Blacksmith makes testing faster when AI writes more code

Artificial intelligence makes it possible to develop software far faster than before. Tools such as ChatGPT Codex and Claude Code can write code, build websites, create apps and develop new features in minutes. But even though AI can do much of the work, someone still has to check that the software actually works.

This is what the American company Blacksmith has built a business around. The company was founded in 2024, has taken part in Y Combinator's startup program and was valued at one billion dollars in October 2026. Blacksmith has raised 66 million dollars in a new funding round, and more than 10,000 companies use the service.

When testing takes longer than development

When a developer writes code, it must be tested before it is put to use. The same applies when AI writes the code. A new feature can work perfectly on its own and still cause problems elsewhere in the software.

That is why developers use automated tests. Among other things, these can check that websites work, that calculations come out right, and that new changes do not break existing features.

This process is called Continuous Integration, abbreviated CI. It means that code changes are checked automatically and continuously, instead of someone having to carry out all the checks manually.

A well-known tool for this is GitHub Actions. It can start automated tests when someone makes changes in a project. The tests run on computers, often located in large data centers.

The problem is that the tests can take time. Sometimes the developer has to wait several minutes just for a free computer to start the job.

When AI agents produce more and more code, the amount of code that must be checked also grows. Testing can then become a bottleneck that slows down the whole development process.

Blacksmith makes testing faster

Blacksmith has specialized in the computers and infrastructure that run these tests.

The company offers an alternative to the servers used by GitHub Actions, among others. Developers can keep using the same automated tests, but let them run on Blacksmith's infrastructure instead.

Blacksmith has built its systems around powerful processors, fast storage and technology that allows work done earlier to be reused. This is called caching.

Imagine a warehouse where employees have to fetch the same goods and documents every time a new order is processed. If everything that is used often is already waiting at the packing table, the job goes much faster.

Blacksmith uses a similar principle. Data that the tests need is stored close to the computers that carry out the checks. That way the system does not have to fetch or build everything from scratch every time.

The founders also found that powerful processors made for gaming were especially well suited to this kind of task. In the beginning, they actually rented servers from a provider that usually supplied capacity for games such as Minecraft and Call of Duty.

One example the company highlights is the customer LangChain. According to Blacksmith, LangChain's developers used to wait between 15 and 20 minutes for free capacity on some test jobs. With Blacksmith, this wait was reduced to about 10–15 seconds. That is the wait before testing begins, not necessarily the time it takes to complete the whole test.

Why has this become a billion-dollar company?

Blacksmith does not primarily sell programming or finished tests. It sells capacity, speed and infrastructure that let companies develop and update software faster.

Its customers include the technology companies Supabase, LangChain and Mercury. In October 2026, Blacksmith stated that its infrastructure handled more than 90 million automated test and build jobs per week.

For a small project, the free test capacity through GitHub can be more than enough. But for companies with many developers, large software systems and thousands of daily changes, waiting time can become expensive. Even small improvements can add up to large savings when they are repeated millions of times.

This development also shows how AI is changing the needs of the software industry. As it becomes easier to produce code, the need to check it grows. Blacksmith has built its business around exactly this shift.

The company is also working on more automated quality assurance, where screenshots and video, among other things, are meant to help document that changes work as expected.

AI can write more and more of the software itself, but that does not mean the result is automatically correct. The more code that is produced, the more important the systems that check it become. This is the part of software development that Blacksmith wants to make faster.

Key takeaways

When AI writes more code, more code has to be checked, and testing can become a bottleneck. Blacksmith lets the same automated tests run on its own infrastructure with powerful processors and caching, so the wait gets shorter. The more code AI writes, the more important the systems that check it become.

Sources and further reading