From Karpathy's LLM Wiki to Brian Halligan's digital twin, how the smartest people in tech are building AI systems that think, remember, and act on their behalf.
Much easier if you run your own company. You *can* mix the two though that is tricky over time, especially if you want to sell said company.
If you have to decide what to shar eeach time, I am not convinced it will work. It is better to be less strict about it - but then you run into confidentiality issues and what can be shared more widely than not. We want to be able to share everything and then have guardrails that filter what is allowed and not allowed between brains. And then allow on an exception basis.
One of the cleanest overviews of this concept I've seen in a while, and I especially appreciate the examples. I've talked to a lot of people about building second brains for their solo businesses, and everyone seems to overcomplicate it. Love the diversity and the examples that you've provided.
Nice breakdown Kieran! I have the same setup you described with separate work and personal brains that don't talk to each other. The permission-layer idea makes so much sense because right now I'm basically context-switching between two completely separate accounts. The Halligan example is wild...lol. An AI agent attending your Zoom meetings is a whole different level 😅
I don’t think this is merely a trend. I think we will get to a point where we can query and chat about anything we need in the company at any moment (data, status update, question). Spin up a deck from it mixed with what we add, things like that. The individualized, stateless agents don’t get smarter over time and they break down the second a document is collaborative and has more than one hop. I think this is how we get value that lasts in a business. Good point on the permissions…that can get hairy. But the knowledge graph and drift detection/self correcting loops solves the rest.
Much easier if you run your own company. You *can* mix the two though that is tricky over time, especially if you want to sell said company.
If you have to decide what to shar eeach time, I am not convinced it will work. It is better to be less strict about it - but then you run into confidentiality issues and what can be shared more widely than not. We want to be able to share everything and then have guardrails that filter what is allowed and not allowed between brains. And then allow on an exception basis.
One of the cleanest overviews of this concept I've seen in a while, and I especially appreciate the examples. I've talked to a lot of people about building second brains for their solo businesses, and everyone seems to overcomplicate it. Love the diversity and the examples that you've provided.
Nice breakdown Kieran! I have the same setup you described with separate work and personal brains that don't talk to each other. The permission-layer idea makes so much sense because right now I'm basically context-switching between two completely separate accounts. The Halligan example is wild...lol. An AI agent attending your Zoom meetings is a whole different level 😅
Loved this, thanks for sharing Kieran..
I am also building one for the marketing companies with whom I am working with..it's fun....will share the updates...
I don’t think this is merely a trend. I think we will get to a point where we can query and chat about anything we need in the company at any moment (data, status update, question). Spin up a deck from it mixed with what we add, things like that. The individualized, stateless agents don’t get smarter over time and they break down the second a document is collaborative and has more than one hop. I think this is how we get value that lasts in a business. Good point on the permissions…that can get hairy. But the knowledge graph and drift detection/self correcting loops solves the rest.