A handful of things I picked up this week

A grab-bag of things I read, heard, or had pointed out to me this week. Not connected to each other. Worth writing down anyway.

1. Naval, on the only sustainable way to sell

From a Naval Ravikant interview: the best way to sell a product is to actually believe in it, have credibility to back it, and be honest about what it does and doesn't do. That's it. No funnels, no clever copy.

It sounds obvious, but it cuts hard the other way too — if you find yourself working overtime on the pitch, that's usually a signal the product (or your belief in it) needs the work, not the slide deck.

2. The 4 Cs — what soft skills look like in the AI era

Caught a talk this week on Ryan Williams' Open to Work. The takeaway I'm still chewing on: as AI flattens the technical playing field, the people who pull ahead are the ones with the soft skills AI can't (yet) replicate. Specifically the 4 Cs:

• Critical thinking — asking the right question, not just answering the one in front of you
• Communication — translating ambiguity into something a team can actually act on
• Collaboration — getting work done through people, not around them
• Creativity — connecting things that don't obviously belong together

LLMs will keep getting better at writing code. They aren't getting better at telling you whether the code is worth writing. That gap is the moat.

3. India is sitting on a quarter of the world's thorium

Random one that surprised me: India holds something like ~25% of the world's known thorium reserves — the largest single share of any country — concentrated mostly in the monazite sands along the Kerala coast.

Thorium is the fuel a lot of next-gen reactor designs are built around (safer, longer fuel cycles, much less weaponizable waste than uranium). India has been quietly building a three-stage nuclear program around it since the 1950s. If thorium reactors ever become commercially viable at scale, the country is positioned in a way nobody talks about.

4. How to actually learn AI (advice I'm taking my own version of)

Best advice I've heard on getting good with AI tools, from a few different places now, all roughly the same shape:

1. Use the tools. Reading about them is not the same. Pick one and ship something tiny by Friday.
2. Forget your job title for a minute. Look at what you actually do in a week — the meetings, the docs, the debugging, the triage. Ask "what part of this could an LLM do at 80%?" That's your first automation.
3. Embrace it as a force multiplier, not a threat. The engineers who end up most valuable in the next five years are the ones who treat their AI tooling like a junior teammate they're coaching, not a competitor they're hiding from.

This one hit because I keep noticing it in my own work — the bottleneck is rarely "can the model do this." It's almost always "have I bothered to ask it."

5. The Wealth Ladder — different rungs, different rules

From Nick Maggiulli's The Wealth Ladder (currently on my nightstand): wealth isn't a single ramp you climb at one constant speed. It's a ladder with distinct rungs, and the strategy that gets you to the next rung is almost never the one that got you to the current one.

The framing that stuck: roughly Level 3 ($1M) is the ceiling of what you can reach by being very good at a salaried job — saving, indexing, compounding, picking employers well. Getting to Level 4 ($10M+) almost always requires something fundamentally different: ownership. Equity in a business, founding something, or a meaningful stake in something that scales without your hours scaling.

The implication isn't "quit your job." It's that the type of work required changes at each level, and most people stall because they keep applying Level-2 tactics to a Level-4 problem. Worth thinking about honestly, even if the answer for now is "stay where I am and execute."