top of page


Probot is an abbreviation for 'programmer robot'.


We envisage a future where human programmers do what they do best while AI programmers assist with controlling the complexity of large scale dependency structure that often limit people. We're focusing on large-scale code structure here, not local linting issues that have a plethora of existing codegen solutions. See for an example of our publicly-facing work in this space.

Without tools in place that can examine dependencies for cycles or detect other breaches of architectural rules (like no test / experimental / known-insecure code in prod), human programmers typically reach a scaling limit whereby unintended side-effects propagate throughout the causality graph of their codebase.

Understanding and controlling dependencies could have easily prevented hundreds of millions of dollars of losses in the Wormhole hack and Knight Capital collapse. Ask us how! Implementing our top suggestions should dramatically reduce the odds of your company being next; we can show you the minimal code moves required to keep your code complexity below your desired risk tolerance (such as at most 2 classes in a cycle, or 1000 lines in a cycle). If you'd have an idea of your desired complexity limit, ensure that you have confidence in your developers being able to fully understand and test isolated code chunks under this limit in one day; most people can't truly understand 1000 lines of code correctly in one sitting and most companies have much larger cycles than this, which is often why everything is breaking all the time. You could also quickly check right now to see how many namespaces / classes / functions are fundamentally required to be in a cycle in your codebase due to recursion.

Automated topological analysis of what your codebase actually looks like via Dependency Structure Matrices is a first step towards managing codebase complexity. This is how you discover the difference between your intended Software Architecture and your actual Software Architecture. This is also how you facilitate incremental testing and understanding of code, allowing newcomers to make substantial contributions quickly. This also helps advanced build systems (Starlark variants) build and test your software rapidly and incrementally, with confidence.

We have internal tools that take Dependency Structure Matrices a step further, highlighting top architectural issues, together with reasoning and steps to fix, providing a truly augmented approach to developing complex codebases that scale effectively.

Our History

Probot LLC has experience analyzing the code of top financial, tech and Web3 companies. We take inspiration from the scaling approaches that Google introduced on the tech side and Goldman Sachs introduced in the financial space.

We know how to make large fragile codebases understandable and testable as quickly as possible. Contact us to find out how!

bottom of page