Yash DabhadeAI Engineer

← Bookshelf

The Accidental CTO

Subhash Choudhary · Engineering

Read

Takeaways

System design told as a war story. How Dukaan scaled to a million stores with a self-taught CTO — every chapter is a tradeoff someone actually had to make at 2 a.m.

Why I Liked It

Most system design books start from the diagram. This one starts from the outage. You get the pressure first, then the pattern that fixed it, which is how you actually learn these things at work.

The Scaling Path, Roughly

  1. single app + Postgres
  2. read replicas
  3. cache hot reads
  4. queues for async work
  5. shard by tenant
  6. observability everywhere
Each step is added only after the previous one hurt. No premature architecture.

Things I Underlined

  • Replication, sharding, caching and queues all have a cost. Know which problem you are solving before reaching for one.
  • Observability is survival: metrics, logs, traces, SLOs. If you cannot see it, you cannot scale it.
  • Design for failure — retries with backoff, circuit breakers, graceful degradation — instead of hoping nothing fails.
  • Cloud bills are an architecture decision. The move towards owning infrastructure was a cost story, not a hype story.
Not having a CS degree was never the bottleneck. Not understanding the tradeoff was.