You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
AI is being given more memory, more access and more ability to act on its own. I believe there is a simple question that has to be answered before any of that happens. Does it still have the authority to do what it is about to do?
That is why I built Trace Continuity.
Trace checks authority at the time an AI tries to act, not just when permission was originally given. It governs what the AI can store and retrieve, protects sensitive information before it is stored, and creates an audit record showing what was allowed or denied and why.
Trace also checks something else that I think is just as important. Even when authority is still valid, conditions can change. Consent can be withdrawn. Risk can change. Legal or compliance conditions can change. Trace checks those conditions again before the action happens and can refuse the action even when the authority itself is still good.
In plain English, Trace asks two questions. Are you still allowed to do this? And even if you are, should this particular action still happen right now?
Trace is already built and live. The current test suite is 86 out of 86 passing, including 22 new Action Admissibility and adversarial tests. The gate bypass scan and typecheck are clean. There is also a public Playground and Break Arena at tracecontinuity.com where people can see the system and try to challenge it for themselves.
What I want to do now is get Trace out of my hands and into the hands of independent technical people who can test it harder than I ever could. I want them to try to break it, find what I missed and tell me where it is strong and where it needs work.
Then I want to take what we learn and put Trace into controlled public interest pilots. Legal aid is one area I believe makes a lot of sense because those organizations are beginning to use AI while handling information that can be extremely sensitive. Trace already has protections for matter boundaries, ethical walls, legal holds, changing authority and audit evidence.
If this project is funded, I am willing to provide Trace at no licensing cost to legal aid or other qualifying public interest organizations participating in these pilots. I do not want this funding to simply help me sell a product. I want it to help answer whether the system I spent the last two and a half years building can actually make AI safer for organizations and people who may not have the money to buy this kind of protection themselves.
If something breaks, we fix it. If something works, we prove it. And what we can responsibly share from that testing should help more people than just me.
My main goal is to take Trace from a system that has performed extremely well in the testing we have done and put it to work in places where it could actually help people.
Trace is built and live. The current test suite is 86 out of 86 passing, including 22 Action Admissibility and adversarial tests. The gate bypass scan and typecheck are clean. We also have a public Playground and Break Arena where people can see and challenge the system themselves.
The next step is real world validation.
One area I believe could benefit from Trace is legal aid. These organizations work with people who may not have anywhere else to turn, while handling information that can be extremely sensitive. At the same time, AI is becoming more useful to organizations like these. The problem is that giving AI more access and memory also creates new risks around who or what can access information and whether that access should still be allowed when circumstances change.
Trace was built around those exact problems.
My goal is to provide Trace at no licensing cost to qualifying legal aid organizations participating in this project and give them access to governance technology they might otherwise never be able to afford. I would like to begin with controlled pilots using synthetic or de identified information and measure how Trace handles changing authority, matter boundaries, ethical walls, legal holds, sensitive information and changing conditions at the moment an AI tries to act.
Depending on the amount of funding, I would also like to include universities, AI safety researchers and other public interest organizations that could benefit from testing Trace or help evaluate what it does.
I want this project to produce real evidence. I want to know how Trace performs outside the environment where it was built, what happens when independent people challenge it, and whether execution time authority and Action Admissibility can provide meaningful protection in real AI systems.
At the end of the project I want to be able to show what was tested, what Trace stopped, what it allowed, what the audit evidence showed and what we learned from putting it into real public interest environments.
I built Trace because I saw a problem with AI memory and authority. The goal now is to find out just how much good we can do with it.
I want to achieve these goals by putting Trace into the hands of organizations that are trying to help people and may not have the money or resources to buy this kind of technology on their own.
Legal aid is where I would like to start. These organizations help people who may be facing eviction, family problems, benefits issues, domestic violence and other situations where having legal help can make a huge difference. They also deal with information that needs to stay private and protected.
As AI becomes more useful to organizations like these, I do not think the organizations with the biggest budgets should be the only ones that get access to strong AI governance.
If this project is funded, I am willing to provide Trace at no licensing cost to qualifying legal aid organizations participating in the project. The funding would help cover the cost of putting Trace into those environments, supporting the integrations, running the infrastructure and giving these organizations the help they need to actually use it.
We would start carefully with controlled uses and synthetic or de identified information where appropriate. From there we can work with each organization to find places where AI can help while Trace protects the memory, controls who or what has authority, keeps matters separated and checks whether an action should still be allowed when conditions change.
If the funding allows it, I would like to expand beyond legal aid and work with other nonprofits, universities or public interest organizations that could benefit from the same protection.
The idea behind this project is pretty simple. I have already spent two and a half years building Trace. Now I would like to use what I built to help organizations that are helping other people.
If we can make AI safer for a legal aid attorney helping somebody keep their home, a nonprofit helping a family or another organization doing work that matters, then this project will have done something worth funding.
The funding will be used to give legal aid and other qualifying public interest organizations access to Trace at no licensing cost and to cover what it takes to actually put it to work for them.
That means helping with the technical integration, cloud and infrastructure costs, setting up controlled pilots and providing the support needed so an organization with a limited technology budget is not handed a platform and then left to figure everything out on its own.
My first priority would be legal aid. I would like the funding to support several organizations if possible, starting carefully with controlled uses and synthetic or de identified information where appropriate. The goal is to let them explore ways AI can help the people they serve while Trace provides governance around memory, authority, sensitive information and changing conditions.
As funding allows, I would like to make the same opportunity available to other nonprofits, universities and public interest organizations where governed AI could make a difference.
Some of the funding would also be used to measure and document what we learn so other organizations can benefit from the work even if they were not part of the original pilots.
I have funded the building of Trace myself for the last two and a half years. I am not asking for funding to build the product.
I am asking for help putting what I already built to work for organizations that may never have the budget to access technology like this otherwise.
Right now I am the team.
I grew up poor and quit school in the ninth grade to go to work as a carpenter. By the time I was 20 I was a journeyman, and by 25 I was a general foreman working on major jobs including aircraft carriers, high rises, refineries and paper mills.
I spent about 35 years in construction. I learned how to build things, solve problems and coordinate people with completely different skills to get one finished job. I never imagined those lessons would eventually help me build an AI governance system, but that is pretty much what happened.
About two and a half years ago I started trying to build the most human like AI bot I could. I wanted to build something that could really get to know somebody and maybe help people dealing with things like addiction, weight loss or depression. I ran into the memory problem pretty fast. An AI cannot really get to know somebody if it cannot remember them.
That started me down a road I never expected to be on.
I had no technical team and no technical education, so I started using five or six different AI systems much like I had used different contractors throughout my construction career. I gave them different jobs, bounced ideas between them, had them challenge each other and kept working through the problems one at a time.
There was a lot of broken code, a lot of starting over and plenty of times I had no idea how I was going to get past the next problem. I just kept going.
Thousands of hours later, that original memory project became Trace Continuity.
Trace is my track record in technology.
It is built and live today. The current test suite is 86 out of 86 passing, including 22 Action Admissibility and adversarial tests. It has a clean gate bypass scan and clean typecheck. The website also has a public Playground and Break Arena where people can see what I built and challenge it themselves.
Trace has also started getting attention from people outside of me. People with backgrounds in technology, business and enterprise systems have looked at the architecture, gone through the website, used the Playground and challenged the Break Arena. The response I have received has given me even more reason to believe this deserves the chance to do some good in the real world.
I am not going to pretend to be something I am not. I do not have a PhD, a technical degree or a Silicon Valley resume. I am a carpenter who spent most of his life building things with his hands and somehow ended up spending the last two and a half years building this.
I know how strange that sounds. Sometimes it still sounds strange to me.
But Trace exists. People can go see it, test it and challenge it for themselves.
I have spent most of my life building things for somebody else. If what I built this time can help legal aid organizations, nonprofits and other people doing work that matters use AI more safely, I would like to give it that chance.
Failure is not an option.
I know that probably sounds like a simple answer, but it is how I have approached Trace for the last two and a half years.
The biggest risk is that once Trace is being used in different real world environments, we learn that some organizations need things that Trace does not currently handle the way they need them handled. Every organization works differently, and there is no way for me to know every situation we may run into before we get there.
I do not consider learning that a failure. That is part of putting something to work in the real world. We learn from it, make Trace better and keep moving.
Another risk is that some of the public interest organizations we want to help may need more technical support or integration work than expected. That is one of the reasons funding matters. I do not want to offer an organization Trace at no licensing cost and then leave them responsible for figuring out how to put it to work.
The worst outcome would be going through this project and finding that Trace does not provide enough real benefit to justify using it. If the evidence showed that, I would have to accept what the evidence says.
But everything I have seen so far gives me reason to believe the opposite. Trace is built, it is live, the current test suite is 86 out of 86 passing, and people outside of me have already spent time looking at and challenging what I built.
I have spent my life building things. Sometimes you find something you did not expect once the job gets into the real world. You deal with it and keep building.
That is exactly how I will approach this project.
Failure is not an option.
I have raised $0 in outside funding.
Every dollar that has gone into Trace has come out of my own pocket. I have bootstrapped this project from the beginning and paid for the development, hosting, infrastructure and everything else it has needed myself.
So when you ask how much money I have raised, the answer is zero.
The only person I have raised money from is me.
If you are a potential funder reading this and you have made it this far, I would ask you to do one more thing. Go to tracecontinuity.com and look at what I built.
Read the Architecture Brief. Open the Playground and see how Trace handles governance decisions. Go into the Break Arena and try to get around the governance yourself.
You do not have to take my word for what Trace is or what it can do. I built the website that way on purpose.
Go look at it. Try it. Challenge it.
Then decide whether you think what I spent the last two and a half years building deserves a chance to help people.
There are no bids on this project.