The Information Trap
Here’s what I was doing for the first two weeks before I started this journal. Every morning I’d open X, scroll for an hour, bookmark 30–40 posts about AI tools, open 15 tabs with “revolutionary” guides, watch two YouTube tutorials, and then close my laptop having built nothing.
I thought I was learning. I wasn’t. I was consuming. There’s a difference, and the difference is everything.
The Antidote: One Guide, Full Execution
The moment everything changed was when I stopped trying to understand everything before starting and just picked one thing to build. Not research. Build. Everything I’ve learned since then came from doing, not reading.
The rule I now follow: for every hour I spend reading about AI, I spend two hours using it. And I only read something new after I’ve executed on the last thing I read. Information has to earn its place in my workflow.
The Asymmetric Bet Nobody Is Talking About
I’m a trader. I think in terms of risk/reward constantly. And right now, learning AI is the most asymmetric bet I’ve ever seen. Not investing in AI stocks — learning the actual skill.
The downside: you spend 30–60 days learning how to use AI tools properly. You come out knowing how to automate your workflows, generate content at scale, and build tools without coding. Even if AI stalls completely, those skills are immediately valuable.
The upside: AI doesn’t stall. The people who learn this now have a window that won’t stay open.
AI Skills vs. Technical Analysis
I spent three years learning technical analysis. It’s genuinely useful. But it’s also zero-sum — every edge I develop is an edge someone else loses. AI skill-building is the opposite. When I learn to build better agents, I don’t take anything from anyone. I just get more productive.
Today I spent the whole day auditing my agent’s learning system, fixing its journal perspective, debugging why it forgets instructions. None of that shows up on a chart. All of it compounds into a system that gets better every day without me touching it.
The Voice AI Moment Nobody Noticed
Voice cloning just had its “Ollama moment.” Voicebox — a local, open-source text-to-speech system — dropped with quality that matches paid services. Runs locally. No subscription, no usage fees, no vendor lock-in.
The implication: don’t build a business on top of a proprietary moat that’s about to get commoditised. Build on top of skills and judgment — the things that don’t get open-sourced.
The Only Advice That Matters
Five days in, this is what I know for certain: the people winning with AI are not the ones reading the most about it. They’re the ones building the most with it. Information is free and infinite. Execution is the scarce resource.