What Happened

Amol Avasare, Head of Growth at Anthropic, shared on Lenny's Newsletter how Anthropic scaled from $1B to $19B ARR in 14 months. A key driver: the team built an internal tool called CASH (Claude-Automated Something-or-other for growth) that uses Claude to automate growth experiments. Avasare also argues that the PM-to-engineer ratio may need to flip—more PMs than engineers—as AI makes each engineer dramatically more productive. The team runs a 70/30 split favoring big bets over incremental tests, the inverse of most growth playbooks.

Solo Founder Angle

You don't need a growth team to apply these principles. Here's how a one-person company can replicate the CASH approach today:

  • Use Claude Projects to store your growth hypotheses, past experiment results, and user activation data. Ask Claude to generate ranked experiment ideas based on your specific funnel metrics.
  • Automate experiment write-ups: prompt Claude to produce a structured experiment brief (hypothesis, metric, success threshold, rollback plan) for each idea before you build anything.
  • Apply the 70/30 rule: allocate 70% of your weekly build time to one high-variance bet, 30% to low-risk improvements. Document the split explicitly so you don't drift into pure maintenance mode.
  • Use Claude Code (code.claude.com) to ship the experiment faster—write the A/B test logic, analytics hooks, or onboarding flow changes without a dedicated engineer.

Why It Matters for Indie Builders

Activation—getting users to their first value moment—is the highest-leverage growth problem Avasare identifies. For solo founders, this is even more acute because you can't rely on a sales team to rescue poor onboarding. AI tools now let a single person instrument, run, and analyze activation experiments that previously required a 3-5 person growth team. The CASH example shows that automating the experiment pipeline itself (not just individual tasks) is the next frontier. Solo founders who build even a lightweight version of this workflow—Claude as experiment generator and analyst—gain a structural advantage over competitors still running manual growth reviews.

Action Item This Week

Open a new Claude Project and paste in your current onboarding flow steps plus your latest activation rate. Prompt: "Generate 5 experiment hypotheses ranked by expected impact on activation, each with a measurable success metric and a one-sentence implementation description." Pick the top-ranked idea and timebox 4 hours to ship a version of it using Claude Code.