What if the biggest market in developer tools was hiding inside everyone's cloud bill?
In this episode, co-founder and CEO JP Jayaprakash argues that the step after code gets written, not the writing itself, is where the next large infrastructure business gets built.
Before AI code generation, writing code and validating it took roughly the same time. Now writing is a prompt away, and most of the software lifecycle goes to validation. JP's argument is that CI was never a small market, it was an invisible one. Teams spend something like $50,000 a year on a CI orchestrator and something like $2 million a year on the infrastructure underneath it, and all of that disappears into an AWS bill. CI workloads are also nothing like what the hyperscalers optimize for. One push can need a few thousand vCPUs for ten to thirty minutes, then drop to zero. Blacksmith built a cloud that deliberately trades durability for performance.
In this conversation, we cover:
- Why "CI was always a large market, they just didn't know about it," and where the spend actually hides
- What the hyperscalers get wrong about CI: "Lego blocks," bursty workloads, and durability over performance
- Why Blacksmith is "an index on software," whichever model or coding agent wins
- How "agents prompting other agents" produced 100 pull requests overnight at one customer
- Why power, not compute, became the constraint, and what a data center outage taught a 30-person team
JP also explains why he is against a Talk to Sales button with no Sign Up button, how a company can move its CI in under five minutes, why he wants sales removing blockers rather than selling the product, and where he thinks the bottleneck lands once agents write most of the code.
Blacksmith is building a purpose-built CI cloud, running the builds and tests that validate code before it ships, for teams whose code is increasingly written by agents.
