If you’re planning the kind of pivot into AI policy that Jarom describes, applications are open through June 21 for Horizon’s Career Accelerator. Our part-time, remote program is designed to help people launch AI policy careers at any stage, offering tailored support, including 1-on-1 mentorship, application advice, opportunity recommendations, and career development funding of up to $100,000.

Imagining your deathbed is a good way to learn what truly matters. Like anyone working in DC, I expect my deathbed mostly to feature fond memories and searing regrets related to my semester-long congressional internship: the coffee chats taken, the coffee chats missed.

This post exists so that you will not experience those same regrets. If you follow this advice, I believe that not only will you create wonderful memories of DC, but you will also GET A JOB IN POLICY!! As I’ve learned from my deathbed reflections, this is the true meaning of life.

All you need to do is (1) get clear on your goals, (2) talk to the right people, and (3) optimize your work for showing off.

What do you even want?

DC interns typically have one of the following responses to this question:

• “A job, duh.”

• “I don’t know what I want; the whole point of this internship is to test my fit for policy.”

• “I will stop at nothing to become President of the United States.”

These are all fine answers, but compare them to:

• “I want to know relevant people at XYZ think tanks and lay claim to at least one solid writing sample I can use to apply to research positions.”

• “By the end of my internship, I want to have strong enough connections on the Hill that I can get flagged for staff assistant openings in offices in my home state.”

Unlike the first set, these more specific goals help you decide which projects to volunteer for and which people to email for coffee.

They also make it easier for other people to help you: if you say “I want to stay on the Hill, and I’m interested in press assistant roles” instead of vaguely mumbling about wanting to work in policy, someone can flag you for an opening or connect you to a relevant contact.

It’s worth scheduling an hour with yourself early in your internship just to think about these goals: what kinds of opportunities do you hope to pursue once the summer has run its course? Some things to keep in mind as you develop and refine your hypothesis:

• Work backwards from longer-term goals: if you have an idea of where you’d like to end up in a few years, try to determine and target the most common paths that get people there.

• The best way to understand these paths1 is to talk to people a few steps ahead of you, the ones currently occupying the kinds of roles you hope to fill in 2-3 years. They’ll often have the most up-to-date and relevant advice.

• Pay attention to what you enjoy. If you find yourself energized by legislative research and banging your head on the table whenever you’re assigned comms work, that’s useful data. Your internship is an experiment! (But please don’t bang your head on the table.)

• Revisit your goals periodically. The person you are at the end of your internship will have much better judgment about this stuff than the person you are at the beginning.

Despite the usefulness of having goals, hold them loosely. Policy careers are winding and unpredictable, and you probably don’t have enough information to be very confident about what you want. Even if you’ve tattooed yourself with the words “Brookings or bust,” you should still be nice to people from the Department of Education.

Network better

The math

To make it in this town, you need to think the way DC people think. DC children learn the concept of a microhire2 in kindergarten.

One microhire = a one-in-a-million chance of eventually getting a job. Every professional interaction during your internship is worth some number of microhires.

Here is a scientific chart to illustrate the concept:

A one-on-one “coffee chat” with someone you’ve specifically chosen (because they work at an organization you’re targeting or have the kind of career you want) might be worth 8,000 microhires. They’ll remember your name, they can flag you for relevant positions, and they’ll probably give you better advice than whoever you happen to sit next to at a happy hour.

Meanwhile, the typical interaction at a large, open networking event is worth maybe 600 microhires. The above-average ones result in pleasant small talk and a crumpled business card to add to your increasingly cluttered purse. You’d need 14 of those interactions to match one coffee.

This doesn’t mean you should never go to events: events can be extremely valuable insofar as you use them to set up coffee chats. You’ll be more successful emailing someone to set up a conversation if you can reference a shared context where you already met. Just remember that the core of your networking strategy should be targeted one-on-ones.

Where to find people

The lowest-hanging fruit is in your own office. You’re already working alongside people who know the policy world, and they’re already expecting you to ask them to chat. The comms director might have a friend at the think tank you’re eyeing, and the legislative director might know that an allied office is about to have an opening. It’s good to set up these conversations early, so the people you work with have time to figure out how to help you, and a reason to.

The most important question to ask in any of your coffee chats may just be, “Is there anyone else you’d recommend I talk to [about finding think tank jobs, learning about appropriations, or whatever it was you just talked about]?”.

Most people will name at least one person, and one coffee begets another until you’ve created an endless pyramid scheme of coffee chats. If someone in your network offers to connect you with someone, follow up on it!

After the connections you make from the connections you already have, anyone from (1) your alumni network, (2) your state, or (3) your chosen policy areas is fair game for networking. Be sure to bring up the connection and why you’d like to hear their particular perspective, and your cold email is fairly likely to work.

And when you’re approached for help (which may happen faster than you anticipate, especially if you’re successful at networking!), remember to pay it back.

The art of the coffee chat

Do your homework before a one-on-one. It can feel like a waste of time to be asked basic questions that could have been answered online or by perusing your interlocutor’s LinkedIn.

If you prepare a short list of questions that you’re most curious about and a honed pitch about your background and goals, you’ll stand a significantly better chance of having a productive conversation and a marginally better chance of uttering the “secret networking question,” a password that activates the Washington Monument’s functioning rocket boosters.3

If you’re early in your career and feeling weird about asking someone more senior to grab coffee, keep in mind that most people are happier to do this than you think. There’s a culture of paying it forward in DC, and taking 20 minutes to give advice to an eager intern is one of the more pleasant parts of anyone’s workweek. I have never encountered anyone who considers chatting with interns to be an imposition.

Despite the name, coffee chats do not need to involve coffee.4

Build a portfolio

Make your office like you

Your supervisor may claim they don’t have favorites, which is a silly lie to tell.

While it’s not strictly necessary to become the favorite intern, this is a good aspiration for a couple of reasons:

1. You’ll often be assigned more substantive work and receive more freedom to steer your own projects.

2. You’ll become the first person your office flags for job openings, or the first they’d consider for an opening of their own.

Being this kind of overachiever requires learning quickly from (sometimes implicit) feedback, anticipating the needs of your supervisor and proactively volunteering for work, then doing that work on time and to a high standard. The occasional thank-you note also helps. This is not especially novel advice.

More important than being the favorite intern is that you don’t become the problem intern.

Being on bad terms with your office makes it hard to get a job, for the mirror image of the reasons being the favorite helps: if you can’t be trusted with the mail, you won’t be trusted with anything closer to policy. A strong reference from someone who’s watched you work can go a long way, and the absence of one will often be taken as a red flag.

The bar for avoiding your supervisor’s ire is lower and basically entails showing up to work on time, being polite, asking for feedback at least occasionally, and refraining from late or sloppy work. Thankfully, I expect all our beloved readers would far exceed this standard.

Produce targeted work

Depending on how much flexibility you have to direct your work (which will be higher if you’ve built up goodwill with your managers!), you might be wondering which kinds of projects to ask for, or which deserve the most effort and care.

This is where your hypothesis from earlier pays off. Many policy jobs will ask for a writing sample, and ideally yours should look like the kind of writing you’d be doing in the role you’re applying for. It’s worth spending significant time, blood, sweat, tears, etc. on 1-2 really good pieces of writing you can show off when applying to jobs.

If you want to work at a think tank, a policy brief or blog post is often the price of admission. If you’re aiming for the Hill, a strong constituent letter will be more useful than a 10-page research paper. Policy memos are the closest thing to a universal currency in policy: nearly every employer wants to see that you can take a complex issue, lay it out clearly, and present options. If you don’t get a chance to write one at work5, it could be worth writing one on your own time.

Hiring managers also want to see that you’ve been exposed to the full range of what the job entails. This means it’s worth volunteering for different types of work even if you only do them a couple times. If you’re in Congress and you’ve never helped with flag requests or constituent casework (common components of entry-level staff assistant roles), just tell your supervisor you’re interested in trying these out. The return on doing something once (and being able to put it on your resume) is much higher than the return on batching emails for the hundredth time.

All of this work is only useful if you can produce it when the time comes, so be sure to document everything. Keep a running list of projects, save copies of your best work, and write down what you contributed while it’s still fresh. At the end of my internship, I forwarded useful samples of work to my personal email6, and this made updating my resume for different job applications much less painful than it otherwise might have been.

Get your materials reviewed

By the end of your internship, everyone will be clamoring to see your resume. They’ve heard the rumors of the Chosen Intern. They want to know that what they’ve heard about your astounding depth and breadth of experience is true.

Before that time comes, get at least two people to tear apart your resume and cover letter, and ideally someone who actually hires in the area where you want to work. A resume that’s been reviewed by a policy hiring manager will predictably look quite different from one that’s been reviewed by a college career counselor who mostly works with engineering students. People in your office are usually very happy to do this.7

Stop and smell the cherry blossoms

Though the point of your short time on Earth is to get a policy job, happiness is nice too. DC is a cool town, and if you haven’t had a lot of fun by the time your internship is over, I will feel like a failure and you will have dishonored the founding fathers. Some things to do include:

• Eating good food: DC has world-class Ethiopian restaurants (I enjoy Chercher). It’s also the world capital of slop bowls. Tyler Cowen has some good recommendations.

• Going outside: Rock Creek Park is enormous, making it a nice escape from the cable news channel which will be playing at all times in your office. Meridian Hill Park is my go-to outdoor spot, partly because it is one of the few parks in DC with correctly spaced trees for hammocking.

• Being patriotic: the Fourth of July is sure to go crazy this year (America’s 250th anniversary).

• Being a tourist: It goes without saying that you should visit the Smithsonians, the top of the Washington Monument, the memorials, etc. They are excellent experiences.

• Wandering around: DC has many a cute neighborhood, and you’re sure to find interesting things by meandering about in Adams Morgan, U Street, Georgetown, Eastern Market, Mt. Pleasant, etc.

• Following your curiosity: If seeing a certain Senator would make you swoon, visit their committee’s hearings! If you’re really interested in a think tank’s work, subscribe to their newsletter and go to their events. Some of the best connections and most engaging conversations come from being at the places that interest you most.

Parting wisdom

Congratulations; by securing a coveted DC internship you’ve proven yourself to be very cool and lucky. You are getting paid (hopefully) to learn how the most powerful government in the world works from the inside.

If you’re interested in sticking around, your best bets are to be specific about what you want, strategic about who you talk to, and deliberate about where you sink your effort. If you have read this far, you’re entitled to connect with me and ask for a coffee chat, and I am bound by the rules of DC to say yes. I wish you the best of luck.

More from Horizon

• Our guide to policy internships

  • And congressional internships, specifically

• Our list of short-term policy programs

• The Career Accelerator: a part-time, remote program designed to help people launch AI policy careers at any stage, offering tailored support, including 1-on-1 mentorship, application advice, opportunity recommendations, and career development funding of up to $100,000. Apply here by June 21.

We were excited to get his takes on the questions we get most from people thinking of making the pivot.

Horizon’s Remco Zwetsloot talked with Neil about:

• How the engineering mindset helps and hinders in policy work

• How quickly you can become an expert (and the case for specializing early)

• Why conservatives and free-market technologists should work in AI policy

• Why you may not need formal credentials as much anymore, but also how they can still help

• How to red-team policy ideas with AI

Neil is the author of Getting Out of Control: Emergent Leadership in a Complex World, and his Substack is Getting Out of Control.

Neil will also be a featured speaker at this August’s AI Innovation and Security Policy Workshop, a three-day DC event hosted by Horizon and the Foundation for American Innovation for people exploring careers in AI policy. Learn more and apply here by June 21.

Remco: Your job title is head of AI policy, and you’ve joked that AI policy can refer to anything from forecasting the future of human civilization to figuring out how municipal governments use software. What’s it like to work on a portfolio that broad?

Neil: It’s challenging. It’s a lot of triage. Part of the reason it’s so broad is that the definition of AI itself is pretty nebulous—people mean very different things when they talk about it, and what it means has changed over time since the term was coined in the 1950s.

How do you pivot from a question about federal preemption of states to interpreting how municipal privacy law applies to using AI? That’s a really wide range—both in topic and in what institutions you’re interacting with and what the potential outcomes might be.

It means I have to bring a consistent set of principles that can apply across that scope. It’s more of an art than a science, and it takes a lot of adaptability.

A lot of people ask us at Horizon whether they should specialize in a narrow area of AI policy or stay broad. What are the main pros and cons they should weigh?

In most areas, I might advise someone to start general to understand what is important and then dig in. But in AI, it’s actually fine to go really deep on a specific area early.

The technology is changing so fast and encountering so many different domains that you’re going to have to become knowledgeable across lots of areas over time anyway. Expertise in one area may not inherently transfer, but going deep early can show people that you can do serious work, and it differentiates you. There are tons of people who have broad opinions about everything AI but haven’t dug deep on anything specific. And because this is a general purpose technology, there are tons of areas where it would be really useful for somebody to have done that deep work.

If you find a space where nobody else has done the work, whatever you do is better than what existed before. That’s a real opportunity for young people especially.

So my short advice: find an area you’re really interested in and dig deep. In the AI space, that’s almost always going to be useful. It’s not a dead end in any way.

You’ve written that the engineers who self-select into policy work often think law and society are as deterministic as computer code—and that as a result, policy wonks with deep technical backgrounds can be some of the best-intentioned advocates of truly terrible ideas. At Horizon we work to bring people with tech knowledge into policy, but your argument is that this isn’t per se good, which is very fair. Can you elaborate on the risk you see here?

I should say up front—I think Horizon helps mitigate that problem. Here’s the problem: The barriers for engineers to go into policy were pretty high, as were the opportunity costs. So there is a selection effect: engineers who were optimistic (or neutral) about technology’s effects on society went into the private sector, while the pessimists went into policy. Add to that that the engineering mindset isn’t a natural fit with the messy, complex emergent system nature of lawmaking. Together this often meant engineer-informed policy ideas would assume legal mechanisms are deterministic, like code or machines: self-enforcing or self-executing, neglecting that other people are involved.

Of course, if you spend time in the policy space or get support from a place like Horizon, you know that the law you pass is not quite the same as how the person enforcing it views it, or how the court reviewing it interprets it, or how the compliance officer at a company tries to implement it. Without that mental model, a lot of the policy ideas generated by people with technical backgrounds don’t account for unintended consequences.

The paradigmatic example is the net neutrality fights. It started as a fairly academic idea—not even that polarized. Most people generally agreed that users should be able to select what content they view without broadband providers acting as gatekeepers. But the policy tools that were ultimately used were so broad that by the end, we were talking about reclassifying all internet service under a regulatory regime intended for telephones. It spiraled out of control despite the best intentions.

I see this happening again in the AI space. People who are new to policymaking from the AI community propose state-level laws and when you push on them—”Well, the state attorney general might interpret what AI means somewhat differently than how you defined it”—they say, “Why wouldn’t they just interpret it the way we all know it means?” And it’s like, because they have different motives than you.

I was briefly at OpenAI back in 2018—one of a very small number of policy people there at the time. I had lunch with a group of engineers and scientists, and one of them asked, “This policy stuff—why does everyone act like it’s so complicated?” I gave what I thought was a simple example from a military context: what if you have an autonomous weapon and it misidentifies a target and attacks it? And this person said, “Well, couldn’t the U.S. and China just write a contract agreeing that if you sink an aircraft carrier, you just pay $5 billion and move on? It’s an accident.” My mind was blown. I didn’t really know where to start.

Exactly.

For people who want to bridge that gap—who want to understand both the technical and the policy side—what combination of skills and mindsets do they actually need?

Neil Chilson: It’s a translation ability. Having the technical expertise, but also understanding what the policy problem is. If you have a technical challenge, the solutions involve research. But if you have a policy or political challenge, the solutions involve building coalitions and connecting to what people need and want. That requires the ability to translate what the technical origin of the problem is into the things that will actually get people to adopt solutions.

Without that connective tissue, you get solutions that are politically satisfying but completely technically problematic. Or you get technologists trying to explain something that’s totally disconnected from the harms people are seeing in the real world or from the political incentives at play. Both sides can feel like they’re getting what they need while the solution doesn’t actually work.

I don’t know exactly how to describe that skill set other than translation. Some of it is learning the vocabulary. A lot of it is understanding the people. We don’t have a great curriculum to teach it—some of it is just experience. Getting ground down through a couple of debates and understanding that it’s not just about winning an argument or even about communicating your idea clearly.

There are people who think that if they just explain their idea well enough, the problem is solved—that others will agree with them. And in some ways, explaining clearly is part of convincing people. But understanding what the other person wants is another really big part.

It’s not enough to assume they’re truth-seeking in the same way you are. I think we’re seeing this right now with some of the public discourse around AI companies and government—well-intentioned public pronouncements that aren’t necessarily helping solve the underlying problem. Sometimes you need to spend more time building trust.

Your personal background spans computer science and software before you pivoted to law school. Has that dual background shaped how you approach policy questions?

Neil Chilson: It has, for sure. Those credentials opened doors for me because it was an unusual combination, especially when I was starting my policy career.

But my policy approach actually traces back even earlier. My motivation for going into computer science was that I got really interested in complex systems theory as a teenager because of James Gleick’s book Chaos. It was a pop science book, but it had all these cool pictures of complex, recursive systems, and I internalized the idea that very small changes can make really big differences and that prediction in complex systems is very difficult.

I didn’t think much about that during my undergrad or grad training in computer science. But in the policy space, it became very clear that this is a complex system where the levers you can pull are unpredictable in their effects. That ingrained in me an inherent skepticism about the strong version of public policy—the idea that you can impose your will on the world through regulation without side effects you can’t anticipate. So even though it’s entwined with my computer science background, it’s actually earlier than my formal education.

We get lots of questions about what credentials are needed to enter AI policy. Has your advice here shifted over time? Someone like Thomas Hochman at FAI came into AI and energy policy right after college and has had a big influence at a young age. Do you think the field has de-credentialed?

Neil Chilson: For the public conversation, it definitely has. The tools of communication are so much broader and more accessible that there’s a much wider aperture for people to contribute as individuals without credentials.

But once that conversation starts to turn into policy action, credentials and experience still help you get in the door. You can be an expert in a space and still not know which committee in Congress to talk to, who the real decision-makers are, or what a good meeting with a staffer looks like. You can blow a lot of the credibility you built through your public work by appearing not to take their job or their time seriously.

So credentialing and experience are still a filter for reaching policymakers, though that may be declining a little. Staff do pay attention to the public conversation, and sometimes they’ll invite people in. At that point, if I were someone who’d never been in a policymaker’s office, I’d find somebody who could give me good advice about how to make the most of my time there.

You’ve spent a lot of your career in right-of-center offices and organizations. We’ve talked before about how there’s a thinner bench of tech policy talent on the right, and the problems that can create in AI policy conversations. Why do you think this is the case?

Neil Chilson: Historically, if you care about technology and you generally believe that markets function well—that there aren’t big problematic societal impacts of the technology, or if there are, there will be market solutions—you’re probably out building things in the market. (This is the selection effect I mentioned earlier.) That tends to be a right-of-center policy framework. So right-of-center talent went into the private sector.

Meanwhile, the left has invested a lot more in training tech policy talent at the university level and incorporating it into congressional offices and institutions. That’s the selection effect.

A lot has changed, though. It’s not obvious anymore that the right is as market-friendly as it historically was, or that it’s less interested in policy intervention. And because tech policy has become such a big component of all policy—with AI cutting across every sector—I think there’s a bipartisan interest in having people with this expertise. The pipelines from academia are still much more robust on the left, but there’s much more demand on the right than there used to be.

We see a lot of right-of-center technologists at Horizon who are curious about policy but unsure whether to make the jump. What’s your pitch to someone on the right who’s building things in industry to consider policy work, even for a couple of years?

Neil Chilson: It’s actually a bit of a hard pitch. I want talented people building technology and businesses because I think that’s the best way to make the world better. But the pitch I’d make to someone who’s already interested in the policy track is: your technical background and interest in this space are in high demand. There are already very loud voices with different opinions than you talking about these issues. Your voice is undersupplied. In a democratic process, we get better results when viewpoints are tested against opposing viewpoints. And you can always go back and build something later.

Another worry we hear a lot is that people feel like they’ve missed the window, that AI policy is already a crowded field and the good seats are taken. What do you tell them?

Neil Chilson: Because AI changes so fast, nobody’s an expert for long. There are so many issues and the technology moves so quickly that someone with good general critical thinking skills and policy instincts could become a top expert on some AI policy issue pretty quickly—because there are just so many issues and nobody’s had time to stake a claim. It’s a better time than ever to jump in. The barriers to entry are pretty low, and you can make a contribution quickly.

Are there particular programs or resources you’d point people to, especially on the right?

Neil Chilson: You can go to abundance.institute—we have a lot that talks about our angle on energy and AI, and we do work at both the state and federal level. Beyond just policy, we’re also trying to shape the cultural conversation around these topics.

There are lots of places that pump out white papers. There aren’t that many that are trying to directly build talent, especially on the right. I’d point to FAI, Mercatus Center, and Tyler Cowen’s Emergent Ventures program. Not as right-coded, but very capable are places like Roots of Progress Institute and the Cosmos Institute. Those are places that are very interested in young talent, helping people upskill, and sharing ideas and mentoring. If I were a young tech policy person on the right, I’d be consuming their content and reaching out to those folks on every platform they’re on.

Last question. A lot of people coming from tech backgrounds increasingly have opportunities not just to think about what AI policy should be, but to apply AI tools to solving policy problems directly. You’ve been doing some great work on that—what does that look like?

Neil Chilson: It’s amazing. The AI tools now make it so anyone can become a data scientist and build a cool dashboard. I saw just this morning that a West Virginia state delegate vibe-coded a site showing the per-county, per-district benefits of having data centers in West Virginia. He’s not a coder.

We built a website, redteam.law, where you can upload any piece of AI legislation and it will identify the stakeholders in that policy, then create red-team personas—the attorney general who wants to someday run for president, the corporate executive trying to maximize quarterly profits—and analyze how each one might use or abuse the law. Then it writes a report. I’ve found it especially useful for people who haven’t engaged in policy before, because it flips that switch: “Oh, there will be people applying this law who have very different motivations from the people who wrote it.”

We’ve done lots of other experiments too. Florida had a bill that would have banned data centers within five miles of schools or residential areas. I quickly had Claude generate a map of Florida showing all the areas that wouldn’t be affected by that restriction—and there were none. The ban covered all of Florida. You could see it visually, instantly. That ability to do rapid data analysis is very compelling.

I honestly think think tanks are going to look very different in a decade, and I’m personally still trying to figure out what that looks like. The tech changes so fast that you can start implementing a new process and then realize there’s a new plugin or feature that just does it all.

It’s a really exciting time to be in the ideas space—but it also means your judgment about how to spend your time matters more than ever.

Big thanks to Neil for taking the time. If you enjoyed this conversation and want to learn more:

• Neil will be speaking at Horizon and FAI’s AI Innovation and Security Policy Workshop this August. Apply here by June 21.

• Neil’s Substack: Getting out of Control

• Abundance Institute

• Redteam.law, Neil’s AI legislative redteaming site

• Horizon’s guide to think tank work on emergingtechpolicy.org

Launchpad is a resource from the Horizon Institute for Public Service, a nonprofit, nonpartisan organization working to address the U.S. government’s critical talent shortage in emerging technologies. Learn more about us and our programs supporting tech policy careers here, and explore our in-depth policy career resources at emergingtechpolicy.org.

ChinaTalk

Doing Big Things in Policy

Want to do big things? Today we’re providing a guide of sorts. Joining me is Remco Zwetsloot of the Horizon Institute for Public Service and Kumar Garg of Renaissance Philanthropy…

Read more

19 days ago · 31 likes · Jordan Schneider and Phoebe Chow

You’re probably underestimating how much you could contribute

Kumar described how people want to look for “white space” in policy—areas where no one is working. He says:

Kumar Garg: One of the conversations I have with donors goes like this—somebody might say, “Let’s do a white space analysis. Where’s the white space?” By that they mean there’s some space where everybody’s working, and another place where no one is working. The sad joke is it’s all white space. You get into these problems, and as you dig in, you very quickly figure out there’s a bunch of stuff that’s quite important and not getting worked on.

We discussed a lot of reasons this happens. One is that there are too few people setting goals for themselves and truly making it their job to hit them. (I also recommend the already-classic post by Nan Ranshohoff on why more people should be a “general manager” for a big problem.)

Remco Zwetsloot: They come in saying, “I want to be a public servant. I’m here to do good. Other people might tell me what that is. I expect to come into the office, be assigned a thing, and do it.” Often they come into a space where there’s no clear agenda. You can spend two years in DC just responding to incoming, doing a thing here and a thing there. At the end, maybe you’ve contributed, but you haven’t really changed anything. For that person, you have to push much harder on: what is the thing you want to be different two years from now?

Kumar Garg: You have to be obsessed with winning—with thinking, “This is really important and I really want it to happen.” A lot of times, people in government fall into this idea of “I own this portfolio.” I don’t like the word “portfolio.” A portfolio is a fancy way of saying this is the range of topics whatever seat I’m in has equities in. It’s better to have goals—“I want to move from here to here.”

Another big reason is that, while humility is incredibly important for policy work, it can sometimes paralyze people into thinking they’re not ready to make the jump and won’t be for a long time.

Remco Zwetsloot: There’s a certain type of person, especially folks coming from academia, who think, “I really want to work in policy and public service. I want to contribute, but I need to understand my area just a little bit better before I make the jump. I need to know the full answer to what should happen with China policy before I go and try to get a job where my task is to say what China policy should be.”

My usual response to this is that, from the outside, it can be very hard to know what you need to learn. I explained this using one of my favorite 2x2s (you can take the guy out of the political science PhD but you can’t take the political science PhD out of the guy):

Remco Zwetsloot: On one axis, you have unconscious versus conscious, and on the other, incompetence versus competence. Most people start out unconsciously incompetent. The first part of the learning journey is becoming consciously incompetent. Then you become consciously competent. The journey culminates in being unconsciously competent.

People really neglect the importance of being consciously incompetent. A lot of experts don’t know all the different things they need to know to have the solution for AI policy, for example. It’s just too complicated. If you’re not in DC yet, you don’t know all the ways you need to think about it. But you can know enough about AI to very quickly know what knowledge gaps you need to fill to say something about open-source versus closed-source AI models, or what China is doing in AI.

One of my colleagues, who was a fellow on the Hill, had a nice saying: “Your job is not to be the expert. Your job is to mobilize expertise.” That is your job as a staffer. To do that well, you need to be consciously incompetent—humble enough to know where your gaps are, then entrepreneurial enough to fill them. That’s a really neglected skill set. It lowers the bar for where someone needs to be to make a contribution in DC. Lower your standards. You probably can contribute so much more than you think just by being aware of the gaps and leaning in early.

The need is bigger than you think

The vast white space in policy means a lot of important ideas no one has explored.

Kumar Garg: We still live in a real deficit of clean ideas. I always used to say, when we were sitting around trying to come up with State of the Union ideas, “Why don’t I have a book from each think tank that says, ‘Here’s everything we wrote in the last year, formulated as a State of the Union idea. Here’s the sentence the president would say. Here’s the logic model of the policy proposal. Here’s a link to all the appendices so you can make it bigger or smaller. Here are the phone numbers of the experts you’d call.’” Instead, I’d be hunting around, “Has anyone written on this?”

So more than ever, the world needs ambitious, humble, thoughtful people excited to tackle big problems with concrete ideas.

Kumar Garg: We’re always talent-blocked. We’re bottlenecked on talent on basically everything.

A recent example—in the past five years, there’s been a huge increase in the number of people who realize lead pollution is a really big deal. Maybe a quarter to a third of the global learning gap between rich countries and poor countries can be explained by lead pollution. When I started talking about this five or seven years ago, I’d get a nodding head—“Yeah, pollution’s a problem.”

How many people work on lead remediation globally full-time? Maybe 100. We’re talking about something that might have a trillion-dollar-plus lifetime impact. We underestimate how many really important things don’t have enough talented people working on them.

Horizon exists to build government capacity in emerging tech—one of many efforts trying to close the gaps Kumar describes.

Remco Zwetsloot: If listeners take one thing away from this conversation, I hope it’s Kumar’s earlier message—this is fundamentally talent-constrained work. I could not name you a problem where I don’t think part of the solution is “many, many more people should work on it.” There are complicated “how” questions depending on your personality and personal constraints. But I can guarantee that for someone trying to do good and thinking about science, technology, and China-related topics, there’s so much impact you can have. People in jobs that don’t feel aligned with their ultimate mission in life should think strongly about how to make the pivot in the next couple of years. A lot is changing in the world, and there’s so much need.

Big thanks to Jordan and Kumar for a great conversation—I hope it inspires people to take concrete steps in pursuit of the change they want to see. Some more great resources on breaking into policy and making an impact:

• Renaissance Philanthropy’s Policy Entrepreneurship playbook

• Jordan Schneider’s Policy Early Career Guide, and ChinaTalk broadly

• Thomas Hochman’s learnings from One Year in DC

• Our resources on pathways into policy and tips for entering the space on emergingtechpolicy.org

This job description doesn’t sound like something most people would associate with government work. But it’s exactly what program managers at federal agencies known as “ARPAs” (Advanced Research Projects Agencies) do.

These roles are some of the most impactful in government, but filling them is hard on both sides. ARPAs often can’t articulate what they’re really looking for. Candidates likewise face a murky process: plenty of qualified people never hear about these roles, or assume they’re reserved for someone else—presumably someone who already has three PhDs and a security clearance and knows ten people on the inside. That might not hurt, but it likely won’t be necessary, either.

Below, I’ll cover what jobs exist at ARPAs—focusing mainly on program managers, around whom ARPAs revolve—and how you can pursue them, whether you’re mid-career, a recent grad, or still a student.

What ARPAs actually are

The most famous ARPA is DARPA: the defense-focused original, created in 1958 in response to Sputnik. DARPA’s founding mandate was, roughly, to ensure the U.S. would never again be blindsided technologically.

Decades of DARPA successes followed, including helping create the internet and GPS, among other achievements. This drove the launch of sister agencies focused on intelligence (IARPA), energy (ARPA-E), and health (ARPA-H), together moving ~$7 billion a year. DARPA makes up the bulk of this budget: ~$4.3 billion run by just 100 or so program managers.

You might say that the federal government runs what is essentially the world’s most heavily funded science fair. Importantly, though, ARPAs don’t run their own labs. They instead fund external “performers”—universities, companies, and research labs that actually carry out the R&D. That structure gives unusual leverage to the program managers who design and run them.

Much ink has been spilled on how ARPAs work, but there’s broad consensus on the key ingredient: empowering talented people to take bets on ideas that could transform entire fields. At most ARPAs, those people are called program managers (or PMs).

As Ben Reinhardt writes in “Why does DARPA work?”: “Every single description of the ARPA model agrees that it’s all about the program managers.”

(I’ll mainly focus on PM roles below, though there are many other great jobs at ARPAs. See our in-depth guide.)

So what do PMs actually do?

What program managers do

Picture the C-suite of a major research institute. Now imagine one of the institute’s scientists, Jane, gets promoted to CEO. Jane’s days, previously filled with test tubes and grant applications, are now spent forging the strategic vision, hiring top talent, building systems, and project managing. The context of her work is still scientific, but her new job is fundamentally operational: she must own the organization’s vision and keep the ball moving.

This is not far from the role of a PM. PMs must have the technical chops to define and pressure-test a hypothesis at the frontier of their field, but the bulk of their day-to-day is organizational and strategic: curating a team of researchers and engineers across national institutions, leading and coordinating that team, allocating the budget and tracking technical milestones, and deciding if and when to make a strategic pivot.

A former DARPA PM described it to us like this: “You’re running a multi-entity consortium that could be as big as the Human Genome Project, with your own staff, budget, and finite runway.”

This makes for a rare skillset, and one that traditional scientific training doesn’t really produce. Where then do PMs come from?

Who are PM candidates?

Given how ambitious but ambiguous the PM role is, ARPAs understandably struggle to describe what qualifications they’re really looking for. DARPA’s own description of its ideal PMs is telling:

“There is no one specific profile for a successful PM. Typically, our ideal candidates have been brilliant, done incredible work, and produced game-changing ideas.”

Accurate, as far as it goes, but not especially helpful guidance for prospective candidates.

The standard assumption of what you need to do to get an ARPA PM role is something like “get a PhD, meet the right people, do some Very Cool and Impressive Things.” This isn’t terrible as generic advice, but in practice PM backgrounds vary widely.

Most PMs hold PhDs, but some have master’s, bachelor’s, or even just associate’s degrees. Their fields range from physics and computer science to aerospace engineering, biotech, linguistics, sociology, and beyond.

Professional backgrounds vary just as much. At DARPA, PMs often come from academia, military, industry R&D, defense contracting, consulting, or startups. Many have prior DARPA or DOD exposure and already hold a security clearance (which is a big pro for the agency but not strictly required). A sample of current DARPA PMs gives you something like this:

• Tristan Tager (BS, Mathematics, Duke) co-founded an AI company building perception systems for DOD applications and served as lead AI research scientist at Genentech.

• Tabitha Dodson (PhD, Applied Physics, Air Force Institute of Technology) worked as a NASA fellow researching nuclear materials for nuclear thermal rockets and as an aerospace engineer for the Air Force on spacecraft propulsion.

• Jeremy Pamplin (MD, Uniformed Services University) is a colonel and critical care physician who built telemedicine systems for military forces and commanded the Army’s lab for automating casualty care.

Different ARPAs seek unique elements. At ARPA-H, for instance, clinical experience is common, as is FDA or NIH exposure. ARPA-E recruits many of its PMs from DOE national labs and corporate R&D. IARPA pulls heavily from intelligence contractors and IC agencies, in part because most already hold clearances.

So it helps to have a post-secondary STEM degree, subject-matter expertise, experience leading something, and some exposure to public-sector work—ideally with an ARPA or its parent agency, but not always.

But beyond some baseline credentials, the most important criteria for becoming a PM are how you think and how integrated you are into the relevant research communities for your field, as these will be major factors in your success in the role.

Becoming a program manager

If you’re convinced that working as an ARPA PM could be incredibly impactful and exciting, what should you do next?

Start thinking like one

As anyone in the private sector knows, when you’re betting on founders, trust is paramount, and choosing poorly can be extremely costly. Credentials matter, but often less than harder-to-vet qualities like prioritization, clarity, agency, and rational risk-taking.

Because of this, ARPAs care deeply about how their PMs think. Below are some ways you can start approaching ideas like PM.

Explore like an outsider. Former ARPA-H Deputy Director Adam Russell speculated that ARPAs succeed by attracting “aliens”—people who venture into new fields out of genuine curiosity, dissatisfaction with their home discipline, and a drive to tackle big-if-true problems that conventional thinking can’t solve.

Aliens tend to be low-ego. They’re less likely to overgeneralize and more likely to stay in a “learner” mindset. This humility makes it easier to spot systematic biases in a field’s literature—by asking naive “why” questions and making implicit assumptions explicit.

Scope ideas that are ambitious but tractable. ARPA programs start as big ideas, then get broken into concrete research agendas capable of delivering paradigm-shifting progress in roughly 2–5 years.

This means finding problems where:

1. Success would be a genuine breakthrough.

2. The technical barriers are hard but not impossible.

3. The problem is underexplored relative to its potential impact, though not entirely unexplored either.

One former PM gave us this tip: “Go to Google Scholar and set the date filter to ‘less than 5 years ago.’ If you can’t defend your idea with literature that recent, it’s not for DARPA.”

Developing this instinct for ‘right-sized’ bets is one of the hardest and most important parts of PM thinking.

Run your ideas through the Catechism. PMs famously refine programs by working through the Heilmeier Catechism—an 8-question framework that starts with a deceptively simple prompt: “What are you trying to do? Articulate your objectives using absolutely no jargon.”

Practice pressure-testing your own ideas with it. Then do that collaboratively, ideally with subject-matter experts (or skeptical colleagues and friends). Program pitches sit at the core of the ARPA model: even hiring largely revolves around how well candidates can articulate and defend their ideas. (Think Shark Tank for nerds.)

Read what PMs have written. It’s one of the fastest ways to absorb how PMs actually think. This ARPA reading list is a great starting point.

If these ideas resonate with you, it’s a good sign about your fit for ARPA work.

Network with the right people

PMs are expected to assemble research teams that span institutions and adjacent disciplines around their ambitious programs. This is why ARPAs look for people who already have networks they could tap, or who’ve shown they can build them quickly. Getting started early helps both in landing a PM role and succeeding once you’re there.

Think about building two kinds of relationships:

1. people in or around your ARPA of interest

2. subject-matter experts who could one day work on your program

A few practical ways to do this:

• Go where R&D people congregate. Attend Proposers’ Days, where ARPAs talk about open funding opportunities and meet prospective grantees, or ARPA-hosted events like the ARPA-E Summit. Conferences are also good for this: DARPA announced winners of its AI Cyber Challenge at DEF CON, for example. You could also start with more general federal R&D events like the AAAS Annual Meeting or NSF PI meetings.

• Do plenty of informational interviews. As much as you can, try separating your “learning” conversations from your “people I want to impress” conversations. If you’re building a network from scratch, start with lower-stakes conversations (e.g. with contractors or research performers) and work up to talking with former or current PMs at your ARPA of interest.

• Do your homework. ARPAs list their PMs and their program details online (with the exception of classified programs, of course). If you might be talking with someone who worked on one, read the description and come with good questions. Subscribe to your ARPA’s newsletters, podcasts, or social media.

Get near the action

You’ll want to position yourself close to ARPA programs, where you can build judgment, credibility, and relationships. A few tactical ways to do that:

• If you already have 5+ years of relevant career experience and are looking to pivot into ARPA or other R&D work now, consider the Big if True Science Accelerator (applications close June 7) or the Brains Accelerator. Both are part-time, Y Combinator-esque programs to help scientists and technologists design R&D programs and network in the federal R&D ecosystem.

• If you’re a recent graduate or early in your career, consider working as a SETA (Systems Engineering and Technical Assistance) contractor or as an ARPA-funded “performer.” Both let you interact heavily with ARPA programs and are among the most common on-ramps to PM roles. (More about them here). Some of the ARPAs also have internal fellowship programs intended for earlier-career folks, such as DARPA’s Innovation Fellows program.

• If you’re a student, try getting exposure to R&D, especially tied to federal missions. You might work on a federally funded project at your school, build a publication record, compete for ARPA-run student opportunities like hackathons and prize competitions, or do an internship or fellowship directly at an ARPA.

These and other suggestions are discussed in more detail in our in-depth guide on working at ARPAs, on which this post is based.

The bottom line is that ARPA PM roles are some of the most autonomous, high-impact, well-resourced, and, yes, potentially fun jobs in the federal government. Few other roles let one person shape the frontier of an entire field.

If you enjoyed this piece and want to learn more:

• Horizon’s guide to working at ARPAs on emergingtechpolicy.org

• An ARPA reading list from the Institute for Progress

• “How to Run Good DARPA Programs”: a Statecraft interview with former DARPA PM Joshua Elliott

• Building a Culture of Risk-Taking with Jennifer E. Gerbi, former Acting Director of ARPA-E

• Big if True Science Accelerator (BiTS) program resources

Launchpad is a resource from the Horizon Institute for Public Service, a nonprofit, nonpartisan organization working to address the U.S. government’s critical talent shortage in emerging technologies. Learn more about us and our programs supporting tech policy careers here, and explore our in-depth policy career resources at emergingtechpolicy.org.

The best ideas on working in emerging tech policy tend to travel through closed networks. After four years spent helping emerging tech experts enter and excel in public service, we’re starting Launchpad to share more of these ideas with you—whether you’re considering a move into the field, already deep in it, or tracking it from afar.

You can expect Q&As with people shaping emerging tech policy, analyses of the field, and practical career guidance drawn from our conversations with hundreds of hiring managers, policy leaders, and people looking to break in.

We’re publishing 3 posts this week to celebrate Launchpad’s launch—this is the first. Subscribe to get new posts in your inbox:

Subscribe now

If you’re considering a think tank job or weighing which ones are worth joining (or even if you already work in one), this post is for you. Horizon’s Remco Zwetsloot sat down with Todd Moss, founder and executive director of the Energy for Growth Hub, because Todd’s written some of the clearest thinking we’ve seen on how think tanks actually have a positive impact (much of it on his Substack, Eat More Electrons). We wanted to pull out the takeaways most useful for people evaluating whether think tank work is right for them—but Todd also shares valuable insights for anyone trying to understand how policy ideas actually get adopted.

Todd runs the Energy for Growth Hub, a nonprofit dedicated to ending global energy poverty. He’s a former Deputy Assistant Secretary of State for African Affairs and previously served as chief operating officer at the Center for Global Development, where he helped create the blueprint for what would become the U.S. Development Finance Corporation. On the side, he writes international political thrillers.

We cover:

• How to tell which think tanks actually have impact

• Why being a translator is often higher-leverage than being an expert

• The “poker game” of getting things done in government

• Who and what Moss bets on when hiring

Remco Zwetsloot: What do you think makes some think tanks more impactful than others?

Todd Moss: All humans respond to the incentives they face. For a lot of people, especially early in their careers, you’re responding to prestige signals from your peers and from the people you admire. If those people are signaling that hosting an armchair chat with a retired Secretary of State is the pinnacle of influence, then that’s what you try to do. If getting a publication in a particular magazine or journal is the thing, then that’s what you aspire to.

But some think tanks actually don’t aim for that. They’re aiming for real, measurable impact in the world, and the incentives inside those institutions encourage experimentation, failure, and really just trying to get things done.

You can pick this up in how think tanks explain their purpose. Some have trophy buildings, annual galas with people in tuxedos, and a stream of household names coming to speak. Others have none of that. They don’t do events. You might not have even heard of them. But they have a record of accomplishments—they’re placing people into government positions, their ideas are getting adopted by governments, and they can point to what they’ve actually achieved in the real world. How think tanks talk about themselves is a tell.

Going a little deeper here, you’ve said that people conflate outputs, impacts, and results all the time. What do you mean by that?

Todd Moss: If people are talking about their impact in the world and their answer is “a 75-page report was published,” they’re not connecting what they’re doing to a specific purpose. Maybe a 75-page report is exactly what the world needed. If you’re working on tweaking the regulations for nuclear safety, maybe a very detailed report could have high impact. But it’s not the report that matters—it’s who is using that report and toward what end. People often get those things conflated.

Moss’s 6-step model for tracking think tank impact, from “What Is Think Tank Impact? (And What Is It Not?)“

Horizon advises a lot of people who are thinking about whether policy is right for them and what jobs they should pursue. If you’re someone trying to decide whether to work at a think tank, how do you assess their work from the outside? An op-ed could be there just because someone wanted to write one, or it could be part of a strategic campaign—and it’s hard to tell which is which.

Todd Moss: If you get to an interview and the person interviewing you has an op-ed in the New York Times, I would ask them straight up: “I read your terrific op-ed. What were you trying to achieve with it?” And then listen. Most people have enough self-awareness and humility to tell you the truth. They’ll say, “Yeah, it got in for some random reason and it didn’t lead to any effect.” Or they’ll say, “Because it was in the Times, I got called by the staff of this senator and it actually wound up impacting this legislation,” or “So-and-so from Treasury called and they actually changed something.”

You will find a lot of very self-reflective people who have—even if they don’t articulate it well—a game plan. There’s a purpose to what they’re doing.

And then there are a lot of people who really think the op-ed is the end in itself, not a means to another end. They think it’ll impress their parents or their grandmother. My grandmother used to call me when I’d get in the newspaper, and that’s a fine feeling, but you’ve got to separate that personal satisfaction from actual policy effect.

For an employer to be honest and open about their purpose—and to be reflective not only on what went well but on what hasn’t worked—that’s what you’re looking for. I’ve spent almost my entire career trying to influence U.S. government policy from the outside, and the failure rate is huge. It’s much more like venture capital than index investing. You want somebody who’s reflective on that.

There’s a saying in DC: those who know don’t talk, and those who talk don’t know. How much think tank impact actually becomes public, versus staying private?

Todd Moss: For some think tanks, everything they do is public—that’s their model. But for the ones having a big effect, the majority of their work is actually done quietly.

At my organization, we put out a good amount of public content, and it helps us build our network, which is the purpose. But almost all of our high policy impact work is quiet. The majority of my policy work consists of one- and two-page memos written for a particular individual, never made public, and often tailored specifically to that single person to get them to do a particular thing.

Doing it in public would actually undermine what we’re trying to do. While we occasionally talk about our impact for fundraising and institutional growth, we actually want the policy changes to be owned by the policymakers.

So being discreet is part of being effective. At the same time, I still think it’s very important for institutions to be able to explain what they do. You can be even more candid with your board and funders. Sometimes there’s a zero-sum tradeoff between being well known and being effective. If you burn a relationship by being too public, you will have lost that channel permanently. I make tons of mistakes—I keep things private that don’t need to be, and I’ve revealed things that probably would have been better kept quiet. You have to make a judgment call on the costs and benefits.

Another question we often get is about how the impact of even well-targeted think tank work can feel uncertain relative to working inside government. You have to wait for a policy window, but you’re not sure if or when it will appear. When you work on a project, what signals do you look for to know whether to persist or pivot?

Todd Moss: The biggest signal is that I’m gaining traction with people on the inside. If you find something that is bugging people inside government—something they can’t solve and can’t even talk about—and you can give voice to that problem and start problem-solving from the outside, that’s how you gain credibility.

One of the things I often do is, long before we go public with an idea, we pressure-test it with people. “We know you’re trying to deal with this problem. Here’s an idea we’re playing with. What do you think?” Usually the first answer is, “That’ll never work because of A and B.” You say okay, come back, and say, “For A, we think we can deal with it this way. For B, we think we can overcome it like this.” As you start iterating, they often go from a skeptic to someone curious to an ally.

Ben Leo and I—my partner in crime in helping create the blueprint for the DFC—went to pitch it first to Senator Corker from Tennessee, a Republican who chaired the Senate Foreign Relations Committee. Literally in the first meeting, he stood up and said, “This is a terrible idea,” and told us why. We came back and said, “This is why you’re worried, but it actually isn’t a problem for these reasons.” We went back and forth, and by the third conversation, his staffer was coming to us with new ideas to make the proposal better.

The idea changed quite a bit through that process, and we had to be open to shifting. But that iterative process—which had to happen behind closed doors—is what allowed the idea to grow, evolve, and become feasible. You can think big thoughts on the outside, which is a freedom staffers don’t have on the inside. But you can also come up with wacky ideas that can’t possibly happen because you haven’t thought of everything. People on the inside are in a much better position to judge political and practical feasibility—they just don’t have the time or space to think big. So you have to build long-standing trusted relationships. Trust is what matters most.

And you can also play a connector role across a siloed government.

Todd Moss: Exactly. It sounds a little like horse trading, but outsiders will hear of things going on across different agencies long before insiders might. You can drop crumbs—”Hey, just a heads up, this is what I heard”—and you’ve built an ally. They’ll come back to you with other information and ideas. I have people I knew as congressional interns who are now in super influential positions, and we’d been joking and sharing tidbits for ten years. It’s playing the long networking game.

When you were newer to policy, what were the formative learning experiences that shaped how you work?

Todd Moss: I’ve got a lot of formative stories, but two really stick with me.

The first is about how to communicate with government people. When I was at the Center for Global Development, the standard tool was a cross-country growth regression—papers with regression tables and statistical analysis. Then I went into government, and not once, ever, did I see a regression table. Instead, what I had was a steady stream of extremely short summaries—usually one page or less—telling me the bare minimum I needed to know, what was credible, and how to make decisions from that information. You’re just overwhelmed.

When I went back to the think tank world, I literally stopped doing cross-country regressions. The real value isn’t always producing unique research—it’s that there’s so much amazing analysis in the world that sits on a shelf, completely inaccessible to policymakers because nobody is translating it into the words and format they can absorb.

So the highest-leverage position is being a translator between the analyst-expert community and the government community, rather than being one of many dozens of experts producing long reports.

Todd Moss: One hundred percent. But it’s also about problem identification. In academia, the upstream decision-making about what to research is driven entirely by publications and methodological questions. But in government, they have a couple of top-line burning questions. If you’re not answering a question that’s already top of mind, your work is irrelevant. So it’s the translating and the problem identification—those are the two very high-leverage points.

The second formative story: On my first day as a State Department official, my boss, Assistant Secretary Jendayi Frazer—a hard-charging, elbow-throwing policymaker who was very effective—grabbed my forearm and said, “I want you to remember: the enemy is right here in the building.” I didn’t know what she was talking about.

What I learned is that the art of policymaking is not convincing some foreign leader to be nicer to the United States. It’s getting different parts of the U.S. government and different parts of the State Department all on the same page. There’s a whole poker game going on—people using influence to stop or advance certain policies—and it takes a while to figure that game out. That’s what she was telling me.

So for someone new to the field, the takeaway is to do work that isn’t just writing reports—do the one-to-two-page memo format, be in rooms where you have to tailor your output to particular decision-makers, and learn the interagency dynamics.

Todd Moss: Yes. And take the networking component seriously as a skill. It’s not a nice-to-have—it’s the channel for actually getting things accomplished. If you’re walking into a meeting and you want to get something done, and you don’t already know what everybody thinks about it, you’ve already lost. Everyone else has pregamed what they’re going to do and where they are. If there’s negotiating to be had, it should have already happened.

If you’re a scientist thrown into some interagency meeting and there are 25 people in the room and you’ve never seen 20 of those faces before, you’re completely lost. Get to know your peers in different parts of your building and the other buildings that matter in your space. Get to know what they care about, what they object to, what motivates them, what their boss is telling them to do. That will make you much more effective.

Never waste a meal or a happy hour. Every lunch is an opportunity to sit down with somebody, break bread, learn about them and their background. You’ll start to learn the culture. That’s where you’re going to network far more effectively than two people across a desk in suits.

You’ve written about Nancy Birdsall betting on young rising talent and giving them freedom to run, and you’ve said you follow the same recipe at the Hub. What do you actually look for when hiring, especially in people without traditional policy backgrounds?

Todd Moss: I look for people who are genuinely curious and self-aware and self-reflective. Think tanks have a lot of freedom, so you need people who are genuinely motivated by the mission and willing to try things and learn. People who are in think tanks to pontificate about their expertise are not that interesting. People who want to learn and problem-solve in groups—that is where you get a lot done.

You certainly want somebody who’s trying to learn the policy process. I often ask people, “What’s your dream job in ten years?” The best answers are an ambitious stretch of a government job, but something that’s doable. One of the great things about the United States is you can be in your twenties and get in a really influential position. That is not true in almost all other governments. I was a deputy assistant secretary at 37, and I was 30 years younger than almost all of my African counterparts.

How can a candidate tactically signal these qualities to a hiring manager?

Todd Moss: People do really well when they’re comfortable with the content but aware that they’re not experts yet. They signal that they’re a sponge and they just want to learn and work hard. DC is full of really smart, young, ambitious people working incredibly hard, and the best candidates want to be part of that next generation making U.S. policy better. I definitely want people who show humility. If you lack humility early on, it creates all kinds of problems.

One thing I’ve started doing: I’ve found that U.S. universities are doing a terrible job of surfacing underappreciated talent. So we’ve started asking applicants to answer three questions, and we read the answers blind to their name and CV. It helps us get away from affinity bias. Part of me would love to hire a Tufts grad—I went to Tufts—but overindexing on that makes my recruiting decisions worse, not better. Some of the best people I’ve ever hired did not come from top schools. They were just really hard workers who stood out in other ways. I try to discount the big name-brand schools and put more weight on grit, creativity, and passion.

A notable absence from your list is deep subject-matter expertise. We see AI experts come to Horizon who are curious about policy but feel like they should stay in the sector longer because they “don’t have all the answers yet” on what they think U.S. AI policy should be. How much should that actually give them pause, versus just jumping in?

Todd Moss: I think there’s a baseline you need of comfort and experience with the basics of the field, and you need some experience with how things have happened in the past. But if you come in thinking you’re the world’s expert on something, the chances that you’ll be effective at changing policy are actually quite low. You won’t have the surrounding context. Policy influence is not about coming up with the ideal solution—it’s trying to make the best of a complete mess. Nothing is working ideally. It’s a chaotic dumpster fire, and you’re trying to get the best outcome you can. The world’s leading expert won’t see that as success, but that is what you’re going for.

The other thing is that you can always buy in expertise. I work a lot on nuclear power. I know nothing about nuclear engineering—nothing. But I have a whole network of people who can tell me what it takes to turn nuclear fuel into a weapon. Those people, though, are not going to be good at designing a policy regime to prevent countries from doing that. They don’t have the policy and political context, or an understanding of how the bureaucracy works.

So it’s knowing what you know and finding out what you don’t know. That speaks again to the networking effect—outsource as much as possible and focus on your core mission: getting to the best outcome you can.

Any final piece of advice or inspiration for someone who’s excited to have policy impact but doesn’t know where to begin?

Todd Moss: Public service can be frustrating, but it’s so rewarding to see your efforts become a reality in government, because the scale is so much greater than anything else you could do. It’s why I work on energy policy instead of running power lines or installing solar panels. I don’t want to just do a few things—I want to try to impact entire sectors and entire countries’ trajectories. That’s what public policy allows you to do.

Big thanks to Todd for taking the time. If you enjoyed this conversation and want to learn more:

• Todd’s Substack: Eat More Electrons

  • How to Get Sh*t Done in Washington DC

  • What Is Think Tank Impact? (And What Is It Not?)

• Horizon’s guide to think tank work on emergingtechpolicy.org

• CGD: How Might Think Tanks Make Real Things Happen? Lessons from the Creation of the DFC

• How to Reverse a Coup, Todd Moss on Statecraft

• Policy Entrepreneurship, Renaissance Philanthropy

Launchpad is a resource from the Horizon Institute for Public Service, a nonprofit, nonpartisan organization working to address the U.S. government’s critical talent shortage in emerging technologies. Learn more about us and our programs supporting tech policy careers here, and explore our in-depth policy career resources at emergingtechpolicy.org.