What Karpathy's struggle tells us about leadership

Andrej Karpathy admitted recently he feels "behind."

One of the world's best programmers (the guy who built Tesla's Autopilot, led AI at OpenAI) can't keep up.

I've been sitting with that for a while now.

Because I think what he's feeling isn't a personal problem. It might be a signal that something fundamental has shifted.

For years, we built our reputations on being the "authors" of our results. We prized the person who knew every line of code or every detail of a contract because they had direct control over the logic. It was deterministic.

But we've introduced this "alien component" into the stack now.

AI isn't a logic machine; it's a probabilistic one. It's brilliant but fallible, which means the old link between effort and leverage is effectively broken.

I'm starting to think the most important skill we need to develop (and frankly, hire for) isn't "technical" in the old sense at all.

It's the ability to preserve authority while delegating generation.

It's like moving from being a craftsman to a systems architect.

In the "Factorio" mindset, nobody cares if you personally hand-crafted a gear. What matters is that you've built a system where gears are produced reliably at scale, even when the individual parts are uncertain.

The ones who figure out how to orchestrate outcomes (rather than just do the work themselves) are the ones who will actually find that 10x leverage we keep hearing about.

It's a strange moment to be a leader.

Did you that NotebookLM make infographics?

LINK HERE=> https://notebooklm.google.com/

Stop treating AI as a "chatbot" and start treating it as a Production Studio.

Google’s NotebookLM just moved past simple summaries. It’s now generating high-fidelity slide decks and infographics directly from your raw data.

Most leaders use AI to "ask questions." Smart leaders are using it to "build assets."

The problem? Most AI-generated presentations look like generic templates from 2012. But NotebookLM is different.

Because it’s grounded in your specific sources—research papers, 20-page strategy docs, or raw data tables—it doesn't hallucinate "fluff." It builds logic.

I recently tested this with a dense research paper. Usually, reading the abstract and conclusion takes 15 minutes, but the "nuance" is buried in the middle. But instead of reading, I prompted it to "Create an infographic summarizing the core technical friction points."

The result? A visual map of the paper’s logic that I could digest in 60 seconds.

Therefore, the workflow for a C-suite executive changes:

1. Upload the "impossible to read" 50-page industry report.

2. Generate a Slide Deck (Presenter Mode) for your next board update.

3. Generate an Infographic for the internal Slack channel to align the team.

4. Listen to the Audio Overview on your drive home to prep for Q&A.

This isn't just "saving time." It’s increasing the velocity of internal clarity.

If you want to try this today, don't just "upload and click." Use a "Specific Lens" prompt.

Instead of: "Make an infographic." Try: "Create an infographic for a non-technical board of directors focusing on the ROI metrics of this implementation."

The tool is free (for now) via Google Labs. Move your research from "unread tabs" to "ready-to-share assets."