[This is a slightly adapted version of https://larsmennen.me/10-x-and-execution/ , applying the idea to the context of EA]

You know you ultimately want to be doing good, and you want to maximise your chances of having maximal impact, but there are many roads to Rome. So how do you do this? I think one should pick a project that has an impact of at least an order of magnitude higher than others and then have great execution.

Working on the right thing

In the EA community, we aim to analyze what the most effective things to work on are, and this often is complicated and hard. Spending enough time to pick the right thing to focus on for large choices is underrated in general, but overrated when comparing options with expected outcomes in the same order of magnitude.

When considering multiple options and one of them has an order of magnitude higher expected impact than the others, you should choose it, even if you don’t know the exact impact. We shouldn’t take these choices lightly. It is worth spending a lot of time thinking about them. Ideally, we each should be picking something that will make all the previous look like a footnote. But this can lead people to apply an excessive amount of analysis to attempt to predict which alternative would be best. Although the size of the decision warrants a considerate analysis, for big decisions what matters is the order of magnitude of impact. So if you find that any of the options are 10x more impactful than the rest, you should go for that, even if the path is fuzzy and uncertain.

Another reason for focusing on such projects is that most projects that do not make an order of magnitude of difference are not worth spending your time on in the long run. A great example of an analysis illustrating this is the case for focusing on economic growth and against randomista development. Now don’t get me wrong: I think the top randomista charities on GiveWell are doing tremendous amounts of good now, and we should spend some of our resources on them. But if you value future generations (nearly) as much as current generations, I believe we primarily need to focus on the riskier order of magnitude projects that deliver high impact in the medium- to long-term.

Finally, when we have multiple options that are comparable in orders of magnitude, our natural inclination as a community is to apply more analysis, but for most projects that are more than 5 years out, this seems difficult to know with a reasonable amount of confidence  (although it is unclear exactly how accurate long-term forecasting is). Realistically, you're not going to get the error bars down significantly, at which point even statistically it doesn't make much sense to draw strong conclusions on the comparison.

This focus on order of magnitude and accepting the intrinsic uncertainty seems undervalued. I observe a lot of people that tend to the extremes here: either focusing on high-certainty medium-impact projects (e.g. the ones we already have evaluation data for) or focusing on extremely low-certainty but high-impact projects (e.g. X-risks). But you can be both risky and realistic simultaneously. In fact, the best ideas are exactly those: they're risky, the path towards them is fuzzy and uncertain, you may not know how to do them, or even where to start. But you do know that it's not thin air, that there likely is some path towards it that takes you less than, say, 10 years, that you're confident you can find this path once you explore it, and that within that time frame you can have a huge impact.

When you have selected one or more order of magnitude projects, everything comes down to execution. When you have multiple candidates, you should ask yourself which of them you think you'll be able to execute better on. This is not hard science. It means answering questions such as: Which skills do I have that are relevant? Which one motivates me more? Do I know people that have experience in this? etc. 

Execution

You've chosen a project. Now it’s all about execution. Everything else will be dwarfed by execution when done well. Paul Graham called failure to execute a bigger risk for most startups than being copied.

Great execution is a collection of skills, but the two most important ones are focus and grit. Focus can be incredibly difficult. If you're moving fast, especially in a growing or starting project, a hundred things will seem both urgent and important every day. In reality (and especially in hindsight) only a handful will be; those that make the biggest strides towards the goals. Focus is the skill to identify and only work on those. This is not something you do once at the start of the project. Great execution requires you to do this every single day, perhaps multiple times a day. You'll have to constantly be asking yourself: "Is this the most impactful thing I can be doing right now?". Focus then will allow you to move faster towards the goal.

Besides focus, grit is essential. Along the way, there will be setbacks, other people telling you that you're doing it wrong, and other hurdles. Without grit, it's difficult to make significant progress over longer periods in light of these. Most order of magnitude projects do require consistent effort over a long period before you start seeing payoffs.

Great execution also has a spillover effect, which can be incredibly powerful, but I think is often neglected. If you see people around you execute well, you'll be more likely to do so. Execution can be part of the culture you set; and to run successful impactful projects it should be. You will need to collaborate with people and get other people to execute on your ideas. Building an organizational culture that promotes this then is critical.

There is a pitfall to extreme focus on the goal: the risk that you don't see the bigger picture anymore, lose the ability to adjust your assumptions, and change course if needed. On the other hand, you don't want to be going back to the drawing board every time you hear another view. This is a tricky but healthy balance to strike. You don't want to take other views for granted, but you do want to effectively judge whether there's a core of truth in there and if so, whether that should result in an adjustment of your plans. This is not a skill that is easy to learn, but I do believe it is learnable. "Gut feeling" is often a synonym people use for using great heuristics they learned with experience and practice, although they may not think of it that way.

If you choose the right project and can execute on it well, many things are possible. We often prefer clear path projects, but those are rarely the order of magnitude projects. I think many more people than are currently working on such problems are capable of tackling these. Fuzzy paths will become clear only once explored, so why not try?

Thanks to Ed Ayers and Mithuna Yoganathan for providing comments on a draft of this.