Learn by doing (good)

When I founded 80,000 Hours, we hadn’t yet come across the concept of career capital. I mainly compared my options in terms of what I thought would be most impactful in the next few years, only thinking vaguely about what lay beyond that. However, if I had been using the concept, it’s possible it would have led me astray. I would have likely concluded that working in investing would lead to better career capital than starting a nonprofit. But looking back today, I gained better career capital by cofounding 80,000 Hours because I learned more, achieved more, and met a lot of interesting people.

Learning-by-doing is often the fastest way to gain a skill, and this can make it the best option for career capital — and even more so if the work you’re doing is directly relevant to having an impact — most often by failing to do so in the way you’d hoped. If you find you lack a certain skill, you can learn that exact skill.

By working on something impactful, you’ll also likely be more motivated, and people will want to help you succeed. You’ll meet people interested in pursuing the same goals as you. And by focusing on issues that are neglected, it can be easier to achieve impressive results. Many people don’t feel like they have the option to do something impactful right away, but if you do, it can be one of your best options for career capital too.

Don’t focus on what might be most impactful in the next year, since most projects take time to succeed. Suppose your goal is building political support for AI safety. Think about how you could build the most support for it over 3–5 years if you aim high. If you think it’s mass media, then start learning communication skills and apply them wherever you can. If you think it’s grassroots advocacy, then just start organising.

This was the path taken by Sneha Revanur. While in high school, she came across a 2016 ProPublica investigation exposing how algorithmic bias was baked into criminal justice software. The findings stuck with her. When California’s Proposition 25 came up — a ballot measure that would have allowed similarly biased AI algorithms to determine who got bail — she was moved to act.

She started where she was: mobilising fellow students. They organised voter outreach, raised awareness about the dangers of algorithmic bias in the justice system, and built a network of young people who cared about these issues. After Prop 25 failed to pass, rather than let that network dissolve, Sneha transformed it into Encode, now the largest youth movement focused on AI risk policy.

From those early student-organising efforts focused on immediate issues of AI bias, Encode has expanded to encompass the full spectrum of AI risks, for instance, by sponsoring SB 53, a new law requiring greater transparency about the training of frontier AI models, which was recently signed by Gavin Newsom. And this has also given her amazing career capital.

Doing what everyone else is doing is rarely the best path for impact, and now we’ve seen it’s not the best path for career capital either. Common defaults like Google, Deloitte, or continuing into graduate school can be OK, but the best career capital often comes from opportunities that require a bit more work to find: the startup that isn’t yet famous, the policy fellowship you found by crashing on your friend’s couch, or even that modelling career you were considering.