Darragh Buckley started Increase in 2020 to fix the problems he faced while building Stripe. Today, Increase moves over $500 billion a year with a team of 33. And they just bought a bank.
He taught me why it’s so hard to buy a bank, how to do it, The Fed’s huge 2025 tech stack migration that no one noticed, and stories from the early days of Stripe, like their overnight build to turn on the first dollar of revenue, and why John and Patrick Collison used to sprint around the office.
Support this Episode’s Sponsors
Monaco: The revenue engine for startups.
Flex: Sign-up for Flex Elite with code TURNER, get $1,000 here.
Clerk: Everything you need for authentication.
Numeral: The end-to-end platform for sales tax and compliance.
Inquire about sponsoring future episodes here.
Timestamps to jump in:
0:00 Stripe's 1st employee bought a bank
1:38 Fixing the stone in his shoe building Stripe
7:59 Banking looks like 2010 payments
11:11 Merging two layers into one, just like Stripe
14:52 Why he finally bought the bank
16:26 How do you actually buy a bank?
20:20 The first bank that got away
22:04 Why banks can’t chase new fads
25:26 Compliance as code, from day one
29:46 Talking straight to the Federal Reserve
32:19 The Fed migration that nobody noticed
33:50 The missing last mile of banking
35:44 Getting customers to switch
38:23 The challenger sales model
40:33 What customers use Increase for
41:48 How Ramp's bill pay works
42:37 $500B a year with 33 people
45:31 Hiring for a high-empowerment culture
47:08 Growing up with the Collisons
50:27 150 test cases in one night
54:23 What really made early Stripe work
57:07 Why Patrick and John sprinted everywhere
1:00:00 “You're not in payments, you're in fraud”
1:02:31 Where AI hits a wall on fraud
1:05:08 Giving payment cards to AI agents
1:09:27 Use boring technology widely
1:11:37 You want your banker to be bored
Find Darragh on X / Twitter and LinkedIn.
👉 Stream on YouTube, Spotify, and Apple
Transcript
Find transcripts of all prior episodes here.
Turner Novak:
Darragh, welcome to the show.
Darragh Buckley:
Cheers. Thank you very much for having me.
Turner Novak:
Thanks for being here. So you were the first employee at Stripe, and you just bought a bank. What’s going on?
Darragh Buckley:
I did, yeah. The bank is now called Increase Bank. I run a company called Increase. Increase exposes US depository functionality by API.
So it’s having your machines talk to my machines to be able to orchestrate your finances. That’s things like opening accounts, sending and receiving transfers, issuing cards, balance and transaction reporting. Our users and use cases are things like payroll, bill payment, loan servicing, neobanking. Some of our more public users are companies like Gusto, which is built on top of us, and some of Ramp is built on top of us.
We do their bill payment, for example. To give a sense of scale, we do a little over half a trillion dollars in payment volume these days. And those users choose to use us because of reliability, transparency, and flexibility. They themselves are payments companies, or folks familiar with payments, who want all of the details and the capabilities. Increase is designed to expose all of the data and capability of US depository functionality by API.
The joke is always that Ramp does not need me in any way, shape, or form to explain ACH payments to them. They’ve got it. They just want all of the data and all of the capability in a way that’s composable into their own product.
Turner Novak:
So could you not do that five or ten years ago?
Darragh Buckley:
Well, if it’s okay here, I’ll speak to some experience from Stripe. While I was building Stripe, we built on top of a cohort of US banks that grew out to global banks, integrating with the card networks.
And the banking, as it was exposed to us, you could always tell that we were always a little bit of an awkward duck. You were always wearing someone else’s shoes. Something always felt a little off, be it the size or the scale.
Turner Novak:
So what felt off specifically?
Darragh Buckley:
So Stripe never did disbursements to merchants by wire. Stripe would receive a chunk of change from the card network, from Visa, and need to push that out to Stripe users. A very natural way of moving those large amounts of money is by wire. But Stripe was always hesitant to do that, because when a wire would fail, we would receive an email from the bank partner saying, “We were unable to deliver this wire. Could you please go repair it?”
And it’s just hard to build a product that you’re proud of on top of those kinds of manual flows. There are dozens of these examples. “Oh, that data element? Yeah, we don’t pass it through to you. Why would you want it?” So you always knew you weren’t in the right place.
You knew you were building on top of someone else’s product. There was a stone in your shoe, and you were trying to explain to someone why the stone in your shoe was super annoying.
Turner Novak:
But they can’t see the stone in your shoe.
Darragh Buckley:
Correct. So they’re just like, “Why are you so worried about your shoe? Just keep walking.” Very much a princess and the pea kind of problem. But you know you’re not in the right place. You know that this is not built for you, and no fault on it. It’s built for someone else. Products are built for their users. You just don’t happen to be in the target users.
And so Increase is the banking technology that I wanted while building Stripe, exposing all of the data and all of the capabilities by API. Of course you want to be able to originate the ACH by API, and of course you want to be able to handle the return of it by API, and of course you want to be able to handle the secondary and tertiary processes. In the case of ACH, this is hold harmless requests and so on, by API.
Turner Novak:
What are secondary and... I don’t actually know what those are.
Darragh Buckley:
It’s the dispute process that comes out at the end.
Turner Novak:
Oh yeah. I get emails from my bank sometimes. It’s like, “Someone just pulled an ACH. Do you approve the ACH pull that has already been pulled?” Well, actually, maybe you could have told me before that.
Darragh Buckley:
The train has left the station. Would you want it to have? Do you want it to come back? And no fault on that. But all of those should be exposed to companies that are looking to programmatically orchestrate their finances by API. The alternative is that you end up growing large manual operations teams, or worse, just not building that product.
Turner Novak:
So did Stripe do that for a while? Did you build around these weird capabilities?
Darragh Buckley:
Oh, certainly. You end up... I know Stripe ended up with a team at one stage with dozens of people on it for ACH return reconciliation. And the reason that was necessary is that there was a particular data element in the ACH file, called a trace ID, that didn’t traverse correctly.
And when you dug into the details, you could see why Wells Fargo was altering this data element. What they were doing was not unreasonable. It wasn’t like the engineers there built it badly. But it wasn’t built under the anticipation that you’d want to compose that within your own product.
Anytime you see a product that almost fits so close, and that we have such buildings built on top of it, one of us should surely go back and do a redo of the foundations. And that’s what prompted Increase.
Turner Novak:
So there have been these infrastructure providers in the fintech space, and it sort of sounds like some of them would solve some of these problems, right? Like, were they not fully solving it by not owning a bank? What was going on?
Darragh Buckley:
The short answer is yes. But let me give you a more interesting answer than the short one.
Turner Novak:
Yeah, we need this to be an interesting podcast. We’ve got like eighty minutes left to go. You cannot just say yes.
Darragh Buckley:
Modern banking has many comparisons to 2010 payments. In 2010 payments, so pre-Stripe and Adyen and similar, back in the day you’d go get what’s known as your merchant account. Your merchant account is your network-level identifier for you to be able to process transactions on the card networks.
Turner Novak:
This is like the equivalent of if you had a card reader when you swipe your Visa or MasterCard. Is that your equivalent of that? Your card reader would reference your merchant account?
Darragh Buckley:
Exactly. You’d get that from Wells Fargo Merchant Services or similar.
Turner Novak:
Weren’t there also some local, really regional sellers of these on-ramp type things?
Darragh Buckley:
Oh, very much so. There’s a whole industry called independent sales organizations, ISOs, which are the boots on the ground and the pavement pounders of the payment card industry. And it’s fascinating where they grew up, in different regions and jurisdictions, specializing in different things.
Turner Novak:
So did they ultimately evolve as a go-to-market function of the card networks back like fifty years ago, whenever they started?
Darragh Buckley:
Indeed, indeed.
Turner Novak:
I very much appreciate this depth-first approach to this question, by the way. Just a hat tip to that. It’s like, I’m always really fascinated by how things came to exist the way that they did. Because you’re like, “That’s kind of messed up. Why do we have all these layers?” There should be an API to just say, “Hey, send a wire,” and it just does it.
Why did that happen? Because then you know how to fix it if you understand how it works.
Darragh Buckley:
Oh, certainly. We’ll even take a step back and go, why is it messed up? That’s a fascinating trigger question for, “Oh, I’m about to learn something.” A solution is arrived at that I don’t understand. And the next question is, “Great, I must not be seeing a constraint. Where are the constraints that I’m not seeing?” That’s always a fun game to play. What am I currently blind to?
Turner Novak:
Yeah, what am I not seeing? The fun thing about the podcast is sometimes you have to hit people at different angles of the same question. “What do most people get wrong?” Or, “What was the most surprising thing you learned about this?” It’s kind of like poker. I’m like, “All right, the third attempt, I got him to finally explain this to us.”
Darragh Buckley:
Yeah, different views reveal different things at times, and fire different neurons.
Turner Novak:
So basically you get this merchant ID to get onto the card networks. But great, you now have this identifier. What are you going to do with it? You’re building your software-as-a-service company. You needed a way to programmatically interact with that.
Darragh Buckley:
You have your identifier. What are you going to do with it? We had our merchant identifier, and then you had what was known as a gateway. These were the things like authorize.net. So back in 2010 payments, what you would do is get your merchant identifier, bring it to authorize.net, and that would be how you’d be able to programmatically interact with the network.
Turner Novak:
And among the insights from Stripe was that these two things are not two separate things, right? Why are you talking to two companies here? This is one. So you combine them essentially.
Darragh Buckley:
Stripe combined them. That was at least one of the initial insights of Stripe. One of these things is not worthwhile without the other, so therefore you should bundle these.
And that’s what allowed, for example, Sunday night signups on Stripe. Everything was within Stripe’s control. It wasn’t like Stripe told you to go off and talk to Wells Fargo Merchant Services and come back when you had your identifier.
Turner Novak:
Stripe would just take care of it all. And would you almost be like a buffer layer between the rougher, harder parts? You’d do that inside of the Stripe back end, so it made it simple for people to do?
Darragh Buckley:
At times, yeah. You’ve got to be careful with those abstractions becoming leaky, but yes. What you would attempt to do is render the terms of the credit card networks into terms that were going to be more usable by developers and engineers.
Stripe creates these beautiful abstractions over all of these things. This is actually in contrast to Increase, but we’ll probably get to that in a bit.
Turner Novak:
Okay, so 2010 payments, you’re in this ecosystem of gateways and merchant providers. And Stripe verticalized that. Current-day banking ends up somewhat similar, right? You end up with these layers that are built on top of banks, and fundamentally you’ve got the bank, the Wells Fargo, running their own what’s known as bank core, their own maintenance of transactions and balances.
Darragh Buckley:
That’s what a bank core is.
Turner Novak:
I’ve actually never known that. I’ve seen the word many times.
Darragh Buckley:
There can be dragons with respect to bank cores, but I think you get a good first-pass model if you just count it as maintenance of balances and transactions.
Turner Novak:
Okay. So it’s almost like a general ledger, like a live, living ledger of the system of the bank.
Darragh Buckley:
Yes. It’s actually their client ledger. And so oftentimes you’d end up with a general ledger behind it. But boy, do I love my pedantic details. Happy to double-click on this anytime.
Turner Novak:
It’s probably good that you know all the details.
Darragh Buckley:
Well, I enjoy digging into these. It’s not like these come to anyone at birth. If your interests take you this direction and you keep swimming, you end up with these fun hobbies.
So current-day banking, when you end up with a technology layer or an API layer that’s talking to a bank’s core, looks very like 2010 payments. And my view is that’s probably wrong. What we need is a bank with a technology muscle. We don’t need to keep adding layers up on top of bank technology. We need to drop down a layer.
Turner Novak:
Now, I’m always pretty skeptical when someone says, “Oh, we need to rebuild that technology,” because usually that’s a bit of an odyssey.
Darragh Buckley:
But we’ve been building on top of banks’ technology successfully for 15 years. It’s actually worthwhile that we just do a double-click as to whether or not there are parts of it we should replace. So that prompted Increase.
Increase verticalizes that. Increase replaces both the bank’s core and the API that you use to talk on top of it, in much the same way that Stripe did this back in 2010.
Turner Novak:
And you started in 2020, but you just bought the bank within the past year. So what was going on during that time?
Darragh Buckley:
My journey into banking started with Increase in 2020, building the technology. The way in which we worked then, and we still do, is we have the technology with a series of bank partners. In particular, we built the core, but having that talk directly to the underlying network, so talking directly to the Federal Reserve for FedACH, Fedwire, Check 21, FedNow, and talking directly to Visa for our card-issuing products.
We’ve had that for years and years. And that was very necessary for me to build first. Increase built that first because that’s the part that’s very important for us as a company. As we grow, we get to know and understand whether it’s just me and my friends who are interested in this, or whether the appeal is broader.
Whether other people felt the same problem and wanted it solved. So the technology has been up and running for years and years, but the bank purchase completed maybe about a year ago at this stage. A small Washington State chartered bank called Twin City Bank. Fun fact, it’s probably the closest bank to my house.
Turner Novak:
I live in Central Oregon.
Darragh Buckley:
And we rebranded it about last month as Increase Bank.
Turner Novak:
So how do you buy a bank? What’s the process of doing that? Because not many people do that. I actually saw there’s an NBA player recently, I don’t know, Harrison Barnes. I think he’s retired, but he bought a bank.
Darragh Buckley:
I came across that when I was looking this up. So I guess some people do buy banks.
Turner Novak:
So how does it go?
Darragh Buckley:
Well, bank investment is actually relatively common. I recognize that bank investment to the stage of controlling a bank might be less so. But fun fact about banks around the US is they can publicly sell stock. They’re eligible for a Securities and Exchange Commission exemption that allows you to publicly sell stock.
Turner Novak:
So you could in theory just buy all the public stock of a publicly traded bank, and then you would own the majority of the shares?
Darragh Buckley:
Yes, in the same way that you could do so with any public company. But the fun parts are obviously when you run into various thresholds. In particular, banking has restrictions on who can control a bank. And there are control thresholds. The first is kind of a 10%, the second is at 25%. And to progress past those thresholds, there are various regulatory approvals that are required.
But how to buy a bank is a fun part about understanding different industry norms. How you might go about buying a technology company, or how you might go about buying the local printer, might be different than how you go about buying a bank.
In particular, banks can’t tell you they’re for sale.
Turner Novak:
Why not?
Darragh Buckley:
Because you end up with a drain of the loan officers from the bank itself. The folks producing loans within the bank would then be known to be on the market, and that would be problematic for the bank.
Turner Novak:
Because that’s how banks generate money. Generally you make a spread on the interest in lending out the capital to people, generating loans.
Darragh Buckley:
Very much so. It’s an asset for the bank. If you’re looking for a good first-pass model of the bank, it’s a lender. Don’t worry about the deposit side from the first pass.
Turner Novak:
Super simple.
Darragh Buckley:
So you can imagine now, if you’re in this situation where you can’t tell anyone that your bank is trying to be sold, what do you do? And so the industry has this humorously named conference called Acquire or Be Acquired. Let’s allow ambiguity. Oh, you might be here to acquire.
Turner Novak:
Yeah, we’re definitely here to acquire. We are not on the market. So everyone shows up. It’s kind of like one of those things, like every VC says they’re top decile or whatever. Every founder says, “Oh, I’m not raising right now.” Every bank that goes to this conference, “Oh, we’re definitely trying to acquire other people.”
Darragh Buckley:
“We’re not trying to be acquired.” And you can imagine, entering a new culture, you’re always going to wonder which social faux pas you’re going to make. How am I going to accidentally look the neophyte novice here?
Turner Novak:
Did you have to wear a suit to this thing? Was that the first time you’d ever worn a suit?
Darragh Buckley:
I’d actually worked in investment banking before, so I had the suit in the closet. But understanding these norms around industry acquisition takes a bit. Typically you end up working through investment bankers, and the investment bankers are the ones who know who’s looking in which direction.
Originally, back in 2023, I’d pursued buying another Washington State chartered bank called Washington Business Bank. We had gotten pretty far in the regulatory approval process, and the message back was, “Listen, you’re new to this industry. You seem nice and all, everything’s together, we’d like to encourage you.”
Turner Novak:
“You should sit on a bank board for a while, as a means of us getting to know each other.”
Darragh Buckley:
And so at that stage, I started sitting on bank boards. The kind of learning that you can get from being on the governance layer and seeing what happens is fantastic.
Turner Novak:
What all did you learn from that?
Darragh Buckley:
You’re learning things as broad and wide as, this is what you need to be looking at within your lending pipeline, and this is how long these deals are going to take. This is what we’re specializing in, and explicitly we are not specializing in these other things, and we’ll walk away from all of these.
Turner Novak:
So, like, we lend to farmers, or we do not lend to farmers?
Darragh Buckley:
You will have a specialization.
Turner Novak:
And it’s dangerous to wander too far outside your specialization?
Darragh Buckley:
Yeah.
Turner Novak:
Why is that? Because when you want to grow and diversify, it kind of sounds riskier technically to be really focused.
Darragh Buckley:
Oh, concentration can be a thing, and you need to watch out for that. What I’m saying is, don’t accidentally wander, as opposed to don’t wander. Choose your directions carefully and see where you have a differential advantage based on what you’re currently good at. So purposefully wander, but don’t just chase the next fad.
Turner Novak:
Does that happen too often, you think?
Darragh Buckley:
Yes, would be the short answer. We ended up with a bunch of banks in the 2010 timeframe, for example, chasing financial technology partnerships. A lot of community banks around the nation saw it as a growth mechanism and didn’t have the appropriate capability or capacity to be able to do so safely and soundly.
And so we ended up with a regulatory pushback against that, and I think quite a lot of that regulatory pushback was appropriate. I think there was skepticism that many of the smaller community banks, especially, knew what they were getting into. And it was also one of the reasons it took us so long to build out Increase Bank.
Turner Novak:
The acquisition closed north of a year ago, and it takes quite a while to put in place the correct capability and capacity. You can’t just slap an API on it and we’re good to go? What’s so hard about that?
Darragh Buckley:
Is the question whether practically you can, or whether safely you can? Because practically, you can certainly strap a jet engine to a go-kart. You will promptly find a wall.
Turner Novak:
So what’s the first wall that you’d be running into? If you strapped a jet engine to a go-kart, or if you exposed a bank by API? Maybe they’re the same thing, I don’t know. I’m assuming you built this API and pretty quickly were like, “All right, we’re good.” Was there anything you had to adjust for?
Darragh Buckley:
Yeah. I think this is where you get into the money movement side, which is very, very important. It’s important to get all of the details and the reliability and the transparency correct there. But how that interacts with compliance regimes is fantastically important.
And it’s very useful if that’s not a post-hoc tag-on. It’s very useful if you’re able to embed the compliance regimes as code early on, and that requires engineers being comfortable reading the source docs for the compliance regimes.
It also requires compliance professionals and experienced practitioners being comfortable jumping into the engineering details. If, for example, there’s a compliance regime that has to do with the data elements that need to traverse when you’re sending a wire, and they apply at a dollar amount threshold of $3,000, it’s called the travel rule.
And it’s very useful if you build that in from the start into your wire processing system, as opposed to when you’re going through your first audit, realizing that you need it and then tagging it on.
Turner Novak:
So it’s almost like you need to make sure that when someone goes to enter in that wire, there’s almost an extra box. “Oh, by the way, we just need you to fill this in also.” And anything else that’s going to get triggered is just part of the product in a way?
Darragh Buckley:
Indeed. If you’re able to build it into the product early, you can make life nice for yourself. There’s this time during an examination, and you may have had this experience yourself with respect to any audits or tests.
You get your list of data requirements. Can you pull these data elements? We’re going to sample from them. But can you pull them all for the time being, so that we’re going to tell? And even though you know that you’ve been doing everything correctly, even though you know that your processes are correct and you’ve got good organizational hygiene with respect to those things, when you see the data correct, there’s this sigh of relief.
Like, great, it all worked. And I think you should try to make that sigh of relief asymptote to zero. That sigh of relief should be as small as possible, and the way you can get there is by running the examination on yourself more frequently.
In much the same way, when we were building engineering products, we’d build a test battery around it. So there’s never any question when you’re rolling out the next version of the API whether you’re correctly capturing that parameter, because you know that it wouldn’t have deployed but for that test having passed. And if you can do the same with respect to compliance regimes, you can help yourself a lot.
Turner Novak:
And the way you get good at doing hard things is by doing them more often. This was kind of a lesson from engineering: how do you get good at the hard thing of rolling out a new version? You do it dozens of times a day. Whatever you do, don’t have any manual step in there.
Darragh Buckley:
And if you’re able to apply the same concept to compliance regimes, how do you get good at doing examinations? You do it on yourself multiple times a day. You run all of that same logic yourself multiple times a day.
Turner Novak:
That seemed kind of way too obvious that it was not a normal kind of thing. Because it just seems like, oh, it looks like one of our processes that we’ve built doesn’t really match what we should be doing. We should probably change it. Like any solution, when you look at it from the correct angle, you’re like, “Ah, that’s it. Cool, cool.” And that is the benefit of being able to look at things from multiple angles.
So one thing you mentioned was you talk directly to the Fed, to the Federal Reserve. I don’t even know how that works. I thought that... doesn’t every bank kind of do that? What’s the importance of that?
Darragh Buckley:
Yeah, in ways. Stripe built on top of Wells Fargo, and the way that Stripe would interact with Wells Fargo is Stripe would send Wells Fargo the messages that Wells Fargo was to send to the Federal Reserve. And if it gets passed through in high fidelity, all is good. If some things change in between, it’s very useful to know what changed.
Turner Novak:
So then you can adjust it to make sure you’re right in the future?
Darragh Buckley:
Correct. Or it helps you understand how things are going to fail. So Stripe was not talking directly to the Federal Reserve there, and that meant it was difficult for Stripe to expose that to others. If Stripe really wanted to expose some data from wires, and Stripe did not receive that from Wells Fargo, it fundamentally couldn’t expose it. It wasn’t Stripe’s gift to give unless it had been given that gift.
Someone else’s abstraction places a limit on what you can do, and that can be frustrating from a product standpoint. So Increase’s role there was to step down and talk to the Federal Reserve directly. The Federal Reserve is also a fantastically reliable and good service provider. I think this was one of the unknowns for me when starting Increase: how good is the fundamental layer?
Turner Novak:
Yeah. If you think, man, that thing was started like two centuries ago, you kind of question it. But...
Darragh Buckley:
Well, the Federal Reserve is later than that. The Federal Reserve is at like 100 years.
Turner Novak:
What year was it, 1907?
Darragh Buckley:
The Federal Reserve is many things. I don’t just mean at its regional banks level. I also mean that it’s got this division called Federal Reserve Services, which is its technology layer. So central bank, bank regulator, and technology provider are not fundamentally components that need to be placed together, but we have chosen to do so as a nation in calling them the Federal Reserve.
So Federal Reserve Services is the technology layer that we end up chatting with a bunch. And they have good technology.
Turner Novak:
And they’ve got fantastic technology?
Darragh Buckley:
And also they’re very good operators. The fun fact here is, about a year and a half ago, the Federal Reserve ran an update to its messaging specification to do with wires. The Fedwire system bumped over to an international standard called ISO 20022. No one knows about this outside of a niche group of interested people. But it required coordination across the financial institutions of the nation.
Turner Novak:
So it’s basically like at one moment in time, everything needed to be adopted for the new methodology?
Darragh Buckley:
Sunday night. Yep.
Turner Novak:
You’ve seen enough technology companies, right? So if anyone described a technology transition like that to you, you likely would do the same thing I do, which is head in hands, like, “Oh, please don’t try to do that. That will go horribly.”
Darragh Buckley:
Yeah, I mean, it’s kind of like Y2K. They were like, “The world is going to end at midnight when the calendar rolls over.” And the Federal Reserve was under constraints, and this was the one that made the most sense, that it was going to be a hard cutover. But they successfully operated that hard cutover through just good planning, good testing, and good operations for months and months prior.
And success here is defined as, you didn’t know about this thing. You didn’t read about it on the front page of the paper that wires across the nation stopped. So the fundamental underlying of the US depository networks is in fact great. We have real-time payments there.
Turner Novak:
We’re missing the last mile delivery of that, right? We’re missing the productization part, or the ability to productize that. So what was missing in that last mile?
Darragh Buckley:
Yeah, high-fidelity pass-through of the data and the capability.
Turner Novak:
So there’d be someone who’s sitting with all the technology and they have access to it all, but then they don’t expose it correctly?
Darragh Buckley:
Exactly. On the button, Ben.
Turner Novak:
Why don’t they just expose it correctly?
Darragh Buckley:
Oh, exposing it correctly isn’t trivial. This is the case where I’ll differentiate between product at Stripe and product at Increase. Product at Stripe is creating these beautiful abstractions. You’re a developer, you care about payments. You do not care about ACH. You do not care about Fedwire. You want this abstraction.
Product at Increase is, oh, this is the FedACH API. You want to know what that parameter is called? It’s called the exact same value as the network calls it. We do not try to be creative and create an abstraction there, because you want high-fidelity pass-through of that data and information. And that’s what I wanted while building Stripe.
Turner Novak:
So you’re almost building... the Increase product is built for the Stripes of the world, and Stripe is built for developers?
Darragh Buckley:
Very much so. Our users and use cases are those technology companies building financial products for others.
Turner Novak:
So I’m curious then, how did this go when Stripe’s a customer? You might have an inside track with Stripe. But with some customers, they may have already been using a product, or they built it all internally, and then you go to them and say, “Hey, you have this whole team, switch to Increase.” How does that go, convincing a lot of these companies to switch over?
Darragh Buckley:
Oftentimes there’s a moment in a sales conversation. Usually it’s someone from product or engineering who is the person interested in picking up Increase. And often there’s a moment in that conversation where they’re like, “Sure, I get that you have these APIs. But do you pass through the IMAD, the Input Message Accountability Data? And when do I get that?”
And you’re like, “Yep, you get that here and in this way.” And after three or four of these skeptical questions, there’s this sigh of relief and dropping of shoulders, which is just like, “Oh, I’m with my people.”
“I found my place. Great. I’m glad that one of us went ahead and did this.” I’m glad that one of us went ahead and dropped down to the base level and surfaced it all with no abstractions. Those are actually pretty fun conversations. Oftentimes there’s a high degree of shared experience.
Turner Novak:
Yeah, actually, the founder I was mentioning earlier that uses Increase, I know him, and he was telling me... I was like, “Oh, Increase, I don’t know if I’ve heard of them before. Why’d you go with those guys? Because there are other solutions on the market.” And he was like, “I don’t know. I felt like they were engineers, and everyone else I talked to was like a salesperson.”
Darragh Buckley:
Yeah.
Turner Novak:
And the Increase sales guy, I have no idea. I guess he talked to you. But he just felt like he understood the problem and understood the technology behind it. And so that’s why I went with them. I was like, “Oh, interesting.”
Darragh Buckley:
Increase is about half engineers. And it’s very, very important to have not just engineering within Increase, but I think everyone takes the same approach. What we strive for is kind of friendly competence. We are willing and capable to talk about the details with everyone.
You want to talk about the ISO 8583 standard for the card network, and you care about this parameter that’s in there?
Turner Novak:
Yep, I know why you care about that. That’s similar to the life that I’ve lived, of wanting to build that into my own product. Like, you probably want it in this way.
Darragh Buckley:
Yeah, great, yes. The intention of Increase is that you’re composing it into your own product.
Turner Novak:
And I had recently, I think it will have come out before this episode goes live, but I guess a peek preview of the future if not. I had Ron Gabrisko, who’s the CRO of Databricks, and he leads their sales team. He joined before they had a million in revenue, and I think they just announced they’re doing $7 billion in run rate or something. Maybe it’s bigger than that, I can’t even remember the number.
But he built the sales team there. And I think their thing is their sales team is very technical. They are not traditional, quote-unquote “salespeople.” They are people who understand the problems that you’re facing.
They understand the technology behind it, and then, how do they solve the problems that you have? And I know that’s always been my favorite way to think about sales. You grow up thinking, ah, sales guy, snake oil salesman, whatever, shoving something that they don’t need in front of them.
But it’s really, somebody has a problem, and if you can solve it, they’ll pay you money for it. So how can I solve your problems? That’s really how you should be doing it at the end of the day.
Darragh Buckley:
Yeah. And there are different modes of salespeople. I very much enjoy the salesperson who can be in a challenger sales model.
Turner Novak:
Oh, so what’s a challenger sales model?
Darragh Buckley:
My understanding, and it’s important to understand that this is not my own skill set, but it’s listening, exploring, sitting with the problem, but then being comfortable saying, “I think you’re doing it wrong. I think that you should be doing this instead.”
“I hear what you’re saying. I’m trying to understand the constraints that are not obvious from the start, but I’ve asked enough questions around it. No, you should do it this way. You should hold the product this way instead.” And it can be very trust-building.
It requires a lot of empathy on behalf of the salesperson. It requires a lot of technical competence on behalf of the salesperson, a lot of understanding. But then, having a trusted third party who’s familiar with the bank side being able to say, “No, you should probably do it this way,” is valuable. You’re not trying to be infinitely flexible. It’s okay that there’s a correct way to do things.
Turner Novak:
So what does a good Increase customer look like, or the average customer? Because the guy we were just talking about is just starting his company, he’s super small. You also have, I mean, I could say Stripe is basically at the scale of a public company. So what does a customer look like of Increase?
Darragh Buckley:
Ours very much tends toward the larger use cases, as opposed to the earlier use cases. I certainly hold many of the early use cases near and dear to my heart and want to continue supporting those. But it’s more for the at-scale use cases like payroll and bill payment, when you’re moving billions of dollars and it is very important to you that you’re able to do it with a high degree of reliability and a high degree of transparency. That’s where we do best.
Turner Novak:
So a lot of the customers you get, they’re probably already doing some of this stuff or using someone else. You’ve got to get them to switch over, either switch over, or if they’re launching a new feature that’s adjacent to something they already have.
Darragh Buckley:
If I was to give a recent example, Ramp’s bill pay is a great example. We actually have a case study up about this. Ramp uses us to pull in money from the payer and push out money to the payee, and we are their means of interacting with the Federal Reserve for that.
In particular, you originate ACH debits, the US mechanism for pulling money from someone’s account, and then how you do the funding to the payee, be it by ACH credit or Fedwire, or even checks, or more modern mechanisms like real-time payments or FedNow, or push to card, and being able to combine all of those nicely within one API. That can be valuable.
Turner Novak:
One thing I think is super interesting about Increase, you mentioned you’re doing over half a trillion in volume, and you have, according to LinkedIn, 33 employees. I have no idea how accurate LinkedIn is, but that’s a pretty high employee-to-volume-moved ratio.
Darragh Buckley:
Yeah. The point to emphasize beforehand is I would not do so with that amount of people unless it was safe to do so with that amount of people. But, preamble aside, it’s fascinating how much you can scale when you’re just building on top of the Federal Reserve, when you don’t have those abstractions.
A lot of the other operational team stuff in the general financial services space is adjusting for the abstractions. Figuring out how to get around them, or figuring out what happens when things leak through. An example from Stripe: ACH reconciliation, or ACH return allocation. You send a transfer, the transfer is not able to arrive at the person’s account, it gets bumped back to you from the receiving bank, and now you need to know which transfer got bumped back to you. Sounds like a trivial problem, right?
Turner Novak:
But of course, how could money just arrive back to you without you knowing why? It would of course be labeled, probably, right?
Darragh Buckley:
Yeah, of course. But it’s not labeled. The primary identifier, the one that you’re meant to use, is called a trace ID. And the trace ID is a numeric string of 15 characters. You don’t get all of those 15 characters, because some of that embeds the routing number, so you’ve got less entropy than you might expect. And so you’ll end up repeating these at pretty high frequency. They’re not actually a unique identifier for you.
And so you’ll end up also building heuristics. Okay, it was this transfer with this ID and to this account and routing number. Great, that allows me to uniquely identify it. And so you’ll build your heuristic mapping of this. That’s just fine. Stripe didn’t have that trace ID because of some systems issues with how Stripe interacted with Wells Fargo.
No fundamental fault on either group, but it meant we ended up using the heuristics very early on. And of course, then there are things that manually fall out of that heuristic queue, and so you build a team that’s responsible for those allocations.
Turner Novak:
So you’re kind of automating and making software to make up for the lack of capabilities in the original software.
Darragh Buckley:
Indeed. And that’s a reasonable thing to do, to a point. And at that point, the correct thing to do is go back and fix the fundamental problem. And hence, Increase.
Turner Novak:
So then how do you think about hiring and adding to the team? There aren’t very many people. How have you been doing that?
Darragh Buckley:
Yeah. I enjoy working in high-empowerment cultures, and I enjoy working with people who enjoy working in high-empowerment cultures. And for me, high empowerment means the ability to wander the world and affect the change that you need to, to build the product.
There are prerequisites for that. You’ve got to hire well, and you also have to trust the people who you hire, and trust that they trust you to go off and effect that change. It does mean that it’s very useful not to silo too strictly. You have to allow a compliance person to wander into engineering and an engineering person to wander into compliance, and more than just allow, you promote and expect them to do so too.
Stripe was very good at this for very many years. Stripe maintained it to a higher headcount than Facebook maintained it to. Facebook maintained it to a higher headcount than Google maintained it to. Google maintained it to a higher headcount than Microsoft maintained it to. So I would aspire and hope that Increase will maintain it to an even larger headcount.
Turner Novak:
But I think it’s what builds good product. It’s what builds effective organizations. One thing we talk about now, going back to Stripe. So you met the Collisons back in Ireland?
Darragh Buckley:
Yeah, we grew up in the same hometown.
Turner Novak:
This is like Limerick?
Darragh Buckley:
In Limerick, yes.
Turner Novak:
Is that a big city?
Darragh Buckley:
Limerick’s not all that big. At the time it was maybe 35,000 people, depending on where you draw the city lines.
Turner Novak:
So for the US, it’s not that big.
Darragh Buckley:
Correct. But it has a very forward-looking university, which I think is very influential on all of us, and it has a whole bunch of multinational companies, and that creates a different window of what’s possible. So we grew up in the same hometown, and spent more time together while at college and university in the US.
And then I ended up joining up with him in about 2010, as Stripe’s first employee.
Turner Novak:
You were kind of the banking guy, like they were, “Deal with the banks”?
Darragh Buckley:
Indeed. I’d worked for Lehman Brothers prior.
Turner Novak:
Oh, shame. Okay. So you had first-hand experience with the banking system.
Darragh Buckley:
Well, I expected that this is where I will show how early Stripe was in its journey with respect to what is a bank. Counting, you know, oh, you worked for an investment bank, surely that is a bank. And, oh, sweet summer child, no. It reminds me of the story.
Turner Novak:
I had Will Gaybrick on the podcast, and he told me a story about how one day Patrick texted him. I forget exactly the conversation, but it was like, “You know finance. Why don’t you be our CFO or something? Here’s this broad regime.”
Darragh Buckley:
John and Patrick, they have many superpowers. One in particular is learning. Their own ability to learn, who they learn from, how they learn. It’s fantastic.
Turner Novak:
And so you’re like, “Oh, this sounds like a good idea. I’ll go be the banking guy at this Stripe company. We, the Irish superheroes, coming over to fix payments in the US.”
Darragh Buckley:
Wouldn’t go that far, but yes. I joined up, and it was quickly evident that what was necessary was less banking and more engineering related to banking. And so that was the first thing that I jumped into. Some fun vignettes here might be... you have likely had this experience.
Like where you’re at a group project, and there’d be four people on the group project, and you’re all going to be working on this thing together, and you’re going to get to know each other. And so I arrive into Stripe, and the early group project is, well, it’s John Collison.
There’s Patrick Collison. There’s Greg Brockman. This is a very high-powered group to be able to learn and build with. And I’m very grateful for that experience.
Turner Novak:
So you were like the fourth, the other guy on the team?
Darragh Buckley:
Just the other guy.
Turner Novak:
So what did you learn working with the early Stripe team?
Darragh Buckley:
Yeah. I think one of the most impactful and meaningful for me was the amount that a small, empowered team can get done. If you’ll forgive a story. There was one night, I suppose it was an afternoon, we jumped on a call with a partner, and this was how Stripe was finally going to be able to talk to the credit card network more directly.
And on the call, the person in charge of certification was like, “Okay, great. The next certification date is six weeks. You’ve got a lot of things to do before six weeks, but you should be able to get it done.” And you can imagine, in startup world, when someone says six weeks to you, you’re like, “Wait, what?”
You might as well have just said 10 years from now. This is beyond our time horizon to contemplate. So someone pipes up in the call, “Oh, is there any time sooner?” “Oh, there is tomorrow at 10:00 AM, but you’re not gonna be ready for that.”
Turner Novak:
I just assumed there’s no way.
Darragh Buckley:
Yeah. And so there’s a meaningful look around the table of, “Oh, no. Sign us up for that. Great, we’ll chat to you then.”
Turner Novak:
“Sure, okay.” I’m just assuming you’re coming back in six weeks.
Darragh Buckley:
Well, yeah. “But sure, I’ll humor you.” Hang up the phone, look around. It’s like, okay, we’ve got 18 hours to get this, you know, 150 test cases done. I guess I know what I’m doing. One of us is starting at the top, one of us is starting at the bottom. Go for it. And you just work through the night.
What do you do after test case 36? You do test case 37. And just keep going throughout the night with this group. And it’s fantastic. Shared stress, strife, and struggle is great for building camaraderie. And that’s always fun.
Maybe at about 8:00 in the morning, in order to talk to the certification service, we needed to do a plain old telephone connection. So I had to run down to Fry’s Electronics store in the South Bay as it’s opening. Knock, knock, knock. Please, I need this part for this thing. Great. I’m now jumping back to Palo Alto.
Turner Novak:
Getting the call at 10:00, by maybe 11:30, “Okay, everything’s passed. Never in my years have I seen anyone go through this certification this quickly, but you’re good to go. You can run your first live transactions.”
Darragh Buckley:
Yeah.
Turner Novak:
So just turning on the first live revenue for the company?
Darragh Buckley:
Yes. The sigh of relief of a group that has achieved something that required just a sprint of effort and endurance is great.
Turner Novak:
Overcoming adversity is great. That’s kind of what drives much of the enjoyment of startups. It’s almost like, no one’s ever done this before, but we just did it. And we didn’t even know this was a thing. We didn’t even know this was a hard problem. We just were like, “All right, we’ll just get it done.”
Darragh Buckley:
Yeah, ignorance can be helpful, in certain circumstances. But it’s more to do with the group. Doing these things alone is fun and whatnot, but it’s more when you’re sharing this experience with a group that’s dedicated to empowering each other, dedicated to helping each other learn and grow.
Turner Novak:
Yeah, I value that.
Darragh Buckley:
And so carrying that forward into Increase was very important for me.
Turner Novak:
Is there anything about Stripe maybe that someone from the outside wouldn’t have seen, or wouldn’t have understood, about the reason that Stripe was so successful and the reason that it worked? Myths about the company that are incorrect, or that should be myths and no one actually knows about yet?
Darragh Buckley:
I think this changes over time. And it’s okay that a company’s story changes over time. I think probably what’s now common: a group of smart, nice, hardworking people who are relatively disciplined around thought and how to give support to each other in the correct ways. Those are useful.
Back in the day, differential characteristics were probably things like a long-form written culture of communication. Writing it down. That wasn’t as common. I don’t think it was as common in 2010 as it currently is. I don’t think I would call it a differential characteristic these days.
Turner Novak:
I mean, now you can just throw some prompt in Claude and get some AI slop and share it in the chat.
Darragh Buckley:
Yes, and you would get all of the value for doing that and none of what actually matters. If you’re using it as the thought partner to grow and develop yourself, yes. But the reason you had long-form writing was because it was helping you learn.
Turner Novak:
It was forcing you to put your ideas down in a legible way.
Darragh Buckley:
Falsifiable statements. And you were going to have to expose all of your logic to someone else, and someone else was going to be able to point out the parts that were more or less shaky.
Turner Novak:
And it’s not like you’re on thesis defense territory. You’re not being attacked for this thing. This is a group project of us all trying to come to the correct answer. But that’s a lot easier when there’s a culture artifact to look at.
Darragh Buckley:
Well, sure. But then it has a culture of everyone looking at the artifact together, and using the artifact to get to a point of learning or truth at the end. And you’re all there in support of the person who’s going to have to own the decision, and you’re going to get behind this decision this person makes.
Turner Novak:
So it’s like working together to get to the right decision.
Darragh Buckley:
Very much so. But the particulars of all of this matter. I feel like there’s nothing objectionable about that statement that we’ve made, and it’s probably not all that interesting until you get to the details.
Turner Novak:
Anything specifically that you learned from Patrick and John?
Darragh Buckley:
Yeah. There’s this fun story, forgive me for storytelling again. That’s more interesting to people, I think. So, as is the tradition of all startups who first get their office, you arrive into the office and the place is empty.
You’re like, “Great, what are we going to do?” IKEA trip. Jump in the car to IKEA, get six of the cheapest desks and cheapest chairs that you can get, put them back in the car, drive back to the office. You’re unloading the car. Pick up a box, put it down, you walk back to the car. That’s not what John and Patrick do.
They pick up the box, walk into the office, put it down, and sprint. And you, as observer, are like, did we park illegally? Did I miss something? Why are you running? They just always have this sense of urgency. There’s always a, “Oh, we can just do it faster.”
Turner Novak:
“Why would we not just do it faster?” Just run.
Darragh Buckley:
We run. And if you’re in an environment where someone else is running, are you going to be the one person who’s not running? Of course not. You’re going to run too. And not in an obnoxious way. Not in a, “I’ll race you, and if you’re slower than me, I’ll rub it in your face.”
Turner Novak:
It very much is in a, “No, this is just all of us here together doing this thing.” You just gotta get there faster. You can probably create an extra, if you’re doing a lot of traveling and movement in the day, the faster you move, the more time you save.
Darragh Buckley:
Well, yes. But I think that cultural prevalence of the importance of urgency, John and Patrick are fantastic at creating. And it’s not an obnoxious, “Why haven’t you done it yet?” It’s just like, “Oh, why wouldn’t we just do it now?” Given in a very friendly, pleasant, but boy, you’re going to want to keep moving, way.
Turner Novak:
And they’re very, very good at doing this nicely. Yeah, I think Patrick has a blog post about... I think it’s just called Fast or something. It’s like his name dot com slash fast, and it’s literally just a list of bullet points about how the Empire State Building got built really quickly.
Darragh Buckley:
Yeah. At times our expectations around speed are just misset, and we can in fact, when you have the right group of people with the right resources, and the right things out of their way, you can do things at unreasonably fast speeds.
Turner Novak:
One thing, going back to the podcast I had with Will when he came on. He said an early thing that you guys learned at Stripe was... it was a really high-up executive or CEO in financial services, and he didn’t actually tell me who it was. Maybe he did, I can’t remember. But something that they told you was, “You guys look all gung ho, like, ‘We’re building a payments company.’”
And he’s like, “Oh, you’re not building a payments company. You’re actually in the fraud business,” or something to that effect. I feel like Will actually told me, but I forget, and I’m interested in hearing it again. So why is it ultimately you’re not actually in the payments business, the money movement business? You’re actually in the fraud business. How does that work?
Darragh Buckley:
Yeah. I think, as you can imagine, cards in particular, but payments generally, are magnets for bad actors and in particular fraud. There’s a lot of money to be extracted.
Turner Novak:
I get emails every day with a fake invoice from some random company. And I’m like, “I guess some people get caught by this.”
Darragh Buckley:
In very unfortunate ways. And there are many things that we can do as an industry to reduce that.
Turner Novak:
Why has it been so hard to solve the fraud problem?
Darragh Buckley:
If you’re able to tell me how to properly solve the identity problem, I’ll be able to tell you how to solve the fraud problem. And I’m not meaning to unfairly throw it back at you. I’m just saying that we haven’t correctly solved online identity, with various versions of it, to reasonable levels of confidence. So the fundamental problem of fraud in payments is identity.
Turner Novak:
It’s identity. You’re pretending to be someone else at the end of the day.
Darragh Buckley:
At times. Now, there are other versions, which I wouldn’t actually chuck in the... or you may or may not count as fraud, like first-party fraud. You received a good and you pretended like you hadn’t received the good.
Turner Novak:
That’s not solved by identity.
Darragh Buckley:
But something that rounds to a correct definition, saying if you solve identity for me, you will vastly solve fraud for me, is correct.
Turner Novak:
And I know one of the things you did really early on was you would literally just manually look at transactions and watch them come in. What did you get from that?
Darragh Buckley:
Yeah. I think this is advice to any payment company who’s starting. You can try to design all of your fraud systems in the absence of data, and try to think deeply around it. But know that you’ll miss it.
The far more important thing is just watching the initial activity. And by initial activity I mean it’s not unreasonable to expect a person at an early-stage payments company to look at every transaction, up until there are like 5,000 transactions a day. You kind of have to do that. I don’t know how to build good risk systems without that.
You do all of the thinking beforehand, of course. Make sure you’re collecting all of the data, of course. But you’re not even going to know what the patterns look like until you start seeing the patterns. And you have a pretty effective neural net between your ears. You should go use it.
Turner Novak:
Does AI and LLMs kind of help with this nowadays?
Darragh Buckley:
I can imagine that there’s quite a lot that can help. I’d be surprised if you still didn’t want to do the person-sitting-in-front thing, saying, “This trucking company should not in fact be receiving this mortgage escrow payment. Something’s up.”
Turner Novak:
I think Will actually told me that Stripe has built their own kind of model internally, where when certain entities make a transaction, you can tell based on learnings over time where they might be grouped, and more chances of these happening. I’ll link a spot to click on to check out that part of the episode, because I’ll probably butcher the way exactly it’s done. But it was pretty interesting. Just like, oh yeah, I guess that does actually make sense.
You can essentially... the models will know, it’s probably not perfect, but they’ll start to pick up on these behaviors and things like this.
Darragh Buckley:
Yeah. Ghostbusters are fun. Like anything within an early company, it can be tempting to overbuild in that particular area. And it can be useful if, instead of coming up with an intricate data pipeline analysis tool, you just look at them at the start.
Turner Novak:
And one thing I think you like to do is you like to give cards to your agents. How does that go? Because a lot of people are kind of like, “Oh, it’s agentic payments.” So how do you do this practically? What do you do?
Darragh Buckley:
Yeah. I feel like I’m particularly grateful that I’m able to do that, perhaps not everyone is able to do yet. You have, similar to others, agents wandering the world doing things for me, but I want them to be able to economically interact. I want them to be able to pay.
Turner Novak:
What are they paying for right now?
Darragh Buckley:
I am currently traveling to San Francisco. I likely need to send a set of flowers home, thanking my wife for being with the kids today. And being able to set the agent upon that task is great.
Turner Novak:
So what are the limitations of that? If an agent goes to the local flower shop, jumps on their site, is able to stack up the checkout, etc., it would be useful if they’re able to complete that. But in order to complete that, they’ve got to have a credit card, right? They don’t need to be able to interact with crypto, right?
Darragh Buckley:
The flower shop already has a checkout. They need to be able to enter the 16 digits of the card on the site. Now I’m able to give my agents payment cards. My agents have their own bank account. And if they need more money, they ask me for more money. “I went through the $100 you gave me. Can I have more, please?”
And so you give them $100 more. And a difference here that I’m pointing toward is I’m not sitting within the synchronous approval process for a transaction. The point is that I’m not thinking about it. It selected what it was buying. It selected the instant coffee from the local grocery that it’s going to send me.
It would be broken for me if I needed to go in and enter the card details. Likely what would happen there is some sort of timeout, because I didn’t get to it quickly enough, and we’d end up in various loops to try to synchronize our schedules. But having the agent have an account, having the agent have a payment card, is great. And the control plane on that for me, the thing that stops it just burning a bunch of money, is it can only spend what’s in its account, and it asks for more money when it needs more money.
Turner Novak:
It sounds like it’s almost like when you have an employee that has a charge card. You can only spend this, these are the rules of what you can spend it on. It’s almost like applying the same principles to an agent.
Darragh Buckley:
Indeed. And doing things, this is what it can and can’t do. There are rules.
Turner Novak:
Yeah.
Darragh Buckley:
And in particular, what I’m pointing toward there is we don’t perhaps need to rebuild all of the checkouts across the internet. We don’t need to plug a bunch of crypto or stablecoin connections into these. It’s okay that we use the existing ones.
Societally, I hope we run both of these experiments. It’d be great to know which one plays out. But for the immediate use case, it’s very useful that we use the card with its known dispute process.
Turner Novak:
Do you have a favorite founder or historical figure that you’ve gotten a lot of inspiration from?
Darragh Buckley:
Oh, fascinating. I think this is going to vary based on different characteristics in different places.
Turner Novak:
Yeah, I mean, there can be 100 of these, essentially.
Darragh Buckley:
Sure. And I think in particular what you’re trying to do there is not overly romanticize the person, but that particular characteristic that you really liked from them. I’ve gotten some off-the-record answers to this question, like a dictator, and it’s because of, like, not what you would think. But it was kind of interesting. I won’t say any of the names.
Turner Novak:
It would be unfair for me to probe deeper on that particular question. But it’s kind of interesting. Like, with being fast and urgency, oh, I can apply that principle to myself. I’m wondering if there’s any non-intuitive, or maybe it’s super obvious, answers to this question.
Darragh Buckley:
Yeah. You know the way when you end up running a small mental model of someone in your head? You’re like, “What would this person do in this situation?” I feel like that’s always a sign of admiration. You’ve carried around this person as a heuristic for you.
One related to engineering, for me, if I’m allowed to go that way, and you can tell me my hobbies are too boring. Tom Brown, who’s one of the Anthropic founders, previously at OpenAI, has this remarkably impactful technical essay from early on, of just use boring technology widely. Don’t try to over-optimize for the correct tool for the correct job. Very much benefit from using a broad tool that can do many things.
And what you’re pointing toward there is the price of complexity. When you’re using the correct tool in the correct place, now you have many correct tools, and your ability to understand the correct tools is hard. So just be very careful about your complexity cost of what it is that you’re doing. I reach for that one maybe daily.
It’s okay to do something in the wholly uneventful way. It doesn’t need to seem bright or clever. It just needs to seem maintainable.
Turner Novak:
It’s almost like there’s sometimes the glamour of building and trying some new hard thing. There’s more social status internally. You get more joy from, “I figured out this really complicated thing,” versus sometimes it’s the beauty of the simplicity of not actually complicating it.
Darragh Buckley:
Don’t get me wrong, I can admire arduous intellectual pursuits. It’s okay that those exist. But we can do that by answering the hard crossword on a Sunday, not by obligating our team to maintain a thing for years. Choosing the uneventful thing, and I especially think this is true in banking.
Imagine if your friend is an air traffic controller, and they show up to meet you for coffee, and they’re saying, “I’ve had the most exciting day at work.” You’d be like, “No, I don’t want excitement from my air traffic controllers.” Like, “It was so boring. Nothing happened.” I will make the statement, I think you want the same from your banker.
I think you want your banker to have an uneventful day. I think you want your banker to, in fact, enjoy uneventful days. There to be a quiet satisfaction of that reliability, regularly getting the things done as expected. I think that personality works well in banking.
Turner Novak:
Yeah, because when I think about what are the biggest events in banking, they’re generally not something that you want to see happening on a daily basis. In much the same way that the underlying utilities of society, electricity or water, it’s okay that these things are uneventful.
Darragh Buckley:
It’s in fact good that they’re uneventful. It’s like any type of infrastructure, you want no events. It just works, as it always does, all the time. And operators and builders and maintainers who value that can be well suited.
Turner Novak:
Well, it’s been a lot of fun. Thanks for coming on the show. It’s been nice, I very much enjoyed it.
Darragh Buckley:
Thank you very much for having me.
Find transcripts of all other episodes here.