So this week I set up a second brain inside Obsidian using Claude Code, and honestly, it's so good it's not even funny.

I can now ask Claude about my own YouTube channel, something like "what are the five recurring arguments across all my videos?", and it just comes back with a real answer. And it cites the videos. Actual citations, which is so valuable.

And the best bit? You literally give it one prompt to set up the infrastructure, then one prompt to ingest your data, and boom, you're straight away talking to your own data and getting real answers. That's it. I'm giving the prompt away for free, it's in the description.

Watch the full build here → https://youtu.be/zjrSoXtlunU

Now, the first thing you're probably thinking is "isn't this just RAG?" And it's not, actually. This is based on the LLM Wiki concept from Andrej Karpathy (one of the best guys in AI, now over at Anthropic). Basically, instead of just storing chunks in a database, Claude reads your stuff and builds an ever-evolving encyclopedia out of it, all linked together. So when you ask a question inside that folder, it reads your own notes first and answers from them, with citations. It's almost like magic, honestly. Works every time.

Now, speaking of giving things a memory, that's exactly what today's sponsor does, except for your AI.

So wouldn't it be great if, when you're building AI apps, you could just give your AI a memory automatically, instead of it forgetting everything the second you close the session? Well, that's basically what mem0 does. Instead of dumping every single message into a database, it automatically pulls out the facts that actually matter, then next session it only feeds back what's relevant to what you're asking right now. So your context window stays lean, your API costs stay low, and your app actually remembers people.

And it's literally just a pip install and a couple of lines of code. pip install mem0ai, drop it into OpenAI or LangChain, done. Over 90,000 devs are building on it, it's YC-backed, SOC 2 compliant, and there's a free tier to get started.

So if you want to give your AI an actual memory, go check it out → https://mem0.ai/?via=luke. Huge thanks to mem0 for sponsoring this one.

Right, last thing. If you want to keep your second brain up to date, just tell Claude something like "create a routine that updates these transcripts every Monday at 9am", run it locally, click allow, and that's you. It stays current on its own.

So go give it a watch and steal the prompt. And hit reply and let me know what you'd point yours at, whether that's your SaaS, your emails, your whole Notion, whatever.

Luke

PS: if you want to actually build this stuff with me, that's what we do inside the Skool community, come join us → https://www.skool.com/luke