Here’s our update for the last three weeks and plans for the next bout. (See our review of the last bout.)
All figures are approximate.
Cash reserves (end April 2018)
Costs last month (April 2018)
Cash reserves divided by costs last month
Months’ runway (rough estimate based on expansion budget, including commitments to donate, end April 2018)
# of salaried employees (FTE, excluding freelancers)
8 (2.535 FTE)
Morale (on a scale of 1-10, full-time staff average)
Primary growth metric
Change in total impact-adjusted significant plan changes (IASPC)
Monthly growth rate: 0% (same as last update)
Annual growth rates:
Change in total IASPC
Growth in TTM compared to this period
Monthly compound rate
Last 12 months
Monthly growth chart - increase in IASPC per month
Our plan for bout 3 in 2 sentences
Within web & research, we’ll focus on improving the podcast, drafting the advanced series, and updating the website, all aimed at attracting people who might make rated-100 plan changes.
On the in-person team, we plan to start a hiring round, continue our AI policy specialist experiment, and put remaining time into job matchmaking.
-----------------------Full update -----------------------
In the rest of this section, we sketch out our plan for bout 3, 2nd May - 6 June, and into the rest of the year. Then, we outline progress over the last 3 weeks.
Our key focus right now is creating a “machine” that produces rated-100+ plan changes.
In brief, we do this by telling people about our priority paths through online content, then help them narrow down these paths and take action through our in-person advice, and by linking them to the EA community.
Our top priority right now is to refine and increase the efficiency of the machine, before we scale up our staff numbers over the coming years. The aim is to achieve 50-100% growth in IASPC per year while doing this.
How we do it
Find new users
1) From the EA community
2) Word of mouth
3) Online content marketing
1.5 million unique visitors per year
Online content - (a)
Convince of “big ideas”
1) Currently: career guide
Future: advanced series
3) Plus, other EA content.
20,000 new newsletter subs per year who read a lot of the content.
5000+ listeners per podcast.
Online content - (b)
Make case for the priority paths, who and how to enter.
2) Problem profiles & career reviews
1) Narrow down path
2) Enter the path
3) Coordinate within
1) Generalist advising
2) Job matchmaking
3) Specialist advising
40 “tier 1” applicants per month.
Increase motivation, knowledge, coordination.
1) Advisors make one-on-one introductions
2) Promote CEA events
3) Promote online EA community
~10% of EA membership attributable to 80k
Record plan changes
1) Generalist advisors follow up.
2) Online surveys.
100-200 IASPC per month
Main strategic updates
During our April team retreat, we clarified how we set strategy.
In brief, we start by focusing on what will yield the most growth in IASPC over the year. We call this the “YC approach”. However, the downside of this approach is that it can lead you into a local optimum or make you unfocused. So, we also consider the “long-term trajectory” approach. This involves making our best guess at a 10 year vision, and then considering the ideal ordering of projects in order to get there. In other words, the first approach looks forward to the best next step, and the second works backwards from where we’d like to end up.
In the long-term approach, ideally we’d focus on no more than three programmes at each time (e.g. podcast, generalist advising, specialist advising). With each, we’d take them through a full development cycle: (i) work out which programme is best at the margin (ii) make it as efficient as possible (iii) scale up capacity (iv) repeat.
Right now, our biggest uncertainty is about which programmes the in-person team should focus on. These are some of our updates on that question:
We did a third round of job matchmaking, which was slightly below our expectations. However, our overall estimate is that job matchmaking still produces 0.5-2 IASPC per focused hour (and perhaps more, since there’s a lot we can do to improve the process). This makes it look similarly or more effective than generalist advising, albeit with a lot of uncertainty due to the lumpy returns. We also think it’s scalable, plays to our comparative advantage, helps build our network, helpful to our most talented users, and isn’t being done by anyone else in the community. So, we’re keen to continue with it.
We seriously considered setting up a scholarship for graduate study for people who want to work on AI policy. However, after looking into it, our best guess now is that it’s less promising than our core programmes. We also think it makes more sense to first solidify our core programmes, then come back to scholarships as a potential “add-on” later.
Overall, we think the in-person team should focus on advising and job matchmaking, and have both generalist and specialist advisors. The generalists help people broadly narrow down their options and get involved; while the specialists help within a specific priority path. This is because we think we can grow 2 to 10-fold in the next three years by improving and scaling up these programmes. They’re also complementary with each other and our online content; they let us build expertise, connections and a platform for many future programmes; and they aren’t being done by anyone else in the EA community.
How high-impact could the in-person team be?
Based on our past efforts, we estimate each full-time advisor/matchmaker can (when combined with online content) generate 0.3 - 3 IASPC per focused hour, and has about 700 focused hours per year, making for 200-2000 IASPC per year.
We think this is worth $1m - $40m donor dollars, for an opportunity cost of about $1m p.a., making it highly cost-effective.
Over five years, we think we could likely fill the capacity of a team of something like 3 generalists, 4 specialists and 2 matchmakers. We might already have a large enough audience to fill this much capacity.
It’ll be even easier to fill this capacity if we keep growing our audience at our historical rate of 50% per year, which seems achievable. We’d need to reach several million per year and about 20,000 listeners per podcast, which would make us only 1-5% as big as Vox, Econtalk, Tim Ferriss, Sam Harris etc., which seems achievable.
This total team of 9 would therefore produce 2000 - 20,000 IASPC per year, or $5m - $400m donor dollars.
It seems plausible we could then grow the team further to over 20.
How soon to hire?
We decided to make hiring 1-3 people to the in-person team a top priority, whereas we weren’t focused on hiring before. This is because we don’t currently have enough capacity to both maintain a reasonable baseline of advising, while also improving the programmes. It’ll also help our annual growth targets.
Otherwise, hiring full-time staff is mostly not a priority, since we think we can grow more quickly through other means (unless we find someone really exceptional and already trusted).
The podcast continues to go well, so we plan to continue with it. We’ve been able to launch one per week, each adds 100-400 subscribers, and old episodes keep getting listens. We think it covers useful content and we get lots of positive feedback about it.
Plan for bout 3 and beyond
Over the next six months, we’ll mostly work on improving the same programmes we did earlier in 2018: the podcast, the website & advanced series, job matchmaking, specialist advising.
Here are our main projects for bout 3:
Web & research
Keep releasing and improving the podcasts
Aim to focus more on answering key questions for potential-100s (rather than growing the audience).
Plan a website update so that it better appeals to potential-100s+.
Keep drafting the advanced series.
The aim is 20h of focused work, which might be enough to finish 1-2 articles.
Run a hiring round, aiming to hire 1-3 generalists and/or job matchmakers, plus any exceptional specialists that come up. This will likely spillover into the next bout, and might not launch until then.
Continue with our AI policy specialist experiment.
In remaining time, continue to do and improve our job matchmaking process. (Or switch to generalist advising depending on who we hire.)
These are the targets we set for the bout, and how we stand two weeks in.
New content releases running total
(target 1.4/week, cumulative)
New podcast subscribers
(target 250/week, stretch 350/week, cumulative)
# coachees with high (6-7 rating) NPS rating
(target 2/week, cumulative)
We also advised 27 people in the last three weeks.
Content released in the last four weeks:
Podcast: Bryan Caplan on education and signalling - this one had especially good feedback and was featured on marginal revolution.
Podcast: Eva Vivalt on the generalisability of RCTs in development - this has some especially interesting new content on the value of being evidence-based.
A list of promising think tanks to work at.
Other key progress in the last 3 weeks
Prepared a mid-year update for donors and met Open Phil for check-in.
Investigated the value of hiring a biorisk specialist advisor.
Completed most of the migration to Gsuite.
Met with several potential hires and offered one a trial.
Pushed forward negotiations with two offices.
Updated our information sharing policies.
Have a great week,