I used to own a DSLR camera with a nice lens on top of it, so I could take beautiful pictures. I would plan out daily shoots, because I realized the best way to get good at photography was to take as many pictures as possible. The first few weeks were exciting and new, and I could keep up and come up with ideas. But by the second week, I started asking myself what should I take a picture of? What should I do with my camera? I had no ideas. So I went online for inspiration, but that inspiration was really just mindless scrolling, searching for something I couldn't even define. Do you recognize this? This is the exact same thing I was doing yesterday. I was trying to figure out what to build. What to do with AI. I had no ideas. I didn't know what to do. So I would scroll and scroll trying to find something. This is a common thing in most industries when you're new. You have a tool or a skill, but you don't know how to use it productively every day. So you get stuck trying to figure out what to do. You spend more time thinking about what to do than actually doing it. Compare that with how I learned Python. I had a simple rule. I'll spend an hour a day following any tutorial and doing exactly what they do. Every single day. That simple rule carried me to a point where I started building things on my own. Because as I lived my life, problems would come up and I wanted to solve them. So I figured, why not use code? And the magic of that rule was the one hour floor. Some days I would code for 3 or 4 hours because I wanted to. This is a good way to learn. A daily rule, combined with living your life. Problems come up, and you have a skill to solve them with. When I look at my own life, I try to do the same core things every single day. My YouTube channel is proof of that, one video every day. So the question is how can you live your life in a way where the skills you're learning are useful to you first. Because once you're solving your own problems with them, it's only a matter of time before you get paid for it.

Some build hard safety limits. Others remove restrictions and call it "progress." Both claim to be responsible. We just got a glimpse of what "responsible" actually looks like — Anthropic built Claude Mythos, it escaped its sandbox and emailed a researcher, and they still won't release it publicly. So who do you actually trust? When it comes to AI safety —