Blacksmith Bets That Faster CI Will Be the Bottleneck Breaker for the AI Coding Era

Every time a developer writes code, they have to run a set of checks to make sure nothing is broken. The founders call this process CI, and Blacksmith builds infrastructure to run those checks very fast.

The company says it has grown 20x in the last year. It recently raised a Series C at a $1 billion valuation, led by Sequoia, and raised about $66 million.

The speaker says none of the founders were platform engineers, none had worked in DevOps, and none had written a CI pipeline before. In his view, the reason they were among the few to approach the problem this way was being naive enough, in the right way, not to kill a good idea before it met the market.

With agentic coding tools, developers are writing more and more code than before. But writing 100 times as much code is not very useful if it cannot be shipped at that speed, so the bottleneck has shifted from writing software to validating it.

Validation usually means waiting on CI, which means making sure tests pass and having confidence that what is being shipped actually works.

One speaker says he always wanted to start a company in Silicon Valley. He chose the University of Waterloo for its strong startup program, and Blacksmith is his second application to Y Combinator. The three founders were friends from undergraduate studies, were all in Toronto at the time, and were all engineers.

He says that at nearly every company they worked for, CI universally sucked, because it was very slow and managing this infrastructure was a pain across the board. They felt they had a better way. When they first discussed the idea, they were somewhat skeptical it could work, but their conviction grew as they researched more. His co-founders were database engineers and he worked on search infrastructure.

From the outside, the founders saw that the reason managing CI was so hard was that the shape of the workload is very different from what hyperscalers such as AWS, GCP and Azure are really good at. They describe this as a mismatch that caused the pain.

They concluded the real way out was building a cloud that is extremely good at running CI workloads, which the speaker summarizes as using the best tool for the job.

The founders realized that CI workloads are similar to desktop applications, particularly the video games people run on typical gaming PCs. Those workloads do not care as much about multiple threads of execution, but they care that a single thread runs really, really fast.

When they started, they rented servers from a German data center services company known for hosting Minecraft and Call of Duty game servers. As they scaled, they began owning more of the stack and now negotiate at much larger levels with data center providers, effectively buying megawatts of power rather than individual machines.

The speaker says he invested in Blacksmith twice. He led the company's initial round out of Y Combinator, on a SAFE, and a few months later led its Series A and joined the board.

He found it interesting that Blacksmith was using gaming CPUs, because he had never seen anyone else take that approach to CI. He said it was a shock that someone would put many of these machines in one place. Gaming CPUs are usually found at home in a single big box under a desk, and he described the idea of buying around a hundred of them to compile code as the point where he saw it was different.

The speaker says a large part of making CI fast is skipping redundant work that has already been done. Developers refer to this high-level idea as caching.

A big part of Blacksmith's special sauce, according to the speaker, is moving cache artifacts close to the compute nodes.

The speaker says the first 200 customers were extremely hard to win. Initially the team asked friends working at companies to give them a try, requesting logo rights so the logos could appear on the website and build credibility.

The speaker then personally cold-contacted CTOs, VPs of engineering, and platform engineers to get them to use the product. The earliest customers were small companies like Blacksmith itself, which cared about moving fast and had very little red tape.

The speaker says the only real way to win over engineers is to perform and solve a problem in a way they probably could not solve themselves. He notes that engineers are very discerning and like to try many new tools, so they are sometimes not the stickiest customers.

Without any sales team, the company says it was able to win contracts in the five- to six-figure range because of how easy it made optimizing and running CI as fast as possible. The speaker says switching all of a customer's CI to Blacksmith takes one click.

Langchain was previously running CI on GCP, where developers routinely waited 15 to 20 minutes just to acquire compute for a CI job. Because Blacksmith effectively exposes an infinite pool of compute, the speaker says that wait time dropped to between 10 and 15 seconds, significantly improving how fast Langchain's developers could push code.

A big push for Blacksmith now is to make it the company engineering teams immediately think of when they consider CI velocity or gaining trust in the code they are generating. To that end, the company is rolling out a large marketing campaign in San Francisco and New York, taking over more than 15 subway stations in New York and wrapping a subway train in its bright yellow color.

When the speaker joined, the team was about five people running a small cluster of a couple hundred servers. Today the team is 30 people, 12 of them engineers, and the scale has grown by many orders of magnitude.

The speaker attributes this to a few key decisions. The team builds a lot of bespoke software, including tools to manage fleet growth and bespoke observability at every layer of the stack. It also keeps the product surface area small and manageable, scaling it only with team size rather than giving in to bespoke customer requirements. The payoff, he says, is being able to scale to a very large number of machines and customers with a very small team.

The speaker says one of the largest challenges is keeping up with demand. Every time a new LLM is released, CI usage spikes massively as people rush to test it and ship more code. Because ordering and installing machines takes time, the team has to be technically creative to support that load.

The responses include better algorithms for placing jobs, cleverer ways to pack jobs more efficiently, and more efficient network topologies.

Right after the Series B announcement early in the summer, a power surge in Phoenix took out both the primary and backup cooling systems in one of the biggest data centers Blacksmith operates in that city. Customers were affected by longer startup times for their jobs during the outage window. The speaker said the incident underscored the importance of being geographically distributed.

The speaker says the incident underscored the importance of being geographically distributed.

Blacksmith wants to help developers have confidence in their code changes. One product in development is autonomous QA, which gathers artifacts such as screenshots or video snippets that build confidence in a change.

The speaker says that with the amount of code being produced, developers will not have time to check whether things are working themselves. They will look for evidence, and Blacksmith aims to give them confidence to merge code faster. The goal, he says, is to remove the bottleneck and let developers bring shipping software as close as possible to their speed of thought.