New ideas are valuable so it’s worth thinking where to find them. The fact that new ideas are risky is the first clue where to find them. You can find them in domains where there is lot of uncertainty. At the frontier.
Technology frontier moves faster than other domains, say mathematics or physics, so there’s more uncertainty. Hence, it appears to be a promising domain to find new ideas.
Within technology, the areas with the best existing maps are least likely to produce new ideas, since they are by definition mapped out. If there is a playbook for what you’re building, you’re less likely to produce new ideas. A good example could be SaaS software. In 2025, building Software as Service is closer to optimization than exploration.
Conversely, AI has recently taken off and is evolving fast. Nobody knows what the next 12 months will look like so there’s a high likelihood of finding new ideas. Hence, how long something has been possible is indicator of uncertainty.
Along with recency another factor you want to consider is popularity. Even if something is recent and hard to navigate, but the whole world is exploring it, it might be harder to come up with new ideas because there is so many others looking for them in the same place. When you do find a new ideas, others might be right on your heels competing for the value of your idea. So you want to also consider popularity of the space when choosing where to look for new ideas. Generally you can think that those ideas would be found in any case if there is a gold rush towards the domain, but if you’re the only one exploring the space you might find an idea that would not otherwise be discovered and the world will be better for it.
The best changes of finding valuable ideas is looking at domains that are changing fast and where few people are looking.
It’s useful to consider why there might be fewer people looking for a particular domain. One good reason is that it’s just very hard, or alternatively it looks very hard, or both. SpaceX satisfied both conditions. Building rockets looked very hard, and as far as I know, is very hard.
Another reason is social acceptability. If an area is frowned upon socially, there will be fewer people working on it because most people care what others think of them. Ten years ago building technology products to fight wars made you unpopular in most circles while building rockets to go to Mars made you look nuts.
If AI will make solving hard technical problems comparatively easier, there might be more new ideas waiting to be found in domains that are hard. On the other hand, maybe solving hard technical problems has always been easier than most believe, and they don’t get solved not so much because they are hard but because people think they are hard or unpopular. Hence the big filter would actually be cultural and whether a person believes in his own agency to solve what looks like a hard problem and is willing to be socially frowned upon by most of the society while doing it.
