Heidelberg AICurriculum

I Built a $1M/y SaaS with Claude Code, Here's How

Nick Saraev ·2026-05-20 ·1 min read

Summary written by us from the video's transcript. The video, and everything in it, is Nick Saraev's work.

Learn how to build a high‑touch SaaS power dialer using Claude Code, from problem definition through idea mining, simulation, and pricing strategies.

Takeaways

  • Identify problems with high client spend and low churn before building a SaaS.
  • Use Claude Code to generate a large pool of ideas, then manually filter for feasibility.
  • Simulate algorithms (e.g., predictive pacing) with historical data and Bayesian optimization before live deployment.
  • Iterate through a loop of definition, ideation, simulation, and real‑world testing; expect failures at any step.
  • Price by simple market tests and aim for high‑touch enterprise customers rather than low‑touch self‑serve models.

Defining a High‑Value Problem

The first step is to identify a clear, lucrative pain point. The team chose an AI‑enabled power dialer for call‑heavy industries because each client has high lifetime value, low churn, and the market had few competitors.

Idea Mining with Claude Code

Claude was prompted to generate hundreds of ideas for improving call pick‑up rates, spawning parallel sub‑agents that suggested mechanisms across algorithmic, behavioral, and regulatory dimensions. Most suggestions were junk, but a few—like temporal propensity models and predictive pacing—proved useful.

Building and Testing Predictive Pacing

Using historical call data, the team simulated multiple simultaneous dials with offset timing to maximize connections. Bayesian optimization refined the optimal offset for batches of calls. The simulation was iterated in Claude Code before a real‑world rollout, revealing practical constraints that required adjustments.

Iterative Development Loop

The workflow repeats: define problem → generate solutions with Claude → human filter feasible ideas → simulate with data → refine parameters via Claude → stress test live. Failures at any stage are expected; real‑world variables often invalidate simulated gains.

Pricing and Market Positioning

Clarivo launched at $250 per seat per month, targeting high‑touch SaaS customers rather than low‑touch self‑serve models. Pricing was adjusted upward based on willingness to pay, emphasizing that simple price testing beats complex AI pricing models.

Framework Choices and Efficiency

The team experimented with many agent frameworks (Hermes, OpenClaw, etc.) but found each added overhead and regression risk. Sticking to the vanilla Claude model yielded more consistent results, as most gains come from the core model rather than wrappers.