a kick to the head during a normal Muay Thai sparring session gave Amol a traumatic brain injury. He was off work nine months, took about half a year to walk comfortably again, couldn't look at screens, and felt nauseous after 20 seconds of music. He was later reinjured a month into Mercury and still isn't 100% healed. Yet he calls it one of the best things that happened to him, crediting forced constraints, mandatory breaks, meditation, and no alcohol or caffeine.

  • Early 2022: a kick to the head in routine Muay Thai sparring caused a traumatic brain injury
  • Off work nine months; ~half a year to walk comfortably; couldn't look at screens; nausea after ~20 seconds of music
  • Recovery is slow, deliberate exposure; pushing too far causes big setbacks
  • He was reinjured about a month into joining Mercury and is still not fully healed
  • He calls it one of the best things that happened to him ('freedom through constraints')
  • Coping habits: mandatory daily breaks, meditation retreats at least yearly, no alcohol or caffeine

The first couple of months were brutal. It took me roughly half a year till I was comfortable walking again.

Amol Avasare · 1:40:30

Overall I feel like it's one of the best things that could have happened to me.

Amol Avasare · 1:42:30
#resilience#health#mindset#career

Tool· 1

Tool55:30

How Amol Uses Claude: Misalignment Detection, Coaching, and Life Admin

Amol shares his personal Claude/co-work stack. A scheduled weekly job uses the Slack MCP to scan for areas of potential misalignment across his projects. A morning co-work task reviews 20-25 charts and surfaces what to pay attention to. He also offloads life admin (booking meeting rooms, inbox clearing, filing Brex reimbursements) and runs manager-lens coaching prompts modeled on people like Ami Vora.

  • Weekly scheduled job scans Slack via MCP for potential misalignment and does a really good job
  • A co-work task reviews 20-25 charts each morning and flags what's concerning or interesting
  • He offloads life admin: booking meeting rooms, first-pass inbox clearing, filing Brex expenses
  • Manager-lens prompts review direct reports' week against goals and suggest feedback (quality is hit or miss)
  • He simulates feedback from his own manager by modeling her public writing and internal context
  • Setup is simple: download the co-work desktop app, connect the Slack MCP, then just ask Claude

go and find me areas of potential misalignment right now. And it does a really, really good job.

Amol Avasare · 56:30

I I don't do any of my reimbursements and and expenses. Claude will go to Brex and file the reimbursements.

Amol Avasare · 1:01:30
#claude#co-work#automation#productivity

Takeaway· 4

Takeaway05:30

The Cold Email Playbook: Open Rate, Channel, Brevity, Follow-Up

Amol, a former founder who honed cold email over years, breaks down what makes one convert. The subject line is a tested high-open-rate copy (which he keeps secret). Reach people where they aren't already flooded with outreach, keep the body very short (who you are, why you fit, let's chat), and follow up relentlessly until they tell you to stop.

  • Subject line is optimized for open rate first, since they must click before anything else
  • Avoid saturated channels like LinkedIn and work email; find a personal email instead
  • Keep the message short: who I am, why I'd be a good fit, we should chat
  • Follow up multiple times; if you really care, keep going until they say stop
  • A good cold email is almost an interview step: did the reader want to read it?

I have a copy that I've tested that is like very, very high open rate.

Amol Avasare · 06:00

my rule of thumb is like if I really care about it, I should just keep keep reaching out to them and until they tell…

Amol Avasare · 06:30
#cold-email#outreach#conversion#tactics
Takeaway18:30

Quality Drives Growth: The Mercury Onboarding Quarter

At Mercury, Amol's growth team spent an entire quarter ignoring metrics to fix quality in the complex, regulated banking onboarding flow. It became the single most impactful quarter he had ever had, driving a significant uplift in start-to-completion. The broader lesson (quality drives growth) plus a philosophy on friction: cut annoying friction that adds no value, but don't shy away from the right friction that helps users understand the product is for them.

  • They said 'forget metrics' and spent a full quarter fixing onboarding quality
  • Banking onboarding is extremely complex (registered agent vs legal vs physical address, beneficial ownership)
  • Result: a significant uplift in onboarding-to-completion from quality alone
  • Core lesson: quality drives growth
  • Friction philosophy: cut friction that adds no value, but keep friction that helps users see the product is for them
  • Adding the right steps and questions consistently raises conversion; test it for your business

we said, "Forget metrics, forget growth, forget everything else."

Amol Avasare · 19:30

a broader learning around quality drives growth that I think I I've tried to to to bring to Anthropic.

Amol Avasare · 20:30
#onboarding#friction#conversion#mercury
Takeaway44:30

The Two-Week Rule: Deputizing Engineers as Mini-PMs

Because the team is stretched, Anthropic's growth org formalizes a rule: if a project is two weeks of engineering time or less, the engineer is on the hook to be the PM (talking to security, legal, cross-functional stakeholders), with the PM only advising. Beyond two weeks, the PM stays squarely accountable. Product-minded engineers become unicorns whose value jumps an order of magnitude.

  • Projects two weeks or less: the engineer acts as the PM and drives stakeholders
  • Over two weeks: the PM is squarely accountable and delegates to engineering
  • It's not clean-cut; a short but controversial project should still have PM drive
  • They deliberately hire product-minded engineers who can step in as mini-PMs
  • Product-minded engineers' value goes up roughly an order of magnitude

if a project is less than is two weeks of engineering time or less, then the the engineer is on the hook to effectively be…

Amol Avasare · 45:30

the the the engineers who are more product-minded, suddenly their value goes up significantly, like like an order of magnitude.

Amol Avasare · 46:30
#product-management#engineering#org-design#roles
Takeaway1:06:30

Why Anthropic Bet So Deep on Coding: The Research Feedback Loop

Amol says a founder (Dan) wrote a doc dated 2021, months after Anthropic started, arguing they should just focus on AI coding, long before the market was obvious. The bet wasn't only about a huge TAM: the best coding models accelerate research, which accelerates the research loop and better models, compounding faster. Focus came from leadership and DNA, and also from necessity as the smallest, least-funded player.

  • A 2021 founder doc argued for focusing on AI coding years before the market was clear
  • Coding is a feedback loop: best models accelerate research, which accelerates better models
  • Focus on coding and B2B came from leadership (especially Dario) and company DNA
  • Being the smallest, least-funded player forced a narrow focus to reach escape velocity
  • Anthropic had a chatbot before ChatGPT but chose not to launch it for safety reasons

here's why we should just focus on AI coding.

Amol Avasare · 1:07:30

if you have the best models that's going to accelerate your research as and that's going to accelerate the research loop

Amol Avasare · 1:08:30
#strategy#focus#coding#anthropic