Suppose you are a technical AI safety researcher who has done some research, but has not yet landed a full-time job. In this post, I argue that taking a contract research manager position (~3 months) on one of the AI safety fellowships could be more valuable for your research career than spending these 3 months on your own research.[1]
Background on research management (RM): to get a sense of what a typical RM role involves, see section 1.3 “Overview of Responsibilities” of Guide to Research Management at Pivotal.
My background: I worked as an RM for 6 research groups on the MARS program Winter 2025-26, currently a researcher on PIBBSS fellowship, considering a full transition to a research manager role.
I will consider two options:
Research manager. You work as an RM on one of the AI safety fellowships for 5 research groups, each consisting of one mentor and 3-4 mentees.
Researcher. You participate in an AI safety fellowship as a mentee in one of the research groups.
The goal of the post is NOT to make a fair comparison, but to highlight some advantages of RM option which are often overlooked. So for the Researcher option, I write shorter and only the main points, thus partially strawmanning it. Also note that the Researcher option is “mentee on AI safety fellowship”, and I do not consider other options like independent research, internship, etc. I also do not cover in depth the main downsides of RM option, which are: you don't advance your own research agenda, you may lose momentum on ongoing projects, and managing others can be emotionally taxing.
Below I list the advantages of each option in three categories: upskilling, career capital, and networking.
Upskilling
RM
You gain skills that are needed to organize and guide your own research group in the future – setting priorities, workload distribution, supporting team morale, scheduling, publishing results.
You will see how 5 research groups work: the different mentoring/research styles, different team dynamics, failure and success modes, common mistakes, etc. Much of this is tacit knowledge, which you can get only by participating in many projects. As a researcher, participating in 5 projects would take you 1+ year, but as an RM, you get this experience in 3 months.
Researcher
You get technical experience, and you develop a deeper understanding of the research area of your current project.
Career capital
RM
RM is a strong line in the CV for application to any senior researcher / team lead / management position, or for asking for grants to start your own research group or research organization.
Your line in the CV could be something like “managed 5 teams, which resulted in three publications accepted to ICML – one main venue, two workshop papers”. Even for researcher positions, this line has some weight.
Researcher
You can get a publication as a first author on the top ML conference. If you have no first-author publications yet, getting one should probably come first — the RM option makes most sense once you have at least one. But if you already have some first-author publications on top ML conferences, then maybe one more such paper adds less marginal value than the RM experience.
Networking
RM
You work closely with 5 mentors, who are already working in different AI safety organizations. You also work with 15-20 mentees, some of whom will go on to work at AI safety organizations. If you ever apply to any of these orgs, you have a warm introduction. You can also get a recommendation letter from mentors for applications to fellowships or jobs.
Researcher
AI safety fellowships often work as internships for AI safety organizations. That is, if you perform well, there is a chance that your mentor will hire you to the organization where this mentor works.
Impact
According to Ryan Kidd, Co-Executive Director at MATS, the AI safety field is bottlenecked on research managers. This suggests that the RM option likely has a higher counterfactual impact than the Researcher option, but of course, it depends on how replaceable you are in each role.
Why I wrote this post
I want to make the case that a contract RM position can actually speed up your research career. If a few researchers try a contract RM position as a result of this post, some will produce direct value for the field, and some may decide to fully transition to RM, which the field needs.
Full career transition from researcher to research manager is also a valid option, but in this post, I assume you return to a researcher position after RM contract.
Suppose you are a technical AI safety researcher who has done some research, but has not yet landed a full-time job. In this post, I argue that taking a contract research manager position (~3 months) on one of the AI safety fellowships could be more valuable for your research career than spending these 3 months on your own research.[1]
Background on research management (RM): to get a sense of what a typical RM role involves, see section 1.3 “Overview of Responsibilities” of Guide to Research Management at Pivotal.
My background: I worked as an RM for 6 research groups on the MARS program Winter 2025-26, currently a researcher on PIBBSS fellowship, considering a full transition to a research manager role.
I will consider two options:
The goal of the post is NOT to make a fair comparison, but to highlight some advantages of RM option which are often overlooked. So for the Researcher option, I write shorter and only the main points, thus partially strawmanning it. Also note that the Researcher option is “mentee on AI safety fellowship”, and I do not consider other options like independent research, internship, etc. I also do not cover in depth the main downsides of RM option, which are: you don't advance your own research agenda, you may lose momentum on ongoing projects, and managing others can be emotionally taxing.
Below I list the advantages of each option in three categories: upskilling, career capital, and networking.
Upskilling
RM
Researcher
Career capital
RM
Researcher
Networking
RM
Researcher
Impact
According to Ryan Kidd, Co-Executive Director at MATS, the AI safety field is bottlenecked on research managers. This suggests that the RM option likely has a higher counterfactual impact than the Researcher option, but of course, it depends on how replaceable you are in each role.
Why I wrote this post
I want to make the case that a contract RM position can actually speed up your research career. If a few researchers try a contract RM position as a result of this post, some will produce direct value for the field, and some may decide to fully transition to RM, which the field needs.
Full career transition from researcher to research manager is also a valid option, but in this post, I assume you return to a researcher position after RM contract.