VCs often say, “you don’t have to be an expert” to start a startup. While this is strictly true, being an expert is the most likely way to be a successful founder.
Since you have a stable income, you have a long time horizon to spin out, so it’s more likely you’ll get the timing right. It’s also easier to train to become a technical expert and once your skills are known, others will seek you out with their problems, rather than the other way around. The downside is that since you only periodically optimize your skills for market needs, you have less chance of building the absolute largest startup . But if you want to have a high probability of having impact and a stable life, being an expert is the best way to go.
This is in contrast to starting your first company as a non-expert. In reality, the types of people that have done this successfully are those that already have experience in influence and decision-making under uncertainty  and happen to be close to a simple changing need. But the cold reality is that most first-time entrepreneurs don't meet these conditions so if they don’t have technical expertise or traction, it’s hard for them to command the respect of their peers or build good products. There’s just too many things to learn at once.
A more natural (and successful) way is to keep shipping products with other people in a single technical role, and gradually become an expert in it. If you get far enough along the frontier of the future and are exposed to large markets along the way, you may naturally see opportunities that can turn into startups, and have the skills and respect to recruit others. The downside is that expertise takes time to develop, so you may acquire responsibilities (family, partner) over time that make it harder to take risks. But on the whole, it’s probably worth the tradeoff, especially since the worst-case scenario is still good so you can take on high-risk projects to grow as a leader.
The best way to develop expertise?
80% of this is just shipping things people want. But the critical components that make it a successful path to expertise are being intentional and consistent in role, and shipping in public with great mentors, which makes it much more likely that your work will be good and impactful.
Being intentional about your identity is the first step to expertise since the world will naturally fragment you. There’s good opportunities to turn down, noise to ignore, and you’ll have to make a choice about your future identity before it becomes apparent to anyone else (or even yourself). Your choice is somewhat arbitrary - most roles have a path towards starting a company if you're good enough at them. However, it's good to bias towards underinvested skills that you'll need to build the first product - frontend design for web startups or computer vision for ML. The most important part, however, is not the choice itself but being consistent  and shipping.
To build skills, you can’t just work hard - most who set out to build skills without a feedback system will fail. Instead, join a platform - a role where you can ship work under your name and an aligned mentor.
Mentors are especially important because underinvested skills typically have underinvested teaching materials. They will tell you what you don’t know you don’t know. Mentors can’t prevent you from making mistakes since humans don’t listen until they get burned, but they can help you correct them 10x more quickly. They also help with branding - famous mentors can go a long way in helping distribute your work if you are loyal to them .
The best platform is a small project with a great mentor. This can exist in many places - Google Brain is a good platform, as was Paypal in the early days. In general, grad school has more good platforms than companies, but this varies significantly. Regardless of your global environment, you'll have to be active in recruiting mentors or finding distribution opportunities if they are missing from your local environment, and in being flexible to change environments as your opportunity-generation machine starts working.
Some rules of thumb, from my experience and many of my peers from Stanford and YC:
A PhD is close to the ideal platform, which isn’t surprising since it’s engineered to grow experts, not products. The downside is that many topics are esoteric. However, for useful skills, grad school is the perfect platform - you can publish in small chunks, other students will give you high fives and your advisor is on your team. Although your advisor may have portfolio incentives that can delay you, they still want to see you graduate and make their clever work widely known.
Grad school is a great platform if you want to learn something that’s studied there and have a good advisor. But stay alert to the utility of what you’re studying.
A common myth is that startups are great learning opportunities because they offer responsibility in spades. While this is true, the downside is that without intentionally recruiting mentors, you can’t grow - you have to build your platform while you’re standing on it.
Startups, in particular, are the best place to make big jumps in influence and management, and VCs are good at connecting founders with role models in these areas. The best way to learn most people skills is by watching someone who's good at them.
If you’re at a big company, you can fight your way up the ranks, or periodically take advantage of internal Hackathons - what Ben Newhouse did to spearhead Dropbox Infinite. However, it’s also possible to build your own platform in your day job as an entry-level engineer.
Find a small project that will teach you an underinvested expert skill, and search for mentors who can give you feedback. You may have to offload some busywork or start saying to no to optional tasks in order to do this, which is mildly dangerous. However, in the long run, it’s actually more dangerous to not develop expertise.
A mentor is critical, but underappreciated. You can’t learn by just reading papers in the real world - instead, seek out experts and win their hearts. A surprising number of times, experts don’t work for the company, but you can still ask them questions if you use your personal time and ask for a phone call or in-person chat as a student.
As you build your platform, start shipping small projects quickly (even if embarassing). The sooner you can show an MVP, the sooner you can get the snowball of traction started, which can make a big difference to the success of what is often a exploratory project. If your project becomes a success and gets supported by a team, this early win will be a big boost to your credibility as a rising star who will get more interesting tasks in that area.
This is probably why VCs downplay expertise. For those good at influence, you can learn areas quickly and partner with experts. However, experts can also be founders and learn to manage.
 Often from investment or promoter backgrounds
 You can build expertise in multiple roles, but typically not at the same time. Instead, try to focus as much as possible on one role and get great at it quickly, then move on to the next role. Smart people eventually will develop a diverse portfolio of a handful of roles this way (2-4 is typical). The fewer the roles, though, the smarter you’ll look and more effective you’ll be, so keep it simple. One of the reasons that Sam Altman has had a breakout career is that he started young and focused on one role for the majority of his career.
 While focusing on distribution can initially feel dirty, it's more about later opportunities than self-promotion. If you want to have impact, the best way is to attract inbound opportunities that you might not have considered, since unaided human creativity is limited. The simplest way to attract others is to provide value to them at scale.