Astro Teller says "most of why we waste time at work is fear." He runs Google X, the division that built self-driving cars at Waymo, and he has spent years building a workplace where people actually behave like anything is achievable. I re-listened to his HBR IdeaCast episode this week and two big ideas stuck with me:

1.) Teller has a thought experiment: If your job is to train a monkey to recite Shakespeare on top of a ten-foot pedestal, which do you do first? Everyone at traditional organizations builds the pedestal, because you can show your boss the pedestal and your boss says "good job," even though you've made zero progress because all of the risk was on training the monkey. If the monkey can't learn Shakespeare, the pedestal is scrap. His operating metric is "learning per dollar," and the logic is simple: "You learn nothing when you succeed, except maybe to do that again. You learn exclusively when you fail... failure is learning. They're identical." Healthcare, legal, and software engineering are already running cheap learning loops like this with AI. The opportunity is there for anyone willing to invest in the experiment.

2.) During the interview, Teller does his one-hour innovation talk in sixty seconds. Choice A: A million dollars of guaranteed value for your business this year. Choice B: A billion dollars, one chance in a hundred. Everyone picks B, and then he says leave your hand up if your manager would support that choice, and every hand in the audience goes down. "You don't need a lecture on innovation. You need a new manager." When projects die at X, the people stay, the code stays, the learning stays. Teller calls it "moonshot compost." Most organizations don't have anything like this. In my world, when a campaign ends, you write an exit memo you barely finish in November after three months of 60-hour weeks and then you walk out the door. A handful of people hold the knowledge, and when they leave, the information is dead. But that's not unique to campaigns. Any organization with high turnover or short project cycles faces the same problem, and Teller's answer is that you have to design your system so nothing is wasted.

Teller rollerblades through Google X every day as one of what he calls the "hundreds and hundreds of signals" you have to send so people actually behave differently, when every professional instinct drives you away from taking billion-dollar risks. I think about what it would look like to send those signals in any organization where the cycles are short and the fear of failure is real. Teller's point is that you can build systems that make it safe to learn, and that the learning is what makes you win.

Thanks to Harvard Business Review, Alison Beard and Astro Teller for a conversation worth the relisten.

I spent an afternoon at the National Gallery sitting in front of Simon Vouet's Saint Jerome and the Angel. Oil on canvas, early 1620s. A painting of a bedraggled old man at a desk.

Jerome is half-naked and tired. His chest is exposed, sinewy, sagging. Papers scattered across the desk. The inkwell open. An hourglass where he can see it. Books, scrolls, the clutter of a man who has been wrestling faith into language for decades.

An angel leans in from behind - not from across the room, not from heaven - right there, pressed close, almost invasive, a trumpet at Jerome's ear. One hand gestures outward as if to say: Aren't you listening? Get all of this down. Jerome's left hand grips the quill. His right hand reaches for something he can't quite hold.

The Caravaggisti light selects Jerome's body, the angel's robes, the papers, and denies everything else. The work happens in a narrow cone of attention surrounded by darkness.

Jerome is famous for translating the Bible into Latin: the Vulgate, the text that shaped Western Christianity for a thousand years. We abstract this into a feat of scholarship. It wasn't. It was an old man forcing divine meaning through a human nervous system, one sentence at a time, while the sand in his hourglass trickled down.

In his Commentary on Ezekiel, written near the end of his life, Jerome described visiting the Roman catacombs as a young student:

Often I would find myself entering those crypts, deep dug in the earth, with their walls on either side lined with the bodies of the dead, where everything was so dark that almost it seemed as though the Psalmist's words were fulfilled, Let them go down quick into Hell. Here and there the light, not entering in through windows, but filtering down from above through shafts, relieved the horror of the darkness. But again, as soon as you found yourself cautiously moving forward, the black night closed around, and there came to my mind the line of Virgil: "Horror ubique animos, simul ipsa silentia terrent."

Horror everywhere seizes the soul, and the very silence terrifies.

He went in anyway and kept going back, and then spent his life doing work that looked, from the outside, like a man at a desk with an inkwell. From the inside it was closer to what Vouet painted: the angel's trumpet that won't stop bleating and the hourglass that won't wait.

Down the hall hangs El Greco's Saint Jerome, where the same saint lives in a different universe.

No desk. No papers. No tools of the scholar. Just the man, stripped to sinew, standing on rock, looking upward into light that doesn't obey physics. His body elongated, twisted, barely on the ground. One foot forward, the other knee back on a rock. He could be stepping toward something or falling backward.

El Greco's contemporaries thought his anatomy was wrong, his proportions distorted, his colors unnatural. Mr. Theotokopoulos didn't care. He was painting what belief weighs, and if that bent the body into shapes that don't exist in nature, so be it. When someone asked what he thought of Michelangelo, he said: "He was a good man, but he did not know how to paint."

El Greco carried a conviction so total it looks like madness, which I find not so far from Jerome descending into the dark.

Vouet's Jerome is the labor of obsession: the desk, the pen, the hours, the angel dictating faster than the hand can move. El Greco's Jerome is the fire of it: the body spent, the ground unstable, the light from somewhere that doesn't have a name.

Painters returned to Jerome century after century. Hundreds of them returned, known and unknown, because what he represents is inexhaustible: a person seized by work they didn't choose so much as submit to. The obsession was their own as much as his, whether their Jeromes hang in storage or were painted over on ancient altarpieces.

Jerome went into the catacombs where the walls were lined with the dead and the silence terrified him. He kept going. He spent the rest of his life translating the ineffable. He didn't know that work would become foundational. He just knew the trumpet was in his ear and his hand was on the pen.

Most people using AI at work are delegating projects they couldn't delegate to a person, then getting frustrated when the computer can't read their mind.

It does exactly what you specify

Last year I started a firm building AI tools for campaigns. The hardest problem is getting people to articulate what they want a project to achieve. Try putting into words how you actually decide on a client's messaging strategy and the research projects you need to support it. It's probably incoherent. It contradicts itself. It relies on unspoken norms we expect staff to absorb osmotically over years. We get away with it because humans fill gaps with intuition.

AI doesn't do that. It has perfect integrity: it does exactly what you specified. The gap between what you wanted and what you got is the gap between what you said and what you actually value.

AI skeptics overestimate how special our corner of "knowledge work" is. AI is doing useful work in radiology, legal research, and software engineering. It's absurd to think what happens in a strip-mall campaign office is innately more complex than that. The underlying problem is that smaller industries haven't invested in product design. We're trained on old processes, arcane bibles for how to Google the right way and format a memo about our findings. The real skill is finding answers to difficult questions, and that's jagged by nature. No process manual covers it well.

Nobody wrote it down

Jaya Gupta and Ashu Garg from Foundation Capital call this the context graph problem: the gap between what professionals know and what they've written down. Your AI isn't in the conference room when you have the idea. It doesn't have access to the Signal thread, the strategy doc you read last week, or the vibes in your heart and head. That's the gap it falls into.

Ethan Mollick gets there differently: "In figuring out how to give these instructions to AI, it turns out you are basically reinventing management." What does "done" look like? Teams defer these hard definitions because urgency doesn't arrive until the polling memo is due. Humans muddle through because they can course-correct late. A language model can't. It just goes and does the thing, and without a committed definition of done, it falls straight into the gap. The user expects magic and receives nonsense.

Be smart or persistent

Most people disappointed with AI treat their first prompt like a delivery instruction: "Build me a messaging strategy that will win the campaign." They're skipping the hard part where you figure out what you're building, for whom, under what constraints, and what you're willing to get wrong. This is a learnable skill. You either need to be smart about it or persistent; best results if you're both.

Smart means answering the product questions before you touch the technology. What makes a research finding useful? When is a project done, defined as something you'd be proud to show your client — versus the deadline hit and whatever you had is what you sent over? Before I got any staffer or AI system to produce work I'd trust, I had to answer these questions. None had anything to do with AI.

Persistent means staying with it when the first output is bad. AI is good at helping you figure out what you actually want. When one produces something wrong and you have to explain what's off, you're extricating your own decision trace. Your first prompt is a hypothesis. The output shows you what you left undefined. You iterate the prompt, not the output. Most people try once, get slop, and walk away. The ones who stay find something useful because bad outputs surface requirements they didn't know they had.

We never had to write any of this down because people were there to fill our gaps. AI is forcing the definition meeting we've been skipping for decades.

When I got my first shot at running a research department, I was terrified I was going to screw it up. So I started making calls to consultants who'd been in the business for decades and campaign managers who'd built and run departments a lot bigger than the one I was about to lead. I asked all of them the same question: how do you hire well?

The answers I got were tricky. When I actually sat down to build a hiring process around them, the advice didn't hold together.

I think part of the reason is that campaigns are like startups that pop up every two years and then evaporate into mist. You're working 60-hour weeks. Mentorship networks are thin. Nobody has time to systematize hiring advice when they're building everything else at the same time. I get why the advice is the way it is. But I think we can do better, and it starts with knowing what you're actually looking for when you sit down across from somebody.

I've spent years now thinking about how I hire, testing it, and passing it along to my staffers as they've moved into managing their own teams. This is the advice I wish somebody had given me.

The advice that didn't hold up

Three things came up when I asked people how they make hiring decisions.

The first was "hire someone manageable." I heard this more than anything else, and I think it points in the wrong direction. What does manageable even mean? How are you going to assess that in one or two interviews and a writing test? I think what happens when you engrave "manageable" into your mind is that you end up hiring people with low agency: people who wait to be told what to do instead of figuring it out. That might make your life easier in the first week, but it's part of why every department is backlogged with work. I want someone who can receive a task and go find the way it should get done. There's a Steve Jobs interview where he says the greatest people are self-managing: Once they know what you want them to do, they figure out how to do it. They need your vision for the department and not your micromanagement.

The second was some version of "my friend recommended them" or "they were on a winning campaign, so they must be good." Referrals are useful, but they're not a decision. You don't know what management system that person's old team was operating in. You don't know whether they contributed to the win or just happened to be in the room. And when that hire doesn't work out, the person who made the referral isn't the one dealing with it. You are.

The third was "You'll just know. It's vibes." I want to be fair about this one. There are managers with a lot more experience than me who have genuinely sharp instincts about people. But when you pass down "you'll just know" as advice to someone who's managing for the first time, you're setting them up to fail. They haven't had years to develop those instincts yet. There's a real difference between someone who is a good hang and someone who is going to help your team fight to expand Medicaid. If you're doing this for the first time, you need a framework.

Humble, hungry, and smart

Jim Collins talks about getting the right people on the bus. Everyone knows some version of that line and nods at it. But nobody tells you how to actually figure out who the right people are. One framework gave me that answer, and it's held up through every hire I've made since: Patrick Lencioni's model of humble, hungry, and smart. (If you want the deep dive, this Coaching for Leaders episode was my starting point.) I built my interview process around these three qualities. I make sure I'm asking a couple of questions for each one so I have a real sense of who I'm talking to.

Humble is about ego, or the lack of it. Does this person share credit? Do they show gratitude to the mentors and bosses and teammates that helped them get where they are? You're listening for whether someone talks about "I did this" or "we built this together." Now, I'll be honest: in an interview, everyone has to be the hero of their own story to some degree. That's the format we all expect. It's on you as the interviewer to make room for humility, not expect candidates to volunteer it against their own interest.

Hungry is about drive. Are they motivated to learn and do more without you having to push them? On campaigns this might be the most important one, because you need people you're willing to be locked in a room with at 11 PM on a Tuesday while votes are being counted. Lencioni makes the point that of the three, hunger is the hardest to teach someone later. That internal drive is pretty much there or it isn't, so interview hard for it.

Smart in this framework is simpler than it sounds. Lencioni describes it as common sense about people: Knowing what's happening in a group situation and how to deal with others effectively. Do they ask good questions? Do they actually listen to the answers? Can they read how their words and actions land on the people around them? For research positions, this shows up in whether someone can take a complicated finding and explain it clearly to a comms director who needs it in 30 seconds not 30 minutes.

Putting it into practice

Be friendly in your interviews. All the campaign hiring guides I've read suggest interviewing way too hard for what they're actually trying to assess. People think the fair thing to do is sit there stone-faced on Zoom and grill your candidate. You'll learn a lot more about somebody if you actually care about who they are and let them feel comfortable enough to be themselves.

Ask more thoughtful questions. Instead of "tell me a time that your boss was critical of your work," try: how do you receive feedback, and what's the best way for someone to give you feedback? That tells you whether someone has ever actually thought about their own learning process. Instead of grilling on your specific methodology, try: Tell me about an interesting research or policy project you've worked on. Then listen for depth and potential.

Expand what counts as relevant experience. Every job in politics is basically an entry-level job because you get asked to assume so much responsibility so fast. I encourage candidates to draw on academic work, extracurriculars, service jobs, whatever. If someone tells me they spent a semester working through Hegel's Phenomenology of Spirit, I don't need them to validate that they've used LexisNexis before. They have the firepower to read new clips. If they've worked at Cheesecake Factory, they can handle an all-staff call.

Test for ceilings, not floors. The way most campaigns run assessments, you end up seeing what candidates can't do instead of what they can. Give people more time than you think they need and see how they operate. You're going to need to coach them for the position anyways, so a candidate's ceiling tells you more than their floor.

A closing thought

If you're managing for the first time, don't be discouraged about the feeling that you're flying blind. Every manager is flying blind because everything the last team figured out disappeared when their campaign ended. You're figuring it all out while still doing most of the production work yourself.

That said: Every person you hire is someone you're going to train, invest in, and go to war with. Having a real framework for figuring out who belongs on your team — even a simple one — will make your life better and theirs better too. Build for high ceilings. Invest in your people. The staffers you hire now are going to become the managers who hire the next generation.