It’s a coordination game…

Let’s imagine you want to found an organisation, project, or something else. The world is currently in an equilibrium where the thing does not exist.

There’s some better equilibrium where the project exists and is flourishing – it has funding, a great team, efficient systems, people see a bright future for it, and it’s doing work that’s really valuable. In between, there are a bunch of not-flourishing states – maybe some of them are also equilibria (but much less valuable ones). Others will devolve back into the ‘not existing’ state.

Your goal is to shift into the great equilibria. To do this, you need to boost talent, funding, expectations etc. in a coordinated way.

How do you do this?

• Have a really strong vision

  • Iterate on this vision as you talk to the funders and potential hires

• Build momentum before the ‘jump’ where you found the project

  • Have good relationships with funders and potential hires

  • Do small hops before the big jump

• Coordinate the jump: (accurately) let funders know that top talent is interested, and let top talent know that funders are interested. To do this with integrity, you need to kind of gradually shift up interest with each of these groups.

… but you should play it with integrity

I think that this sort of dynamic is (a part of) what caused Theranos to become a fraud. They were trying to do this coordinated jump to a world with great talent working on an important problem. The jump failed, but they kept claiming that it was succeeding.

They had a cool vision! But rather than selling the potential, they lied (even from early on) about how close they were to that vision (and they probably deceived themselves towards believing that lie.) Then they tried to cover up the lies and pretend that they’d made the jump when they hadn’t.

You can have a bold vision and also be clear that you haven’t reached it yet (and that you might fail and never reach it): this is often enough to get great support for your project.

Subscribe now

Some made-up examples:

1. Frankie doesn’t want to work at 80k anymore – and has started applying elsewhere.

2. Arpita’s manager Abe doesn’t think that Arpita is currently strong enough at overseeing 80k’s tech security and is sceptical that she can grow into the role.

3. Gavin is frustrated that 80k has been failing to evolve in the way his Tour of Duty indicated it would.

4. Leda would really like to work on podcast editing – but that would mean being managed by Joe, so she’d prefer to stay on Kiara’s team.

When people are new to management at 80k, they often lean away from addressing these kinds of topics.

This is very understandable! The topics are somewhat unpleasant; it feels like there might be some harm done just by looking at them or by making them salient; it’s not clear that they’re tractable; and it’s possible to bungle the conversation or hurt someone’s feelings.

However, I claim that it’s usually better to address uncomfortable management ‘truths’ – and that experienced managers1 tend to update towards this view.

Here are some of the positive things that can happen when you address these truths:

1. Sometimes you’re mistaken about an aspect of it.

2. You feel a sense of relief. The discomfort you felt avoiding it is reduced.

3. It moves the problem away from being an issue with the other person to a shared problem which the two of you can look at together ‘as a team’ and determine how best to approach.

4. The standard benefits of organisational transparency.

   1. Creates a culture where people don’t need to guess what’s happening / waste their energy on it.

   2. Increases the # of people with context who can help solve things.

   3. etc.

5. You now have the ‘language’ with which to talk about the uncomfortable thing – such that you can name it when you talk about how decisions are being made.

6. You can get started on actions necessary to deal with the situation sooner – succession planning, moving someone into a new role, etc.

There’s lots to say about having these conversations well – doing so requires a substantive set of skills!

If I could point you to three resources on this subject, they would be these:

1. The book Crucial Conversations

   1. The authors say a crucial conversation is “a discussion between two or more people where (1) stakes are high, (2) opinions vary, and (3) emotions run strong”

2. The field of Non-Violent Communication ∑

3. 15 Commitments of Conscious Leadership (workshops / books / coaching)

If I could only say three sentences to you, they would be these:

During this conversation, you need to radiate a sense that the two of you are fundamentally on the same team and that the two of you are going to treat one another with kindness as you navigate this shared problem. If you can’t think of a way into naming the issue, allude to it in a few sentences and invite them to describe how they see the issue. They will very likely have done some work on characterising it, you’ll learn a lot just by hearing how they’re thinking about it, and starting the conversation with a framing they’ve been using is a great way in.

Subscribe now

First, I find that it’s an arena for becoming the sort of person I want to become.

Doing strategy forces me to think hard about what is going on in the world, look into the darkness, and come up with plans that might work anyway. Performance management means delivering hard feedback with care. Coaching helps me notice my own brokenness, give tough feedback directly but kindly, and try to imagine and nurture a better version of people. Leading a team means I need to listen, take care of culture, and take ownership for the group.

Second, I get to try to help others become who they want to be, even while I’m learning myself.

I’m far from perfect at all of this, but I feel lucky that my work is intertwined with the quest to become a stronger, wiser person.

• Give them big ambitious targets that you’re not sure they’ll be able to achieve. Support them and make sure it’s not catastrophic (either practically or emotionally) if they fail.

  • Look for things that are a stretch but not impossible

  • Look for things that lean into their strengths

• Really give them responsibility for the areas that they’re owning. Set context, rather than controlling.

  • Do:

    • You can (and should) coach them

    • You should stay synced up with them on what their goals are, which goals are most important, metrics for success, etc. (writing is the best way to keep this transparent)

    • If things don’t go well, you can change their responsibilities such that they no longer own the area (or are fired or whatever)

    • Give them feedback about how you think things are going, and your overall assessment of their performance

    • You can even share input as long as you’re very clear that they can ignore you

  • Don’t

    • Do their job!

    • Hand down decisions

    • Insist on co-creating plans for hitting their goals

    • Meddle with them doing their job!

For senior staff who you trust, you should also try to get them involved and feeling responsible for running the organization with you: set up leadership team meetings such that you’re discussing the biggest issues together, and taking some collective responsibility for making things great (though the buck stops with you). If you can create a leadership team that’s working together to run the organization, that’s the best step you can take towards making yourself replaceable, and it’s probably better than having a single talented successor with a weak leadership team.

The familiar version of the worry is about whether the organisation will hire someone else in your place. If it will, then the difference you make by taking the job is only the difference between you and them.

There are a couple of well-known answers to that. The person hired in your place probably cares about the world too, and would go on to do something useful with their time instead (Ben Todd wrote about this in 2015). And in a hiring round, the top candidate is often a lot stronger than the next-best one (Rob Wiblin worked this through in 2019).

These days, I think it’s very common for organisations to do bar hiring rather than slot hiring: if we don’t hire you, we won’t hire anyone else.

If someone was considering joining as a podcast host or a writer, that would resolve the replaceability question: it’s clear that their addition to 80k is approximately equal to their podcast episodes or the pieces they write.

But for people in managerial or generalist roles, we often hear a different response to the replaceability worry, which goes something like this:

“80k is already doing pretty well. It’s big, stable, and it has a few committed senior people. If 80k doesn’t hire me, the staff who are already there will keep looking after whatever I would have done. That means I won’t have much impact there and I shouldn’t work there.”

I claim that this line of thinking is mistaken, because it doesn’t reason past the first-order counterfactual.

Of course, it’s literally true that a given responsibility could otherwise be done by someone else.

For example, you could join our People Operations team and run our September payroll. If you didn’t do it, then our Head of People Operations would run payroll. If she couldn’t do it then our Director of Operations would do it (or ask someone on her team to do it). And if she couldn’t do it then our Chief of Staff would do it (or ask someone else to do it).

It’s very unlikely that we don’t do our September payroll, so the first-order analysis shows that it’s entirely replaceable work.

Thinking through the second-order analysis (but skipping the third)1, you’ll notice that our Head of People Operations might otherwise have done a deep dive into a tension in a management relationship and substantively improved how the two of them communicate, our Director of Operations might have reviewed the case for a solo Bay Area office and changed her mind on two aspects of it, and our Chief of Staff might otherwise have taken two meetings with speculative candidates to start new programmes.

Somewhere at the bottom of that chain is a piece of work which doesn’t get done: the deep dive, the office review, or the meetings with would-be programme founders.

It isn’t trivial to know what 80k will get done because someone is now running our payroll. But it is trivial to reason that something extra will be done. Similarly, if you accept an offer to join, it isn’t trivial for us to reason through what extra will be done because you came. But it is trivial to notice that, in expectation, we’ll be able to achieve more if we’re 56 people than if we’re 55.2

This means that if you are working at 80k, or in fact at any large organisation in the AI safety ecosystem3, you’re really likely to be making a contribution which wouldn’t happen without you.

These might seem like straightforward points.

And they’ve been made before and more rigorously than I’m making them here.

But I do still run into the concern that working at a given organisation or in the field generally is replaceable.

So I’ve decided to emphasise what I’m saying here by whining about what it feels like to me when someone tells me that they don’t want to work at 80k because we don’t need them.

What follows is a description of the bottom of this chain at 80k: the things we think are worth doing, and aren’t doing.

Rant

We’ve been working on helping talented people tackle the world’s most pressing problems for 15 years, but the pressing problems still seem erm… Pretty bloody talent constrained.

Like, we are really really trying, and maybe we look like we’ve got our shit together, but in comparison to what our mission and the world are demanding from us, we just suck at what we do. And so we need your help.

I think that most of our existing programmes are markedly below the scale and impact which they could be operating at. I think our research and movement building strategy is further from the bleeding edge than it used to be. I think that we’ve probably lost ground relative to other efforts aimed at getting people into the field in the past couple of years, such that if we repeated the cross-sectional surveys we’d look worse than we previously did.

We totally missed the boat on fellowships. These turned into the ~best scaling movement building innovation of the past 5 years, but we never set one up — even though they’ve been on our radar since I joined in 2017.

We launched our AI career advisor in August 2026, a mere (checks notes) 2.75 years after we first prototyped the idea.

Marketing took us years to get off the ground and we’re still below where we’d like to be by a factor of somewhere between 3 and 30.4

I could go on.

I will go on.

We’ve repeatedly failed to spin up a strong headhunting service for the ecosystem. We had three false starts under different leads, using various approaches, and we’ve only recently begun scaling the approach we’re currently using. We’re well below our potential with the current (more scalable) version of headhunting — by at least a factor of 4.

We have wanted a YouTube channel like AI in Context for years (everyone was nagging us about it); we only managed to launch one last summer. And now that we have one, we wish we were publishing way more on it, being more ambitious with it, and running more channels. Also, we’re still not doing well at shortform video (TikTok etc.) — but surely there’s gotta be a viable strategy here…

The number of people we’ve given one-on-one advice to has flatlined since 2022.

We gave away half a million in talent development grants, then lost the ability to make grants when we spun out of Effective Ventures. We haven’t prioritised rebuilding it, as it would take some work (and probably also because we suck).

We didn’t spin out of Effective Ventures as early as we should have, our recruiting team started in January 2026 (after one false start), and we don’t have a US entity because we don’t have the capacity to set one up.

/rant

Overall, I just think it’s bloody contingent that we ended up being this level of good.

I think we’ve averaged something like 30% growth5 a year across 80k.

Said like that, it sounds lovely and smooth, perhaps even overdetermined.

But I don’t think it’s been overdetermined, and I don’t think much of it has been a ‘top down’ decision based on reasoning through when we expect programmes to hit diminishing returns.

Instead, what we’ve been able to do seems extremely contingent on particular staff who managed to make particular contributions, and on particular staff who we failed to hire who didn’t make contributions.

It’s really easy for me to imagine an adjacent version of history where our overall growth rate was more like 50%, or where we shrank. Indeed, in some years, some programmes have more than doubled, in other years programmes have shrunk.

And it seems very possible to me that the returns on resources invested into 80k have increased about as often as they’ve decreased. I think it’s quite possible that our video programme is more cost effective than the rest of 80k: such that 80k saw increasing returns when we added it. And I’m fairly confident that our headhunting team is going through a period of increasing returns.

To summarise: I claim that, properly considered, your work at a given large AI safety ecosystem org is ~not replaceable.

I’ve given a few arguments for that, and I’ve done a lot of whining. I hope that you found at least one of those convincing.

There’s some (net) benefit to doing a task. That benefit generally1 declines over time, but the functional form varies a bit.

For some tasks, there’s this kinda cliff-edge drop-off in benefit. For example, with a tax return, it’s basically not any better to submit it Dec 1st vs. Dec 31st, but it’s way better to submit Dec 31st vs. Jan 1st.

For other things, I think that the drop-off is pretty slow, or even totally smooth, and it’s just about getting it done sooner. I think that most things are more like this.

Against fake deadlines

In cases where the value of a task diminishes smoothly, I think that setting deadlines at all is a distraction/stress-inducer without much benefit.

I’d rather just know that I’ll do it when I can. This makes me less stressed.

I think not having a deadline also better reflects the reality of the benefit/time curve, which means I can prioritise better: e.g. if I know that the benefit of a task is dropping off smoothly, and a new high-priority thing comes up, it’s easier to switch to the high-priority thing, rather than getting stuck on meeting an arbitrary deadline for the lower-priority thing.

Why still have deadlines?

Sometimes you actually have a cliff-edge case! In this case, I think deadlines are good, and I use them, but this is rare (in my work).

Some people seem to like having a clear deadline to work towards, even if it’s (in some sense) fake. People’s psychology on this varies, so if deadlines are useful to you I think you should use them.

Deadlines can also be a tool for coordination, and it’s important to manage deadlines cooperatively. For instance, if someone else gives you a deadline, try to meet it or renegotiate it.

Two types of deadlines: “Best before” and “Expiry”

I now basically don’t use deadlines at all, but one thing that I tried for a bit and liked is having two dates for tasks:

• “Best before”: roughly, the date at which the value of completing the task is most rapidly declining.

• “Expiry”: the date after which there is approximately no value in completing the task.

(In analogy to food labels.)

Some tasks have both best before and expiry dates (e.g. it’s best to fill in our meeting agenda by Monday evening, and it’s useless to fill in the meeting agenda after we’ve had the meeting). Some just have an expiry date (in which case, if you failed to do the task by then — for example, because it wasn’t super important — you can just delete it from the list at that point). Some just have a best before (e.g. it would be most valuable to share an update with the team before all hands, but it would basically always be valuable to share it).

This is still simplifying the true thing which is the benefit/time curve, but I think it lets a bit more nuance in.

Most centrally, hiring a CEO.

But it could also be hiring another very senior role, or hiring for outlier talent (e.g. research talent). Some of the below also applies to hiring managers in general.

Why is it different?

I think executive hiring is very different from hiring for more junior roles.

For instance, often:

• You will have very few plausibly qualified candidates.

• You may well know all of them, and likely the top candidate is at most two connections away, because you’re hiring for expertise in a particular field.

• Those candidates are likely to be very busy (because they’re probably already doing high-impact work).

• You will be assessing their ability to succeed at tasks that take many months or years (managing people and teams, setting a strategy, developing a research agenda), so trial tasks won’t work for assessing their performance.

• More of the alpha is in persuading the very best people to be excited about the role, rather than in filtering people (though filtering is still crucial!)

What does this change about how you should hire?

• Filter first: You should do most of the filtering before you reach out to people, and then do a very tailored, white-glove process for those you reach out to.

  • How do you filter before you reach out to people? Ask people in your network for recommendations, then try to talk to people who have worked with them, and basically get an informal reference. Look at public track record where applicable (easy with researchers, harder with execs).

  • Then when you reach out you can say “You’re one of <10 people who we’re excited about after having combed through our networks. Would you be up for a call?”, which is more flattering for the candidate and means it’s more likely to be worth their while to engage with the process (since they have a >10% chance of getting the job).

  • Then the process should be extremely white-glove and tailored. You shouldn’t waste candidates’ time, and you shouldn’t test things that you’ve already seen. Instead, you should reflect on your key uncertainties and then do specific things to get more info on them. And you should ask about, and test, their cruxes.

• References become much more important, because they help you assess candidates’ abilities at the longer-time-horizon tasks. E.g. you can’t get easily test candidates’ ability to do the full strategy-setting process at your org, because that takes months. But you can hear how well they did this at their previous company.

  • Getting useful signal from references is a skill, about which I might write more some time, but you can get useful info from references.

I think that this book is useful. It’s not perfect, but it is usefully different from “standard EA” takes (lots of work tests etc.) It’s more tailored to executive hiring, and has some useful templates for reference calls and interviews.

This means that the time of people with context and relevant skills is very valuable: usually more valuable than their salary.

When taking this effect into account, it turns out that:

1. The net cost to the ecosystem of an organisation made up of high-context and relevant-skill people is much larger than the financial costs of that organisation.

2. The net cost to the ecosystem of an organisation which trains many staff might be much lower than the financial costs of that organisation.

3. The return curves of these organisations can get pretty whacky, in a way which will sometimes affect strategic decisions.

4. Giving your staff a good experience in the ecosystem and enabling them to increase their context / skills is important: it’s likely a meaningful fraction of your organisation’s impact.

Here are some hypothetical figures to illustrate what these effects might look like.

If you’d like, you can play around with these figures here.

The rest of the post is focused on better explaining the above and belabouring implications. I expect that if the above is intuitive to you and you’ve already absorbed the strategic implications, then you should skip reading further.

What a hire costs

Let’s consider the net costs to the ecosystem of an organisation employing a staff member:

1. Salary

2. Opportunity cost

   1. Best quantified from the point of view of aligned funding: roughly, how much an aligned funder would be willing to pay for the impact they expect this person could produce if they weren’t working at their current org.

3. Capacity-building value

   1. Working at the organisation affects the expected value of the rest of that staff member’s career: they gain skills and context, and based on their experience there, the job increases/decreases their inclination to remain in the ecosystem.1

   2. They’re also an easier hiring decision for the next organisation: they know that they’re a good fit for the work and they’ve learned how to do it, and their first employer in the ecosystem is able to vouch for these facts.2

So the net cost of a hire to the ecosystem is their salary, plus their opportunity cost, minus the capacity-building value added by them being employed by you.

A worked example

Let’s imagine you’re a generic organisation in our ecosystem with five staff. The figures are made up (they’re the ones in ‘What each hire costs the ecosystem’ above; you can play around with them here).

The founder’s opportunity cost to the ecosystem is very large, because this person would otherwise be founding another organisation, or running some other critical project. In fact, most of their cost is from opportunity cost, rather than their salary.

There’s a large capacity-building benefit, as their career capital (skills, connections, credentials, and financial runway) increases through their work.

The operations coordinator is the same.

For the two researchers and the graphic designer, the most important term is capacity building: because there was a chance they wouldn’t work in the ecosystem if they didn’t take your role. Working for you makes them more likely to take a role in the ecosystem after they leave your workplace.

For the researcher who came to the field mid-career, it brings the net cost down to about £10k. For the research fellow and the graphic designer, their capacity-building value is larger than their salary plus their opportunity cost, so their net cost to the ecosystem is below zero.

Why the return curves get whacky

If you include the opportunity costs of people who have been in the ecosystem for a long time, and the benefits of training people new to the ecosystem, the economics can change a lot!

Properly considered, the lean organisation staffed by a few old-timers turns out to be very expensive in ecosystem resources, while a very financially expensive organisation staffed by loads of people new to the ecosystem turns out to be ‘cheap’. They might even pay for their financial cost in capacity-building value.

To compare returns, I’ve said that the founder working alone would produce 100 impact points a year. Each hire then adds some points as they join: 50 for the operations coordinator, 40 for the researcher, 20 for the graphic designer, and 10 for the research fellow. Per person, the returns go down. But per £100k of ecosystem resources, they go up. The founder produces 11 impact points per £100k they cost the ecosystem, the operations coordinator 17, and the researcher 400. The other two cost the ecosystem less than nothing, so they’re free before you count any of their work.

Here are those return-on-investment graphs again. Look at all this silliness:

Put your own numbers in

I’m somewhat confident that the effects described above hold.

The specific figures I used are totally made up though: if you’d like to get a sense of how these effects should change your strategy, you can play with a model here — brentonmayer.com/talent-leverage.

It’s tempting to just try to hire board members who have the skills you need.

Instead, you should design a board like you build a team: get a combination of experts with the skills you need, then help them to work well together.

Selection

I think there should be about 5 people on the board.

Criteria:

1. Everyone on the board has:

   1. Some basic alignment with the org (believe in and understand its mission at a high level)

   2. Broadly smart and sensible.

   3. Ability to coordinate and be reasonably concise

2. Three of five are quite strategically aligned (e.g. at the level of a mid- to senior-level staff member at an organization doing similar work)

3. Active trustee: At least one of those three has the management experience and capacity to do a serious annual review of the CEO’s performance (ideally experience managing managers, at least 10 person-management-years)

4. Chair: At least one of the five has the time and leadership skills to chair: run meetings efficiently, keep people on task, etc.

5. Treasurer: At least one of the five has some professional experience dealing with legal and/or finances and the capacity to seriously engage with the organization’s audit reports etc.

6. Nice to have

   1. Comms

   2. Crisis management

   3. Professional experience in things related to the organization’s work products (e.g. an organization focused on producing research should have someone with research expertise on the board)

Track and avoid AI COIs.

I think boards should mature over time:

• For a small startup I think it’s mostly fine for the org to be founders + people the founders trust: if the founders are good, they should get support and be unblocked; if the founders are bad, the org will fail regardless of the board.

• Once the org becomes more mature and controls more resources, the board should have more external people, and do more scrutiny of whether the founder/CEO is still the right person to lead the org through the next stage of growth.

Using the board

• The active trustee should meet with the CEO every 1-4 weeks (depending on how well the CEO is doing and how much support they need).

• The treasurer should do check-ins with the ops team every 1-2 months, and dive into audit reports, financials, etc.

• The full board should meet every quarter for 2h

  • Before this meeting

    • It might be good for the CEO to meet one-on-one with most board members to brief them and ask questions.

    • It’s useful to have a ~5-page report sent to the board a week before the meeting, so that you can do updates asynchronously before the meeting.

  • In the meeting:

    • First, the board should meet without staff to discuss the CEO’s performance. It’s good to do this reliably even if there are no concerns before meeting.

    • Then staff should join and you should discuss any questions, issues, or concerns from the update. The board should really dig into how the org is performing and feel on the hook for making sure it’s good.

    • Then the CEO might ask for input on any active conundrums, or things that need board-level discussion.

    • (Maybe the board needs to sign off on some things, but this should normally be done asynchronously / shouldn’t take too much time.)

• Annually the active trustee should write a performance review for the CEO, which the rest of the board should review and sign off on.

  • As preparation for this, the active trustee should review an anonymous staff survey, probably chat to a handful of the senior people, look at what the org has produced, and ask the treasurer how ops stuff is going.

This means that leadership should approach capacity-building as if they were at a startup, rather than a steady-state organisation.

Here’s a stark way to put this is:

Aim to be at a point where you’re not required to do much work to maintain the functions you oversee. And then build more. All functions you’ve built should be ‘owned’ by someone who isn’t you. Those owners should independently be doing a strong job of overseeing their work.

Your oversight should look like: inputting into the highest-stakes decisions within those functions, generally figureheading, coaching and developing staff, determining team direction, overseeing coordination across the team, and so on. You’ll often dive into the object-level details of a team you oversee to gather information for assessment or coaching, or to execute on one of the few projects which the owner can’t get to yet.1

For a team in our ecosystem, which is likely a) growing and b) facing a complicated strategic environment, I’d expect that if someone managed to get their team into the situation above then it would look like them:

1. Spending ~20 hours per week on team leadership.

2. Being able to drop out of work for 3 months without anything breaking.2

Now, of course, at the point that someone reaches the position above, they typically won’t just drop out of the office for 3 months. Instead, they’ll put their time elsewhere, for example:

1. Spinning up a new function. For example, I spent around half of my proactive time in the second half of last year spinning up a strategic analysis function (which didn’t really previously exist), and our Head of Growth aimed a quarter at looking into new verticals to spin up.

2. Taking on one of their boss’s responsibilities. Obviously which responsibilities you can take on depends on your respective fits and career aims. But in most manager–direct relationships there is usually some responsibility which the direct could plausibly grow to take on.

3. Firefighting. Dealing with someone on the team quitting unexpectedly or getting sick, or some other crisis.

4. Recovery. Going through a period of ‘jogging’ rather than sprinting, to recover and enable them to sprint hard in the future. This might include some combination of working less and developing personally.

5. Leaving. If none of the above seems valuable enough to do, then working to hand the team over to someone else and leaving.

I’m worried that people might misread this and decide that they should just hire someone, declare that they own a responsibility, assume that it will go well, and leave them to it. That isn’t sufficient. Instead, you need to do the oversight work described above. This is a lot of work!

In fact, you might never get to the point where you’d be able to drop out of the office for 3 months. But being marginally closer to that point and being able to drop marginally more time into options 1-5 above is valuable, so you should build towards it anyway.

I think that my considered view is that most leaders in our ecosystem should be aiming for a setup like this3, and that ~all operations leaders should be.

If you lead an operations team at one of the organisations we write for and you don’t currently share that view, there’s a good chance that speaking to me about it will change the orientation you take to your work, and increase the value of your work. If you’d like to have that conversation, fill in this short form and I’ll be in touch.

At 80,000 Hours, we’ve recently been puzzling over the opposite effect: as we’ve grown, we’ve felt more agile.

The podcast team, now with 4 hosts and a stronger production function, has been able to experiment more with different formats and release pieces quickly to keep up with the news cycle. Two members of the senior leadership team are spending this quarter thinking about strategy and new programmes to spin up, we launched our AI career advisor, and we’re trying out a second video programme.

What’s going on here?

Let’s define being nimble as being able to change the work you’re doing, or to execute on work which you weren’t previously doing.

To do that, you need:

1. Staff who are flexible enough to execute on a change.

2. The strategic capacity to notice that there should be a change, and to design that change.

3. The organisational skill and will to make the change. This includes cultural norms around disagreeing and committing, enough clear authority that decisions aren’t ‘made by committee’, and willingness for staff to make and execute on calls which might relatively advantage or disadvantage some staff or worldviews.

To me, this gives clear insight into why large organisations tend to be less nimble:

1. Team members are not selected for flexibility. Founding teams are extremely flexible — to a first approximation, the founding team should be able to do (or cause to happen) ~everything required for the org. As you grow, you can hire folks who are more specialised and less able to do ‘whatever comes up’.1

2. Strategic capacity can fall, because senior staff carry large management loads. (Though of course, the total number of people who could do strong strategy work with their time increases. So if organisations can preserve that, then strategic capacity could increase as they get larger.)

3. Organisational skill and will tend to decline. The average staff member will be less attached to the mission, trust throughout the org will fall, committees will tend to pop up, etc.2

Where does this leave a large organisation?

A large organisation should never expect to be as nimble as a recently founded one:

• Not all of 80k’s staff could (for example) pivot to running an AI safety fellowship, and we’d be making a mistake if we only hired people who could (though it continues to be very useful to hire flexible staff).

• Not all of our staff’s time can be reallocated to other work (and we’d be making a mistake if we dropped some of the highest-value work we have access to, like if the job board stopped posting jobs).

However, they should value their ability to be nimble, and cultivate it. Here’s what we’re aiming for at 80k:

• Celebrating not just when people raise thoughtful objections, but also when they disagree and commit.

• Celebrating when we cut down bureaucracy.

• Being particularly proud of our finance procedures, which don’t limit teams to a set spending amount, and which usually allow people to use their judgement rather than waiting for sign-off.

• Transparency, strategic clarity, and high ownership: people can choose to approach their responsibilities in whatever way will best advance 80k’s mission.

• Encouraging people to take on more responsibility and move to higher ownership.

• Preserving leadership time for strategy and new initiatives.

I know of places which have chosen to stay small in order to preserve agility. I expect some people might hesitate to work at large organisations as we approach the singularity, for fear that their bureaucracy will slow them down too much to react well. Our experience at 80,000 Hours, and the arguments above, make me more sceptical that they are making the right decision.

Most organisations aren’t set up to work well in a crisis.

At 80,000 Hours, we’ve found that work slows down across the org as staff are worried and distracted. Also, the CEO ends up stuck with figuring out what to do more than they should.1

A crisis task force solves these problems.

Set it up

The main move is a second org chart, which sits alongside the normal one for as long as the crisis lasts.

1. One person leads it. They own the crisis, and they report to the CEO (or whoever it would otherwise have landed on).

2. A few other people join it, and it becomes their job. Their normal responsibilities get handed to someone else. At the start, pull them out completely: they should feel like the crisis is their job and they’ve left their team to deal with it. Once things settle down but aren’t fully wrapped up, it’s fine to move task force members mostly back to their team, with an hour a day or so on the task force.

3. Name it after the crisis. For example, ‘the Hugging Face task force’.

4. The task force takes on any project which the crisis throws up. At least, by default. If there’s a piece of work which someone in the rest of the org is clearly the right person for, and it won’t wreck their week, the task force scopes it and hands it over. But that’s the exception: the rest of the org’s job is to keep the org running.

5. Ideally, it’s literally a room. Everyone in the same place, lots of back and forth. Failing that, you could constantly be on a call or in a virtual workspace like Gather.

Who to put in it

Which of these you need depends on the crisis. Go down the list and pick.

1. Someone on the news. Following what’s coming out, so nobody else has to.

2. Someone working out the strategy. What’s actually going on, and what it means for us.

3. People running projects.

4. External comms.

5. Comms to the rest of the org. Someone who owns the daily message (see below).

6. Legal. Someone who can approve comms and give advice fast, with context, rather than a lawyer who’s hearing about it for the first time. Push them hard for clear guidance on what you can and can’t do, so that you can get moving. This is always a hassle.

7. Technical expertise.

8. Someone from the team the crisis will change most. So that afterwards, that team has someone who knows the whole story.

Run it

1. A daily meeting. What did we learn? What did we do? What are we doing today?

2. Much faster, much more synchronous comms than you run in peacetime. At 80k, much of the org is set up with a Cal Newport-like communication cadence optimised for letting people focus. In a crisis the picture keeps changing, so we’re optimising for pushing bottlenecks forward more intensely as the situation changes, and that needs a lot more talking.2

3. A daily message to the whole org. Someone posts an update every day: what’s happened, what we now think the overall strategic picture is, and what the go is from here. That lets everyone else get on with their normal jobs, because they can see it’s handled and don’t each need to be following the news.

4. The lead manages rather than does. The value of the lead’s work is the value of what the task force produces. In a complicated and changing situation, that usually means they spend their time understanding the strategic picture, updating on it, and delegating and managing the work, rather than doing it themselves. I find that hard to do. What I try to remind myself of is the team leader in a resuscitation. They must not start putting in drips or examining the patient. Their job is to physically stand at the end of the bed, keep track of the evolving picture of what’s going on and what to do about it (the diagnosis and the treatment), and for goodness’ sake keep their head screwed on.

5. The lead is responsible for the people on the task force, the same way a manager normally is: their output, and how they’re holding up. If someone is cooked and can’t function, tell them to go and get some sleep. If someone needs to go harder, ask them to step up.

6. The lead keeps asking when to wrap it up. Aim to wrap up the task force, or move people back to their teams, as soon as practical. There’ll usually be a period where people are partly in the task force and partly back in their old jobs, but I think it’s useful to ‘extremise’ a little, so that people are either fully in or fully out.

When does a crisis task force make sense?

1. Your org isn’t set up for crises by default, the way an election campaign, an emergency room, or a comms org is.

2. You’ve got more than about 10 people.

3. Crises where the news keeps changing and each change has strategic implications for the org: a sweeping executive order, FTX collapsing, Hugging Face, your major funder pulling out, etc.

There are lots of systems like this. 80k started using it more or less at random after a former civil servant who’d worked in them brought it into 80k.3 Claude has written up a few of the other approaches here.

80,000 Hours has used crisis task forces since FTX collapsed, and if I had my time again I’d have set the first one up faster and more definitively.

Play with the model here.

Underhiring to the 80,000 Hours operations team was one of the most consequential career mistakes I’ve made.

One reason I made the error was relying too much on ‘bottom-up’ reasoning.

That mental manoeuvre was something like:

Figure out which projects need to be done, how long they’ll take, and figure out how many people we’ll need to execute.

If I had thought about that question well, then I could have answered it with the correct answer – and come to the right hiring strategy.

But I didn’t think about it well.

I wasn’t able to think through the details enough to notice everything that would make the work difficult: unforeseen projects, how one’s manager’s small tasks become their big tasks, coordination costs, shocks to capacity like people leaving or going on leave, how much more complicated a 60-person retreat is than a 10-person one, that we would move to a faster growth rate which requires more operations staff, and so on.1

This was made more difficult because my intuition led me astray. Part of me still thought that 80k was an organisation of 10 people, and that part of me thinks that it’s silly that we would need more than one person working on HR. Compound growth do be crazy.

If you’re involved in operations capacity planning, try out this model I made.

You input assumptions about your organisation, whether you’d like to increase or decrease the capacity of your operations team (relative to the org), and it outputs an expected number of hires to make.

It’s only one of several useful lenses on capacity planning, but personally, I find it more useful than the ‘bottom-up’ approach.

Let’s consider two organisations. They each have 100 people, but one growing at 25% and the other is growing at 50%.

Under reasonable assumptions, we should expect that the faster-growth org hires 50% more people, needs to redesign its systems nearly twice as often, and its operations staff are 25% more likely to be inexperienced at the organisation.

These, and other effects, mean that the faster-growth org needs more operations staff.

80k has been hiring at ~50% per year recently, rather than the ~25% per year of our historical average. I think we’ve really felt these effects, and they’ve been larger than my intuition thought they would be! But I think that if I’d actually stared at some of these models for a while, I might have been able to anticipate them better.

I’ve built a tool to help you build your intuitions around the sizes of the effects operating here. I hope it helps!

Organisations often hire people to work on specific responsibilities. In this model, you figure out what responsibilities you need someone to oversee and then evaluate candidates on how well they could perform those responsibilities.

An alternative approach is to hire based on talent profiles. In this model, you look for sets of traits which allow someone to contribute to your organisation. Then you then evaluate candidates on whether they demonstrate those traits.

It’s especially useful if you can find talent profiles which are ‘robust’ — in that someone with that set of traits will nearly always find valuable work to do at the organisation.

Example talent profiles

At 80,000 Hours, we have three particularly robust talent profiles:

• Advisors

  • Capabilities: an understanding of 80k’s views and the interpersonal / analytic skills to navigate a complicated (and ideally life-changing!) conversation.

  • Example responsibilities: (primarily) advising calls, but also running workshops, headhunting for other high-impact organisations, and contributing to our public-facing writing.

• Operations generalists

  • Capabilities: project management, high and reliable output, and flexibility across projects.

  • Example responsibilities: organising a retreat, negotiating an office lease, running a hiring round, scaling up a large book giveaway, fundraising.

• Writer/researchers

  • Capabilities: good writing, careful research, and an understanding of 80k’s views.

  • Example responsibilities: writing career profiles, writing problem profiles, research for the podcast.

I think that Coefficient Giving’s approach to its generalist grantmakers and its business operations staff is also consistent with ‘robust talent profiles’.

Most of 80k’s hires satisfy at least one of the three talent profiles, but only a minority of our staff have the name of their talent profile in their title.

Sometimes, the role only demands the characteristics of the robust talent profile, but we already know what people’s initial responsibilities will be, so we go for the extra specificity.

Sometimes, the role demands someone with the baseline profile as well as some extra requirements.

For example, when hiring recruiters, we’re looking for ‘Operations Generalist profile + recruiting-related skills/experience/qualities’, and we’ll advertise the role as looking for recruiters.

This can also happen when people move from a role which only demanded the baseline profile over to a role which needs extras. For example, Balázs Rapi was a lawyer who started with us as an operations generalist. We didn’t hire him specifically because of his experience in law, consulting, or tech policy, but we expected that they’d be useful. When we became large enough to need more capacity in governance, we moved him to a role that would use that background, and he’s now our Legal & Governance Operations Specialist.

Why define robust talent profiles

Easier workforce planning

If someone has one of these sets of capabilities and is otherwise a good fit for 80k, they’re likely to be able to contribute a lot to our mission.

This is because a) there is a great deal of valuable work for these profiles to do and b) we’re well-practised at figuring out if someone does have one of these profiles. (In this way, 80k satisfies two of the three enablers of ‘Bar hiring’: there’s no steep drop-off in value from additional hires, and we can execute on hiring at the bar well).

This manoeuvre means that we’re not trying to predict when a particular set of responsibilities will come up and designing roles around those particular responsibilities. Instead, we’re thinking, ‘We might need approximately this many people, of approximately this shape’ — which is simpler.

This gives you a lot of flexibility too. For example, we were able to set Luisa Rodriguez’s start date to ‘whenever she finishes work on the book What We Owe the Future’. We were very uncertain about the exact date she’d finish, but that detail didn’t matter much, because we knew that we’re ~always keen to hire more writer/researchers.

You learn how to work with the profile

You can reuse the hiring process — the same work tests and interview templates — which makes it easier for your recruiting team to learn to run hiring rounds (rather than needing the hiring manager to). You reuse the hiring bar: you learn to recognise what good looks like for each profile, and you can apply that from one hiring round to the next. Over time you can create profile-based development rubrics; managers who are managing similar types of people can more quickly develop intuitions about how best to support that kind of person. You even start to learn how people with the profile tend to move and progress over time.

You can hire many people at once

We can advertise for multiple roles which require (for example) the baseline operations generalist profile (e.g. legal specialist, recruiter, business operations, finance associate). We assess everyone as operations generalists (plus some people for legal / finance / recruiting ability), and then make offers to candidates for each of the positions that they meet the bar for. This roughly halves the organisational capacity required to make a given hire.

It appeals to (some) candidates

Lots of people, especially junior people, are looking for their early roles to give them some chance to explore. If they are hired to fit a profile rather than to carry out a specific responsibility, it’s clearer to them that we will approach their development with this in mind.1 This includes hires moving between teams, which has been common for us (for example, Jess Binksmith (operations generalist → Chief of Staff) and Sam Haskell (tech-focused operations generalist → tech systems on headhunting)).

Limitations of talent profiles

It won’t get you all the skill sets you need

You might be able to tack assessments for a given skill set on top of one of the robust talent profiles, but as the skills get rarer and more specialised, this seems increasingly infeasible.

It only appeals to some candidates

The flip side of the ‘exploration’ benefit: people who want a specific, well-defined role can find profile-shaped positions less attractive.

Role discussions are an ongoing task

You're effectively kicking the can down the road on some role decisions, so you’ll need to figure out what comes next for your staff earlier. That's a whole thing. At 80k, we deal with it through tour-of-duty-style commitments, coordination of responsibilities during quarterly planning, and conversations where we try to be open about organisational priorities and people’s role satisfaction.

Hiring by talent profiles isn’t a novel approach! I was first introduced to it by a friend at a startup, and many large-scale organisations have their own versions — for example, the UK civil service’s ‘Fast Stream’, or LVMH’s manager graduate programmes. But it’s a frame that I want to signal-boost among the organisations we write for.

If you have ambitions to scale and you’re spending a lot of time designing and hiring for bespoke roles, consider hiring by talent profiles instead. It might help you hire faster, reduce workforce planning complexity, and give you a supply of people who can turn their hands to whatever the future of the organisation throws at them.

For example, one person had applied for funding to get a master’s degree in machine learning. They needed to commit to the degree within a few weeks, but they hadn’t received confirmation about the funding yet. I was pretty confident that they’d be awarded a grant, but they had little enough runway that the uncertainty might have been enough to stop them going ahead with the degree — and if they did so, it would have been very stressful for them. So I put a promise in writing that I’d send them the money for the master’s if the funding didn’t come through.

I’m glad that I made these offers! It meant that these people were able to go ahead with their projects, they felt less anxious, and — because I thought that they’d probably get funding in the end — I was only spending something like 1/10th of the cost on expectation.

In general, I’ve been quite enthusiastic about making ‘donations’ to unblock someone who was prevented from doing important work due to their financial situation, in cases where I’m in a high-trust relationship with them and we’re able to stomach the social weirdness of it all. Some examples of similar things I’ve done include:

1. Paying for conferences

2. Paying graduate school test fees

3. Equipment for work — where the category is broad enough to include ‘air conditioners’

While these are awfully tax-inefficient for donations, I think that I’ve managed to unblock a lot of valuable work as a result.

I claim that the most obvious question is not the most useful question.

Sure, if you’ve made a clear mistake, then you can correctly make an update and avoid the mistake in future. That’s great. But between you and your manager, you’re probably not making all that many clear (ex ante) mistakes.

So your options are:

1. Making a minor update to your approach. For example, you might slightly upweight a given characteristic in your next hiring round.

2. Making an incorrect update to your approach. This update would have been better in the particular case you’re thinking about, but for future cases it will on average worsen your decision-making.

3. Not updating at all (and just concluding that you got unlucky).

As a result, in a case like this, you don’t actually learn much by thinking about how you could have improved your selection between Option A and Option B. Almost all of the valuable learning lies beyond that.

Instead, I think a better prompt for retrospectives is to imagine yourself as a total badass.

Ask yourself this question instead:

‘What would a dramatically more capable version of me have done?’

Importantly, the person you imagine should be a more competent version of you — not someone you admire with a completely different skillset and personality and without any flaws. This isn’t someone who’s stratospherically out of your league, but the person you’re striving to become over the next few years.

When I take this approach, I tend to notice that there are actually an enormous number of decision points where I could have done something differently, each of which could have caused the situation to go better.

Here’s an incomplete list of ways that a dramatically more capable you might have done better:

Mechanism 1: They saw more options at the final decision point

Instead of framing the hiring decision as ‘make an offer to Candidate A’ or ‘make an offer to Candidate B’, a more capable you might have: offered Candidate A a slightly different role that was better suited to her strengths; done a one-month trial; hired both people; asked the new employee to work on a particular project which they’re a strong fit for; or something else.

Mechanism 2: They set you up to have better options at the final decision point

The stronger version of you would more deeply understand what your team really needed from the new role. They would also have a clearer sense of what types of people were available in the hiring pool and how to appeal to them. As a result, their job ad and strategy for reaching out to people would have been stronger.

Mechanism 3: Their badassery gave them better options by default

Who you are — your skills, judgement, speed, reputation, relationships — shapes which options are actually available.

A more capable version of you would have had stronger candidates than these two trying to work with you.

Mechanism 4: They developed a clear ‘diagnosis and approach’ faster

A novice chess player in a complex situation needs to think ‘manually’.

‘If I move my bishop there, he could take it with the knight... but then I’d take back with my queen, and then he might move his rook to the open file, which would threaten my king… Is my king safe enough?... ’

Similarly, as a junior doctor, when considering a patient’s symptoms, I would have to manually think through all the options before arriving at a diagnosis. Experienced physicians and grandmasters, on the other hand, have such sophisticated mental representations that they’re able to mentally move much faster: they just have a sense that that bishop move is wrong, or this diagnostic pathway is more promising. They don’t need to think through the options as explicitly. Using this skill, experienced doctors can ask the right questions to navigate to the diagnosis, and grandmasters can quickly make strong moves.

In the hiring situation, a more competent you would have understood the job requirements, created a strategy, and considered candidates faster and more instinctively, allowing them to complete the whole process faster.

Unfortunately, for this effect to happen, you need to ‘get the reps in’ - do something, reflect on it, and repeat. Fortunately, for this to happen, you just need to ‘get the reps in’.

(H/t Peak for this mental model.)

Mechanism 5: They built infrastructure that allowed them to move faster

Maybe a consulting arrangement for your prospective hire was a good idea in principle, but in practice it would have taken three weeks for you to set up — figuring out the contract, the rate, the scope, the review timeline. A more experienced you might have drafted template contracts, scoped consulting work, and had rates approved before, so it would have taken them only two days.

Maybe you would have ideally gone to a conference where you could have met a standout candidate. A more capable you might have created a pipeline-focused recruiting team for your organisation, so that person would be in your pipeline already.

Competent people build the infrastructure that makes options cheap enough to deploy that they become real options.

Mechanism 6: They had more and better information at the decision point

For example, their work test better distinguished the top two candidates, or they knew to drill down into something a candidate mentioned in an interview.

How to relate to this emotionally

When I do a retrospective which only focuses on a particular choice point, it feels very unsatisfying. It feels like it’s really hard to make things go better in future — I struggle to draw out lessons which will really improve my performance. This ‘more capable me’ frame, on the other hand, generates interest and curiosity for me.

The orientation is something like: I understand I’m ~never doing anything perfectly, but that’s ok. I’m navigating super complex choice points, doing a pretty good job, and learning along the way about how to do better. Also, there’s just so much to learn, and so many processes that could be better. How exciting!

Among people I’ve worked with closely, the most remarkable example of this is Niel, the CEO of 80,000 Hours.

He understands in his bones that he’s making mistakes all the time, but his instinctive reaction to feedback is rarely defensive.

Each time he learns about a specific way he fell short, he’s curious to understand what happened and how he can improve — and after over a decade of having that attitude, it seems like the guy has become good at approximately everything.

But as I’ve got to know more of these people over the years, both inside and outside EA, I’ve been struck by how badly they suck at some stuff. However, these flaws are not obvious to the casual observer and they’re not tanking these people’s performance, because they use kludges: they patch over their weaknesses with workarounds.

• A friend has a quant PhD from a top school and built a >$100m company. He can’t give feedback on people’s work in person. When he thinks that the work is bad, he is visibly self-conscious and anxious, he doesn’t know how to help the person in front of him, and everyone involved has a bad time. He now only gives feedback asynchronously.

• Another friend has about the best education you could imagine for her field and runs a $10m/year non-profit. She’s a shocking speller, so her PA reads every email before it goes out.

• A researcher I admire struggles to get his thoughts out without immediately shooting them down. When he sits down to write any document, he starts with a disclaimer along the lines of: ‘This might be rubbish, I don’t have to share it with anyone if I don’t want to, these are just hot takes I’m exploring’.

I feel surprised about cases like these. My intuition used to be that badass people would have worked on these weaknesses and resolved the underlying issue, or perhaps just been born without them in the first place. I now think that this intuition is wrong.

When I was failing to work on something useful in the morning, or overthinking an email, or fumbling small talk at family dinners, my intuition told me that this meant I couldn’t be one of those intimidatingly badass people who are able to do important things. These issues just don’t feel like the kind of things that those people would struggle with — so maybe a) I could never get there or b) I would need to somehow generally get to a higher level of Enlightenment and Self-Actualisation such that these things wouldn’t happen to me.

But as it turns out, there’s no need to fix everything: there’s a route to excellence through hacky workarounds.

I think a great way to end is to recognize one’s counterpart for something they genuinely contributed..

For instance:

“[name] it has been so nice speaking to you, and really appreciated your insights about [thing they actually provided insight on that you actually appreciate] — I do unfortunately need to run [on the hour/at the scheduled time].”

Ideally followed by some into-the-future thing you both want (”Let’s keep in touch/curious to see how xyz goes’/looking forward to our next call on [day]”)

Other options

• At the beginning - “So we’ve got 50m today, what shall we focus on?”

• Towards the end:

  • “We’ve got X minutes left”

  • “As we’re coming to the end, …”

If you don’t have a watch/computer/HUD

… so you can’t see the time easily.

Maybe set a timer on your phone for a couple of minutes before the end of the meeting.

If you want to run over

That’s fine, but check explicitly that the other person is also free to do that.

Stragglers will quickly learn that if they want to make the start of the meeting they need to be there by then.