Thanks, man 🤟 You can, but you have to write a different data collection class (or just manually copy your notes into JSON files that follow the same structure)
Great technical deep dive on RAG architectures, Paul. This kind of knowledge is what separates truly useful AI assistants from glorified chatbots.
I've been thinking a lot about the build-vs-buy decision for personal AI assistants. Clawdbot (44,000+ GitHub stars) represents the "use someone else's work" approach, while your course represents "understand the fundamentals."
I researched Clawdbot extensively - the coverage is mostly hype. Reality: $300+ burned in 2 days by some users. 180M tokens in Viticci's first week. 512 security findings.
I ended up building my own system (Wiz) using Claude Code. The memory architecture you describe is exactly what makes these systems actually useful.
This is very interesting, and while I didn't get a lot of it, you may have unblocked my own LLM building journey by introducing Gradio, so thanks for that!
As a zero tech, How do I build one. I have some many ideas, questions around the Idea of Second Brain / AI Me. Anyone interested Dm me for any thoughts!
Pretty cool and nerdy but beyond the amusement and masturbation of chunking machines together, what value do you genuinely find in this? I know "learning" and what not can be a common refrain, but that seems more like a cope than anything. Learning requires active engagement and changes to neurons, not chatting with a bot.
I'll look this over. Would be interesting to develop this using my Obsidian notes and other sources rather than Notion.
As long you format your Obsidian notes in Markdown + extract the metadata in the same JSON format, you can hook any source you like
Good to know.
100% Chris - if you are open, you can contribute with an article on how to adapt the course to Obsidian or other data sources
Wow awesome Paul. Definitely follow the series. One more question: can i change notion to apple note?
Thanks, man 🤟 You can, but you have to write a different data collection class (or just manually copy your notes into JSON files that follow the same structure)
Excellent. Thanks.
Thanks 🤟
🖖
Great technical deep dive on RAG architectures, Paul. This kind of knowledge is what separates truly useful AI assistants from glorified chatbots.
I've been thinking a lot about the build-vs-buy decision for personal AI assistants. Clawdbot (44,000+ GitHub stars) represents the "use someone else's work" approach, while your course represents "understand the fundamentals."
I researched Clawdbot extensively - the coverage is mostly hype. Reality: $300+ burned in 2 days by some users. 180M tokens in Viticci's first week. 512 security findings.
I ended up building my own system (Wiz) using Claude Code. The memory architecture you describe is exactly what makes these systems actually useful.
Clawdbot vs building your own: https://thoughts.jock.pl/p/clawdbot-deep-dive-personal-ai-assistant-2026
Oh dang, nice! I have to take a look
This is very interesting, and while I didn't get a lot of it, you may have unblocked my own LLM building journey by introducing Gradio, so thanks for that!
Haha, it's my pleasure. Follow the next lessons to better understand the details and zoom in on each component with theory and code
it is awesome.
Thanks 🤘 This Thursday we release lesson 2
It is not a “brain”, but maybe pseudo-brain.
The Second Brain 🧠😂
great job Paul, excited for the coming lessons!
Thanks Elie 🤘 This Thursday we release lesson 2
Thanks
Great content. Really appreciate
Great article! 🙌
Loved this. 🔥
So I shared it in my community: https://app.daily.dev/posts/build-your-second-brain-ai-assistant-jtlgaltlr
Lovely article. Did you use AI to generate the voice ?
I think companies are not ready go implement system such as does. Centralized information is the key factor for business performance
As a zero tech, How do I build one. I have some many ideas, questions around the Idea of Second Brain / AI Me. Anyone interested Dm me for any thoughts!
Pretty cool and nerdy but beyond the amusement and masturbation of chunking machines together, what value do you genuinely find in this? I know "learning" and what not can be a common refrain, but that seems more like a cope than anything. Learning requires active engagement and changes to neurons, not chatting with a bot.