A feature-based tech tree for progressing towards real Digital Assistants

December 14, 2025

What are we actually trying to build with Personal AI?

Where is it heading?

How far along are we?

What parts of that ecosystem are ChatGPT and Claude Code?

What are the next steps on that path?

—

These are the questions I've been asking myself as I've been building out the Personal AI Infrastructure (PAI) project on Github.

The project's purpose is to provide a customizable AI Infrastructure and Digital Assistant system that can help anyone, regardless of background or means, with what matters to them as humans.

PAI has a ton of functionality already, but questions like the above help me think about what's still missing and therefore what we should be building towards.

This means things like:

The idea is to have handles for talking about and building this stuff. So what we did here was break these into a list of granular features that we can actually pursue in GitHub, and then group those roughly into a set of phases or tiers to use for a maturity model.

First, the granular ones.

1. : You ask a thing in text, and it can answer back.
2. : The system knows a decent amount about you, so it can customize its results based on its knowledge.
3. : This can always get better but became pretty solid in 2024.
4. Zero Friction Access: You can use your assistant when you're away from your primary interface, like your home computer, its terminal, or even a particular website.
5. Continuous Activities: Our assistant can continue working for arbitrary amounts of time while we're doing other things.
6. Persistent Voice: Any time, anywhere, you can just whisper or speak to have your assistant start working for you.
7. Persistent Sight: Your assistant can see what you see, and also behind you (as well as around you, if it has access to other cameras).
8. Full Computer Use: You can navigate with voice and gesture while your assistant does the work using your system.

I would say we are roughly at #4, which is precisely the feature that I'm working on next for the PAI project.

For the Phases / Tiers of the Maturity Model, I think these are a decent start.

1. : Basic text interaction—you ask, it answers.
2. : The system knows who you are and customizes responses accordingly.
3. : API-enabled. Can take actions in the world—search, code, browse, create.
4. Presence: Is always with you when you need it, as opposed to requiring a special system or interface.
5. Proactive: Anticipates needs and acts without being asked.
6. Senses: Always able to listen and see what's going on around you—persistent voice and vision.
7. Advocates: Negotiates, represents, and acts on your behalf in the world.