You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
I started Corridor while while working on a separate project. I wanted to understand why certain internal mechanisms were not behaving as designed. I built an instrumentation system to see what was happening inside of the program. That instrumentation is what became Corridor. Corridor was designed to observe the actions and telemetry of an AI system to see what failure patterns were occurring over time. Corridor has two layers. The first can observe agent actions, tool calls, results, blocking decisions and task completion. It also has a deeper layer with 3 detectors, FS1, FS2 and Invariant Persistence. These were designed to identify structural problems that may develop across many cycles rather than just one. These detectors observe unresolved prediction tension (FS1), falsification debt(FS2) and persistent violations of structural invariants (Invariant Persistence).
I’m trying to determine if Corridors deeper layers can work on another system and not just in a controlled environment where I’m certain the telemetry already exists. I need to first find what telemetry Corridors needs such as unresolved prediction tension, how much influence competing predictive models have, structural invariants and enough provenance to understand how those things change from one point to another.
A big part of this project will be observability by design, which as AI systems become more advanced and ungoverned, will be the future of AI agentic design. I want to work with researchers and engineers that are at the beginning stages of a system and can design the telemetry to be exposed from the beginning rather than using a mature system where obtaining the telemetry is substantially harder. I am searching now for a partner who is in these stages for collaboration, if I don’t find this partner, I will build a system to see if the telemetry can be exposed in order to get Corridors detectors to fire, which I have already done in a controlled environment with synthetic test suites with a positive outcome.
The funding would mainly be used to give me more time and resources to turn Corridor into a more serious external research and validation effort. Up to this point I have funded Corridor completely on my own, around my regular work and obligations. I would be able to devote a larger portion of my time to the project.
Another reason for the need is I would bring in specialized help where needed in targeted areas such as software engineering, AI research and independent technical review. Some of the funding would go towards technical expenses such as computer, API and model access, cloud services and data storage.
The point of funding is not to add more features to Corridor but to find out if I can get Corridors deep detectors to work on independently designed systems.
As of now, I am the only one on the team. Corridor is a founder led project derived from the necessity for better, deeper observation tools for my own AI agentic projects. I do not come from a traditional academic computer science or engineering background. I use AI assisted engineering tools along with GitHub, Replit, Codex with Python based testing. My track record is mainly Corridor itself, although I have built several agents using the Open AI SDK frameworks, they all were out on the back burner once I realized the possibilities what Corridor could become.
I have a published paper on SSRN the part of the theory behind Corridor and its deep detectors and the idea of false structural relief in agentic systems
I think the biggest way this project could fail is if the deeper diagnostic detectors do not work anywhere but in the controlled environment where I know the telemetry exists. However, as AI systems become more capable and consequential, developers are likely going to need evidence of what those systems are doing internally. That may create pressure to expose more structured telemetry and Corridor is trying to find what that telemetry will look like.
I have not raised any outside money for Corridor. I have invested only my personal time and money to date.