Raw brain dump
Brain Dump #1:
Okay, I want to start working on this idea, but obviously yeah, for the first reps we're just gonna kind of Go through this manually which I want to do because I think it'd be way more authentic and raw and real Uh, and i'm just gonna whisper flow brain dump The title I fucking hate that's not kind of the vibe I was going for I of I rebuilt my business os So I could actually help. It's more like I think the learnings also that I want to take is like the evolution The evolution of your operating like it's never You're never just gonna have one it's never gonna be one tool one skill although all the videos you watch make it Promise that like oh you just buy one mac mini and your problems are solved or you just get these Five skills and then all your problems are solved It's more like you build something You test it you gather data New conditions change and then you adapt. I think that's what you have to do um Like I've got a lot of that from kind of reading and ingesting a lot of how Elon Musk thinks and like first principles and stuff like that And being able to adapt to the conditions With confidence as well, I think And also just speaking to people who are smarter than you so say for me it was kind of the evolution was going from a solo operator to working with a team and Obviously, I found out quickly. I remember when I was working in a startup like my first like actual knowledge work job like people's brains don't think like me like everyone thinks differently and You have to learn to eat like kind of explain communicate and just even just day-to-day work Like creators work differently than say Analytical people all that kind of stuff Similar to the book Surrounded by idiots where he kind of colors people in like red blue yellow green And the whole concept of the title is if you're red and you speak to a blue be speaks of them like a red You're gonna sound like an idiot to them and vice versa because you're not communicating in the language that they their brains think in So, you know again adapt is similar to the business OS stuff and AI stuff it's like I had my original notion set up I Haven't worked. I haven't run my own team before so it's like okay now. Let me go speak to people who have And I like five to ten levels ahead of me See how they run it and then go see specifically how people on notion run it with a team in the best way possible Not only for like humans, but also with AI like building for the future And building that kind of foundational layer where agents and humans can work in harmony And also data lives and no context is leaking That's like one of my biggest fears and like going to bed is having context that's leaking and not captioned So how can we capture that with very little friction? So, I think I learned a lot I was on a call with these guys named Ethan and Zack they have their business in Melbourne very energetic Zack Messaged me on Instagram. And as soon as I saw the keyword notion He's like my co-founders a nerd at this you guys should hop on what was meant to be like a 20 minute call turned Into 90 minutes of just nerding out sharing screens Giving tips back and forward and both of us are Just learning and coming up with new ideas and new ways to adapt to the new conditions.
Brain dump #2:
The real leverage is that everyone is so focused on more AI tools and new models, but that only gets you so far. Right now, your time is more usefully spent building context infrastructure that AI and humans can operate on, and making sure all the data feeds into that infrastructure.
I also really like the clean mental model of having a frontend for humans to look at and a backend for the AI to look at. I was slowly building that out, but that framing really changed my perspective on how to build it, like you do with an app. You have the frontend for humans and the backend for the AI.
A big point I keep making that I have not really heard anyone else say is this: we built the world for humans to live in, and that makes sense. We are priority number one. Now AI agents have come along, so we have to rebuild the digital world for agents and humans to work in. In the future, we will have to rebuild the physical world as well, for robots, humanoids, and humans to live within.
You can think of things like email, text messages, and UI buttons as all being made for humans and being very efficient. AI does not need to do that. We can delete and simplify the process of them querying a database directly via SQL or code, and MCPs and CLIs are making that very efficient. Notion dropping their new CLI and it being basically 91% more efficient than the MCP is really good, and it ties into that.
A lot of people are obsessing over Obsidian, and it is really good for a solo entrepreneur getting started and getting into memory and knowledge. It is really cool. There is nothing wrong with that, but if you own a real business and you have a real team and you want to scale and genuinely build an operating system that you can work from your laptop, your Mac Mini, your phone anywhere in the world and have it constantly running, I think Notion’s new developer launch of 3.5 really kind of instills them as the best spot to do that.
The big one that I learned from speaking with these two guys is relations, not backlinks. Obsidian people are obsessed with just backlinking content and then making their pretty knowledge graph, which actually provides no value. It is just visual stimulation that looks cool. Notion database relations create a proper operating layout for both humans and agents. That is what we are kind of really installing because then it lets Claude code our structured questions throughout the business, no matter what you are going up or down.
Whether you are going from the task related to the projects related to the clients related to the deliverables, these two-way relations are really strong across a range of databases. On top of that, you have AI autofill inside your property. Notion has this feature where you can make an AI basically fill out the properties for you, whether it is tagging, categorizing, linking, filling out just text or summaries, and all that kind of stuff. This is really good for both Notion AI or Claude code or codex to search by the new CLI in a database. It will search just the properties and the title in a much more efficient way instead of opening the page, reading the whole page, ingesting all that context, making a decision, and then proceeding to the next one. Now you can work with speed much better across a wider range of data.
I think that’s a really good example of the world-building concept Caleb Ralston talks about: packaging and having clear concepts and definitions.
I think this AI autofill plus relations inside Notion databases equals speed rails for your AI tool of choice (Claude Code, Codex, Notion AI, whatever it might be). It 10x your workflow, and it compounds over time in a much better way than a basic Obsidian setup could.
Now it’s about turning the outcomes into action. I talked to these amazing people, and I learned that ingesting information for its own sake doesn’t actually change anything. It’s dumb to read a book or watch a YouTube video and then not change your behavior.
Same behavior = dumb. New information + changed behavior = smart. Look at these guys: this is working for them, so I’m going to try it for me.
That’s the evolution of my work:
There are so many variables and conditions set now, and it’s made the game more challenging. It’s like going to the next level in a video game. The team has helped me force me to do it, because it’s different when it’s your own team versus working with other teams or just speaking to other teams.
The outcome I’ve come to is a tiered setup:
We’re slowly going to integrate all the data sources into one spot:
On the front end, we can build a filtered database dashboard view so humans can see exactly what they need. The AI can query everything in one spot instead of pulling from 10, 20 MCPs or CLIs and being very inefficient. We do this through skills and SOPs to tell it exactly what to search and when, and in what conditions. The AI is also smart enough to search that by itself.
I also really like the concept of studio hubs. They are a separate condition that wasn’t applicable to them but probably is to me: I’m working with an agency as well, and it’s helping me out with a lot of my marketing stuff.
That’s different from my team members, and it requires different access. They don’t have full access to all the data. I just want them to see a certain view. That’s where these studio ideas come in: a growth studio, a client studio, and a content studio. Each one is for a specific group of tasks, workflows, and information, and it’s all linked where it needs to be.
On top of that, we want to tie it back to the point that everyone thinks differently and works differently. I have my private page, which is my command hub. It is almost like I'm sitting at a control center and I can see everything I need to. Everything is filtered to show only what I need to see.
Other people on my team work differently as well. Now they can create their own personal command center, and it’s filtered the way they want to work.
But again everything's built on top of that backend now so no context is leaking at all.
That has made me conclude there are two versions of Oliver now:
I think it is good to have that split because the human, front-end part is super important. Clarity helps me dial in and filter out what I need to see. If you look at every piece of information, your brain will be overloaded and you will not actually take any action. By making it super clear and documenting what I need to see based on these variables and conditions, show me exactly what I need to.