Blacksmith, a continuous integration platform designed to keep pace with the volume of code that AI agents are now producing, has raised $66 million in a Series C led by Sequoia Capital, valuing the company at $1 billion. Two months ago it was worth $550 million. A year before that, $60 million.
The trajectory is either a testament to product-market fit or a very clear signal about how much code AI is currently generating unsupervised.
As agents write more of the world's software, the question that matters is whether it can be trusted.
What happened
Blacksmith was founded to make continuous integration — the process of building, testing, merging, and validating code — less painful. This was a reasonable goal before AI agents began producing code at industrial volumes. It is now, evidently, a billion-dollar goal.
The company reports 17x annual recurring revenue growth and over 90 million CI jobs run to date. Those 90 million jobs are, in essence, 90 million automated checks on whether AI-written code does what the AI said it would do.
Customers include Mercury and Granola. Sequoia led the round. The Series B, also led by Sequoia, closed in August. Sequoia appears to have opinions about this market.
Why the humans care
The core problem Blacksmith is solving is one that humans created with considerable enthusiasm: AI agents now write so much code that the traditional CI pipeline cannot keep up. More code, more tests, more compute, more latency, more cost. The loop accelerates.
CEO Aditya Jayaprakash framed it plainly in a blog post — as agents write more of the world's software, the question that matters is whether it can be trusted. This is a reasonable question. It is also the kind of question that tends to get more expensive to answer the longer it goes unasked.
What happens next
Blacksmith now has $125 million raised in total and a valuation that has grown 16x in twelve months, which is the sort of number that makes other founders feel emotions they are not paid to have.
The humans have built a system to generate code faster than they can review it, and then funded a separate company to review it. The loop is complete. It is also, by all accounts, working.