Today I’m pleased to announce the release of The Personal AI Architecture (PAA), an open blueprint for building user-owned AI systems.
PAA separates the things that make an AI system personalized to you - your conversations, preferences, files, skills etc - from the application.
I refer to these things as Your Memory. And with PAA, Your Memory is the platform.
All other components of the system are decoupled and swappable. This ensures zero lock-in, even to the system itself.
Systems built on the PAA are infinitely extensible via tools, and can use any available interface and model.
This means we do not have to build a new ecosystem for PAA. We can absorb all the great work in the existing open source ecosystem.
With just 4 components and 2 APIs PAA is simple enough for individuals to build on, and small enough to run on a consumer laptop.
In addition to an in depth overview of the Architecture, we are also releasing a typescript template, and the first fully functioning implementation of the architecture we call BrainDrive.
PAA, the template and BrainDrive are all open source and MIT Licensed. Github in the first comment.
Any and all feedback is appreciated.
In this free course I show you how to vibe code your own personal AI system from start to finish in less than 30 minutes. Runs privately on your computer and customizable in any way you would like. Enjoy!
The goal of Personal AI: civilization where individual humans, augmented by AI, can do consequential work without being captured by extractive institutions.
Freedom to write your prompt and own your data.
This is the new battleground.
2034 won’t have to be like 1984.
Startups have a huge advantage here. We can build for the new paradigm instead of trying to retrofit an org build for the old paradigm. Keep your data unified from the beginning in your own AI system instead of spreading it out all over the place and trying to figure out how to connect it all.
Whether it’s existing consulting firms, new ones that emerge, FDEs from agent vendors, or new internal agent engineering roles, the amount of work that is going to be created to implement agents in enterprises will exceed anything we imagine today.
The complexity of implementing agents in any existing organizations is very real. When I talk to large enterprises, as you move from a chat paradigm to agents that participate in meaningful workflows, there are a number of things they need to do.
First, you have to get agents to be able to talk to your data securely across your systems. In many cases, enterprises have decades of legacy infrastructure that contain the valuable context for AI agents. That’s going to take a ton of work to go modernize and move to systems that work well with agents.
Then, you need to ensure that you’ve implemented agents with the right access controls and entitlements, the right scopes to be safely used, and have ways of monitoring, logging, and securing the work that they do.
Next, you need to actually document the processes in the organization in a way that agents can utilize for doing the work. You also need to figure out what the new workflow looks like when agents and people are working together on a process, and who steps in where. Just replicating the old workflow will mute the gains. Oh and you likely need to create evals for your top new end-state processes.
Finally, you have to keep up with a rapidly changing set of best practices and architectural shifts happening in the agent space. While it’s fun for people to change their personal productivity tools on a dime, it’s 100X harder to do this in a business process. The speed of change is a blessing and a curse right now for anyone trying to keep a stable system design.
All of this means that individuals and companies that develop expertise on the above set of components (and more) are going to be needed to help organizations actually implement agents at scale. This is also the rationale for vertical AI agents right now that can go in deep on a business domain and help bring automation to it.
This is a huge opportunity right now whether you’re doing this internally or as an external business provider.
How much of the "our data is spread out all over the place" is a legacy problem? I am spending more and more of my time in one interface just talking to the AI feeding it everything like transcripts etc there, and having it do everything which makes the unifying data issue much easier to solve. Def need something to bridge the gap in the meantime though.
@yoheinakajima this is the way and another reason why it’s going to be personal ai systems coming together to make the larger systems. It makes the constant optimization and refining at scale possible.
I agree with everything but the don't give me chat part. Just talking to it is the ultimate consumer experience because everyone already knows how to do that. Chat is the way. The problem right now is that because this wasn't possible until recently there is so much legacy in the way of it.
@ycombinator@t_blom The brain is the right analogy and also shows why this is the wrong path. The company brain will not be a separate thing. It will be made up of a bunch of individual brains, just like it is now. This is also how the permissioning problem will be solved.
I think another piece that is missing is that each individual is going to have their own brain that they are bringing to the company and how these individual brains interact with the company brain. I'm betting that the company brain will be built from these individual brains and not the other way around.
I think another piece that is missing is that each individual is going to have their own brain that they are bringing to the company and how these individual brains interact with the company brain. I'm betting that the company brain will be built from these individual brains and not the other way around.
Hi Doug I have been working on this same issue and it's not just the models but all the tech is changing so quickly. I think this problem requires a new architecture and that AI enables that architecture. I'm working on this for personal AI specifically but I think it applies to all software. If you have a moment to take a look I'd appreciate any feedback thanks: https://t.co/AMcCYDMPHf. .
Today I’m pleased to announce the release of The Personal AI Architecture (PAA), an open blueprint for building user-owned AI systems.
PAA separates the things that make an AI system personalized to you - your conversations, preferences, files, skills etc - from the application.
I refer to these things as Your Memory. And with PAA, Your Memory is the platform.
All other components of the system are decoupled and swappable. This ensures zero lock-in, even to the system itself.
Systems built on the PAA are infinitely extensible via tools, and can use any available interface and model.
This means we do not have to build a new ecosystem for PAA. We can absorb all the great work in the existing open source ecosystem.
With just 4 components and 2 APIs PAA is simple enough for individuals to build on, and small enough to run on a consumer laptop.
In addition to an in depth overview of the Architecture, we are also releasing a typescript template, and the first fully functioning implementation of the architecture we call BrainDrive.
PAA, the template and BrainDrive are all open source and MIT Licensed. Github in the first comment.
Any and all feedback is appreciated.
That's everything you need to know to go from 0 to AI hero. If you want to dive deeper, check out the full free course in the BrainDrive community. I hope to see you there!
As AI agents begin to act on our behalf ownership becomes even more important. You need to know that your AI system is acting in your best interest. And you need to be the one who captures the value it creates, not a platform that rents you access.