ctrl+r #10: AI & cognition, Obsidian+Claude Code

Training brains in the age of shortcuts

🧠 The paradox of getting “smarter with AI”

With many people debating whether AGI is near—or even whether LLMs are truly intelligent—there’s an interesting paradox: will we stay intelligent while using them?

A French documentary (don’t worry, there’s auto-dub) called “La Fabrique à Idiots” (The Idiot Factory) explores the impact of AI in schools—how students use it for cheating and why it’s becoming nearly impossible to assign homework anymore.

My kid is 6, and it makes me think: how can I help him keep training his brain while still embracing the technology?

Someone online told me, “Well, LLMs are just compressed knowledge, like a book.” That feels like an understatement, as LLMs can go much further than simply delivering information.

And there’s science behind this. As mentioned in the documentary, learning requires three neural stages:

• Encoding: the hippocampus creates new connections

• Retrieval practice: the basal ganglia strengthen pathways through repeated recall

• Error correction: dopaminergic neurons fire prediction error signals that tell the brain what to reinforce or prune

When AI handles tasks for you, none of these mechanisms engage. We should watch out for ending up with an atrophied brain—and the choice is up to us.

🛠️ Obsidian and Claude Code

I’ve been playing with the Terminal community plugin for Obsidian to leverage more AI within Obsidian. Not necessarily for writing, but rather for linting and searching. Yes, Obsidian has good search mechanisms and Dataview (so the query I show in the screenshot is actually pretty basic). But I’m looking more for skills like “add a summary of this large note” or “fill in the missing metadata from the source URL.”

📚 What I read/watched

• Why paying DevRel matters: A reminder that DevRel is NOT an entry-level role. Yes, you can create marketing content as a junior, but true DevRel work requires deep technical knowledge and industry experience.

• Creator economy’s abundance crisis: Some good observations about how content abundance has surpassed demand (this applies to tech too) and how you can stay “authentic” when AI can generate so much content so easily.

• OpenAI went ALL IN on this Rust framework: Fun to see that TUI in Rust is so hot right now, given all the AI hype.

• Clawdbot is a security nightmare: Clawdbot (now renamed “Moltbot“), an open-source framework for running your personal AI assistant, gained more than 50k stars in just a few weeks. But as with many new open-source projects gaining quick traction, there are often security holes—so be careful!

• MCP vs Agent Skills: Kaxil (from Astronomer.io) explains the main differences between both in this thread. Will steal and reuse!

---

---

Most important AI note from one of my latest meetings. Yes, I’m lucky I can get decent sleep already.