When thinking about tech that will be built out / needs to be built out, it is refreshing to have a long time horizon. Like decades, not years

You get to tune out the noise (aka X). The AI agent stuff, the new architecture tweaks, the inference build out, hell even deployed robotics. You become tuned to the long term step-functions: nuclear energy, brain-tech, quantum computing, space travel.

Who capitalizes this? you need an agent with long enough time horizon, because the agent should be coupled to incentives on those time horizons. I don't think VCs can do this, I don't think it works in capital markets. It might be something only the 'state' can finance (it already does, through research)

Focus on work that is slow but something that will really transform the future. Otherwise you get caught up finding cheap wins in the short term, and that's just .. uninteresting

Again, the advice converges to: play long term games with long term people