Triplebyte is fighting credentialism in software engineering
A conversation with newly named CEO, Ammon Bartram
I sat down with the founder and newly named CEO of Triplebyte, Ammon Bartram, to discuss the mission and how Triplebyte is bringing about a more fair, data-backed credential that is more representative of skills.
This turns out to be a topic near and dear to Ammon’s heart simply because he lived this problem himself— being homeschooled meant he got rejected from every university he applied to. When he did graduate, he was nearly shut out of highly paid software engineering positions because the whole industry over-relies on resumes.
Sitting down with Ammon, it was clear to me this was one of the largest value-unlocking startups in the world, just lying in plain sight.
Transcript below, but you can also watch the full video here:
Garry Tan, cofounder and managing partner at Initialized: Triplebyte is a treasure in Silicon Valley for helping people get amazing software engineering jobs, but without any special credential beyond your raw skill. That’s so powerful. Why did you start this and how did you get started in Silicon Valley?
Ammon Bartram, cofounder and CEO of Triplebyte: Those things are very connected for me. My background as a software engineer is self-taught. I, like a lot of people. I started programming when I was a teenager and kind of got involved with programming and demo seeing it online.
Garry: Yeah, demo scene. That’s awesome. Being into computers when we were young, that’s like winning the lottery ticket and not even really understanding that.
Ammon: Yeah, I didn’t understand it. My parents didn’t understand it. I’d recently gotten into programing with spending lots of my time on the computer working on these games and I remember my parents sitting me down and saying, “You know Ammon, we like that you’re interested in this, but we think you’re spending too much time programming.”
Garry: “We’re worried about you.”
Ammon: “We’re worried about you.”
Garry: I got that, too.
Ammon: We need you to study some useful life skills.
Garry: Turns out this is the most useful thing that you could possibly be into.
Ammon: That’s where the connection back to the Triplebyte is for me. It is — some people are self-taught — so many programmers, there’s this hurdle to actually get a good job. Credentials end up playing a pretty large role and so you know, I was self-taught, I was actually homeschooled, so I wasn’t going to school.
Ammon: I was being taught by my parents and spending all my time reading and teaching myself. That was a big childhood. I loved it, but it made it very hard for me to get to university. I had just none of the background, all I had was my parents.
Garry: Where was this again?
Ammon: On a farm in Western New York state and that was, I think, my first encounter with this kind of credentials and the role that played, was trying to get to college. The first year I applied to colleges, I’m embarrassed to say, I was rejected from every school I applied to and couldn’t go to college. That was hard to take at 18 years old. I had to work really hard and apply to a bunch of the least selective schools.
And I actually got in, so I went to a very low ranked school in New York state, and I did well. I really enjoyed school. Even at lower ranked schools, there are really smart people, like the professors who were really passionate about their areas. I took programming courses.
Garry: You probably got amazing, very direct access to the best professors that way. You weren’t one of thousands of people, the sea awash with people.
Ammon: And actually the C.S. Department at my school, I think had a total of 22 students enrolled at the time. Many of my classes were five students, so I had good access to the professors. My experience became much, much worse when I graduated. I remember my school had a career fair. I think the only company attending that had any claim to be a tech company was a, sort of, outsource tech support call center that was located…
Garry: So the big guys didn’t show up, the Microsofts and the Googles didn’t show up.
Ammon: I don’t believe there was a single company hiring engineer’s at the career fair. You know, they were hiring tech support people to get on the phone to help people fix their computers.
Garry: And yet there were very smart people, you, your classmates who were very skilled, and could program anything.
Ammon: Yeah. And many of whom… I have this career in Silicon Valley, and some of my classmates I know work at NASA and other startups.
Garry: So the credentialing system is actually failing people.
Ammon: It is. What really hit home for me during my job search process. So post graduating I was then, I had to get a job and I went home living in my parents attic and applying to companies and it just, months, I would send out resume after resume, just not hear back.
Ammon: Because you know my resume, I didn’t look very good. All I had, I had four years out of a school that no one had heard of. That’s the filter that people use.
Garry: This credentialing process is just clearly broken in some fundamental way for really smart people.
Ammon: What’s so insidious about it, is that it’s not irrational. So then the flip side of this is I eventually got lucky and I got a job at Twitch when it was a small startup, then called justin.tv. I got a role there because they were small enough to not really have a team of recruiters and have an application process. And they had a programming challenge. You could do their challenge and if you completed the challenge, they would interview you.
Garry: If you were great, you’re great.
Ammon: So I solved that. Two weeks later I got this one-line email from an intern there actually inviting me in to interview. I got the job, got my foot in the door. What really paints the whole picture for me, is kind of actually what I experienced next. Right?
Ammon: Once I was employed there, pretty quickly I was leading a team and hiring people under me, experiencing the flip side, right? And realizing that it’s easy to say credentials are stupid, people shouldn’t look at them, but when you’re hiring, it’s more complicated than that. It’s really expensive for your team to bring somebody in. Credentials do correlate. Right? What’s so insidious about this problem is that…
Garry: Right, you’re trying to skip ahead because before Triplebyte, before an objective measurer that’s run by software, there wasn’t really a scalable way to create this credential period.
Ammon: You’re a recruiter or hiring manager, you’re hiring for your team and flying someone in, you bring them for, eight hours of engineering time, whatever it is, it’s extremely expensive and you feel terrible for bringing someone in and they do poorly.
Given those incentives, there’s a reason, people with fancy degrees and internships and past experience at big companies as a group are better, right? It does mean something to have an MIT degree. It’s really hard to get into MIT, world class professors, great, great, great classmates, it’s a great place, that means something. The problem is, that only a tiny fraction of people had gone to MIT. Yes, it means something, but the vast majority of programmers in the United States don’t have those credentials. And so if you have a process that just rules them out, missing tons of people, right? So really high false negative, right.
Garry: Contrast this with what the biggest tech companies constantly say. We can’t hire enough smart people. There aren’t enough smart people out there. This is just clearly false then.
Ammon: Yes, there is a large pool of people who are being passed over and the problem compounds, right? I like thinking about my trajectory. I was extremely lucky and it was blind luck. The only job offer I got was that this tiny startup, I hadn’t heard of Y Combinator, or I didn’t understand the startup space, I got this job and that company ended up growing into Twitch. So now I have this awesome resume, early engineer at Twitch responsible for critical systems, there early on.
Garry: What a ride.
Ammon: But had I received an offer from CVS corporate to work on their employee management systems and wherever CVS is located, I don’t know. I would have taken that. I needed a job and I would have taken that, and I would have poured myself into it. I would’ve probably done a good job. But the result then would’ve been that my resume would have shown, degree from no-name school and then three, four years working at CVS corporate.
I can tell you from working Triplebyte, that resume actually looks worse than… That resume is absolutely going to be ignored by nearly every recruiting team at top companies, so the problem compounds. And there is a real signal there, it’s not irrational.
Garry: But people get tracked…
Ammon: Yes, people get trapped.
Garry: Yeah, they get tracked and they get trapped, actually by that tracking. There’s, I guess, society has this great sort.
Ammon: And of course it correlates with a lot of other things, right? Socioeconomic status, race, gender, your, parents can’t support you through college.
Garry: Yeah, totally. It’s a real thing.
Ammon: It’s that much more likely that you drop out and don’t have that credential and go down that track where you’re then trapped.
Ammon: What drew me to Triplebyte, what led me to start this company was this problem. Having seen both sides of it, it’s relatively obvious what the solution is, right? This is a data problem, right? We know how to solve data problems. The solution is to come up with the… is to directly measure the skills and create a process, a test, an assessment, identify skill wherever it lies.
Ammon: And so we basically start out with that as the foundation of Triplebyte. We are going to conduct lots of interviews, gather data, and build a process that is fair and more accurate and open to people from all sorts of backgrounds.
Garry: And that’s what you’ve done now. How many people have gone through the Triplebyte assessment at this…
Ammon: Somewhere around 200,000 folks have gone through the assessment, and what’s awesome about that is again, this is a data problem and so the more data we have, the more accurate we can make the assessment.
Ammon: Fundamentally, we’re in the business of classifying humans and that is an impossibly hard thing to do, right? Human potential is incredibly complicated. The key observation is that we’re competing against teams of people who are just making gut calls. Right? And the actual research shows that’s pretty close to random. We are still very far from perfect, but our process is de-buggable. When there’s a problem in our process, we can go in and say, “Oh, this was incorrect. We’re going to change our model to fix this problem.” And, it gets better over time as more candidates and more companies use our product.
Garry: And this is separate from the AI efforts that some of the bigger companies, like Amazon, have done. You’re not using past data, or past resumes, or past background, to predict whether or not someone will actually get hired, simply because that’s institutionalizing basically the biases of the past. So, this is actually about skills, not background, not resume.
Ammon: Yes, I think it’s really important point to make about kind of the traps you can fall into here. So, you have to be careful, training on past data runs this risk of Naval-gazing, of doing all this work to just explain exactly in what way this previous system was inaccurate or biased. A number of efforts have been made, to do large-scale data science that predictive modeling on top of resumes.
Ammon: You look at those projects, they tend to fail and they tend to fail just because the signal fundamentally isn’t there. A core result from the rise of AI, and machine learning, is that data tends to trump better modeling. If your data is noisy and bad, you can hire the best achievement people in the world, and build the most sophisticated models, and you’re going to have limited success. If you have high-quality accurate data, especially labeled data where you sit and you do a supervise ML method, simpler models with access to better data are likely to beat most sophisticated models with worse data. So our approach here is really all about building the best data set.
Garry: So your model will actually work because it’s more about skills. It’s actually a larger data set, and you’re very focused on this more fair credential in some way.
Ammon: And that’s working. Triplebyte is just a different beast, right? We find people who have few other options and we get them jobs where they earn, three, four, five X the salary they earned in the past. We are changing people’s lives. It’s extremely rewarding. Extremely exciting.
Garry: The Triplebyte blog is full of incredible stories of people who… they were working on an oil rig or they were working in fast food and they had a kid. And how do you actually support them? But you know, in their spare time they learn to code. They loved writing software. They loved making, they loved being an engineer and they had all the skills to do it.
Ammon: Early on I have my, in my mind, the early days, somebody who was working in a warehouse. Their job involved driving a forklift around the warehouse. They had one program online and never met another programer in person. And so this person showed up for their interview, at Instacart I believe. Their first face-to-face interaction that ever had been with another programmer. But, they had the skills and they got the job.
Garry: And they’re here in San Francisco now.
Ammon: They’re here in San Francisco.
Garry: That’s incredible. So we talked a lot about the candidate side. What about the hiring side? We know a lot of people say that they can’t hire enough people. This must be transformative for them as well.
Ammon: Yeah, it is. Hiring is one of the biggest problems that every company faces. You know, startups early on, but as you get bigger, hiring the best people is just existentially important for every company.
Garry: And people spend a lot of time…
Ammon: They spend a lot of time and a lot of money. One of the stats that I’m most proud of, candidate who companies source through Triplebyte in an interview, go on to pass those interviews at about twice the rate of candidates who they bring it, sourced through their other channels.
Garry: So right there you’re basically cutting in half the amount of time you have to spend on on-sites. And the only people who can do those on-sites are actually your best engineers.
Ammon: And so that’s a huge time save. It’s a huge money save, it’s a huge attention save. Engineers dislike interviewing. And they especially dislike interviewing people who don’t do well. I’m really excited about — from a business perspective — we’ve been able to help some of our clients significantly grow their engineering teams. And I think this is a pretty massive opportunity, especially as software eats the world, right? So software eats the world means that more and more businesses need hire software engineers.
Garry: At Initialized, we have a thesis that you look at the overall GDP, so all products in all transactions in the world, and you can look at that pie chart and only a fraction of that pie chart is actually being driven by software right now. And you’re seeing Airbnb in housing and you’re seeing Uber in transportation, bit by bit, larger and larger parts of that GDP will be taken over by software and who is there except software engineers to actually build that? So they’re truly the limiting re-agent on the future of all of business.
Ammon: Yeah. And there aren’t enough currently, right? There aren’t enough software engineers accessible to companies. And I think it’s why it’s so important to look for the skills directly. Just from a business perspective. I talked a lot about why it’s socially important and that’s the mission, that’s what drives us, that’s what motivates me. But it’s also just important for the economy. As software eats the world, what that means is that, you know, grocery store chains need hire software engineers, right? If you’re a medium sized grocery store chain in Illinois, you have to compete with Walmart.
Garry: That’s our whole thesis for investing in both Instacart and Standard Cognition.
Ammon: Exactly. We view our role there, as making it possible for the little guys to compete. And that’s where we’ve seen a lot of growth recently. It’s moving out of just the core, Dropbox, Stripe, tech companies, into helping the rest of the economy compete.
Garry: What is it that Dropbox or Airbnb can do in terms of hiring software engineers that incumbents, that are not tech companies, they just get wrong?
Ammon: I think the number one thing is that they aren’t currently employing software engineers. So we’ve talked a lot about credentialism, and credentialism is a problem, but at least at Dropbox or Stripe, there is a deep pool of talent. Once you get through the problem of the credential process, there’s an interviewer there who can identify actual talent. And the problem we see at non-tech companies trying to move into that space, is that they really don’t know what to do. And so often they fall back on purely credentialism.
Garry: We see that problem with early stage startups all the time. You have a non-technical co-founder who cannot evaluate their technical co-founder.
Ammon: And so they just fall back on credentials and say, okay, we’re going to hire somebody who has a degree from.. a CS from a well known school…
Garry: They’re like Stanford, Google, Stanford, Google…
Ammon: And that misses… So I said earlier that degrees do mean something and correlate, and that’s true, but there’s also false positives. Hate to say it, but people who are not very good programmers sometimes get degrees from Stanford and MIT.
Garry: I went to Stanford and there some people who are totally brilliant and some people…
Ammon: Yeah, I think what we’re doing is really important for the economy, and our vision actually is broader than software engineering. Software engineering is a $100 billion market, and so this is where we’re pretty laser-focused right now, but the social impact of skills assessment is profound. It runs far deeper than software engineering, right?
This problem of credentialing is rampant. Look at how teachers are hired, look how lawyers are hired. Lawyers are hired, as far as I can tell, 100% on credentials. If you don’t go to a top law school, you simply do not have access to jobs.
Our vision is skills assessment, background skills assessment, kind of killing credentialism and our goal on a 10 year timeframe, is to help everyone hire it in this way that’s open, that’s fair, and that lets people move upward from whatever their background, get jobs based on their skills.
Garry: That sounds like the biggest idea that actually exists. Billions of people who have been tracked since they were 15, 16. You’re tracked and then trapped.
Ammon: It’s profound. It is profoundly important.
Garry: Well, thank you for doing this good work, man. This really important work, honestly.
Ammon: Thank you.
If you liked this interview, be sure to subscribe to Garry’s YouTube channel for more on startups working on big missions. https://YouTube.com/garrytan
Learn more about Triplebyte at https://triplebyte.com