Strategies for the Evolving CTO: A Discussion with Peter Bell

David's Notes:

In today's rapidly evolving technological landscape, the role of the Chief Technology Officer (CTO) has become increasingly multifaceted. From leading large engineering organizations to being a thought leader and public face, the responsibilities of a CTO vary greatly between organizations and only continue to diversify. How do we stay up-to-date with ever-changing technological advancements, and remain flexible with the ways that these advancements affect the nature of our jobs?

I was thrilled to be able to share this conversation with Peter Bell, the brain behind Geeks Who Lead, a learning hub for senior engineering and technology leaders across the globe. This is not the first time Peter and I have mused about the role and responsibilities of a CTO - he is a friend and colleague. Peter’s insights about these questions, and his emphasis on adaptability and innovation, have frequently been invaluable to me.

In this conversation, not only do Peter and I discuss the effect of AI on the role of the CTO, but we also talk about how remote work has changed the technology industry’s culture as a whole. How will these processes change our day to day? Do they have the potential to not only be used for advancement but be used for good? Peter’s analyses of these issues, combined with his rich history in the technology sphere, make this fireside chat a must-listen for anyone interested in the intersection of technology, leadership, and innovation. I am glad that we were able to include you in this conversation, and to learn from Peter Bell’s wisdom as I have. Enjoy!

Transcript:

[00:00:00] David Subar: Hello, everybody. Thanks for joining us at this fireside chat. Today, I'm talking with my friend, Peter Bell. Peter's a great guy. He's an entrepreneur, a visionary technologist, and an accomplished engineering leader. Maybe most importantly, he's the mind behind Geeks Who Lead. If you don't know about Geeks Who Lead, you should.

It's the only global learning community exclusively for senior engineering leaders, senior data leaders, and influencers at larger companies. It's for people that lead at scale. In addition to that, twice a year, Peter hosts CTO summits, developer impact leadership summits, and data platform summits.

These are complementary single day events for the same folks that are in Geeks Who Lead. The idea is to bring them together as opposed to just an online community so they can catalyze thought. And create connections to each other.

Thank you, Peter, for being here.

[00:01:06] Peter Bell: David, thank you so much for inviting me.

[00:01:09] David Subar: So today we're going to talk all about CTOs. What's going on in the CTO world. What's changing, how people keep up. Using your impact and your knowledge that you have from Geeks Who Lead. So let's start off by laying some groundwork. What are the different types of CTOs that exist?

[00:01:29] Peter Bell: There are lots of different types of CTOs, and there have been some blog posts that are like "the seven types of CTOs". It practices a balance, but there are a number of dimensions in which a CTO can engage. One is they can run a large organization and be primarily responsible for ensuring the trains run on time, that they have teams, that they're focused on the right business priority.

A lot of companies though, especially at scale, more of that will be handed over to FVPs or SVPs of engineering and the CTO is then often free to go in one of a number of dimensions. Sometimes it's a sales CTO. Their primary job is to deal with potential customers, to engage with them and get them excited about the technical solutions that their company's offering.

In other cases, they're a hardcore architect solving hard technical challenges and attracting people to the team who are interested in that. In other cases, sometimes they're a public figurehead. They are the person who does the presentations, the conferences, the blogging, really being a thought leader to attract everything from customers, investors of its VC back and the top talent to the company. So there are a number of ways that a CTO can work and it really depends upon their strength. And their primary responsibility though, is to say, how can we solve our business problems by leveraging technology? And then how can I put in place the team that I need to ensure that we solve all of those problems, even if I'm not the strongest person at doing all those things.

[00:03:08] David Subar: Okay, so you listed four different types, if I got that right. For those four types, what do they need to keep themselves up to date? Things are changing very quickly. Their roles - those four different types of CTOs have different kinds of roles. I assume the things they need to know are different in the world.

Are they actually different? And if so, how do they keep up?

[00:03:34] Peter Bell: Yeah. I mean. If you are primarily a thought leader, bizarrely, you actually need to keep up with presentation technologies and video recording technologies and sets of things which, unless you have a large team around you, that allow you to be a more effective presenter, both on stage and online. If you are primarily an architectural CTO, of course, you need to keep up primarily with the technologies, the changes in the capabilities of the systems that you rely upon.

If you are primarily a sales CTO, you need to continue to understand the challenges faced by the businesses or individuals who use your software. And also what's going on in the competitive landscape and make sure that you continue to have appropriate differentiators. And if you're primarily responsible for actually leading and engaging the org, then of course it's a matter of keeping up to date with best thoughts and practices.

It's funny, people have been the same for a while, but our expectations of our leaders continue to change and grow as we learn more about getting the most out of people, right? It used to be the floggings will continue until morale improves. Now it's very much Dan Pink drive, autonomy, mastery, purpose, servant leadership. Being really thoughtful about creating an environment where people can bring their whole selves to work and looking at all of the kinds of diversity you need within your org. And while those aren't the focuses, if you look at the top line, we run summits and we're probably not going to be doing a bunch of DEI talks right now, because this is the year of efficiency.

This is where everyone's looking at how they can report to the CFO more effectively, deliver more business value with less. And yet at the same time, it's incredibly short sighted to say that the answer to that is just go back to historic management patterns. You still need to lead with empathy and awareness of the amazing individuals on your team so that you can attract and retain the best talent, which is always going to be the differentiator in the long run.

[00:05:34] David Subar: But that sounds, those changes sound like changes in emphasis. Rather than direction. Would you agree with that?

[00:05:42] Peter Bell: The, the vast majority, yes. Though I guess it's like as you hit past the 90 degree point, it's like - oh! And suddenly that's a change in direction. Yeah it's - I think there will always be changes in focus. There'll always be seasons. But I think it's important across all of these domains to continue to say: how can we be customer focused and how can we do it in a way that attracts and supports the best talent to drive those initiatives?

[00:06:10] David Subar: Okay, I buy that. That makes sense. So I want to, I'm going to put a pin in that for a minute and dig into some technical changes and then we'll come back to that. So I went to a presentation last night and we were talking about - what does everyone talk about today? Gen AI. But we were really talking about Gen AI Ops.

[00:06:30] Peter Bell: Hmm.

[00:06:30] David Subar: And some of the interesting takeaways are - if you assume that Gen AI falls under the CTO, data science falls under the CTO, that may not always be true. You have Data Engineering, you have Data Ops, you have Data Science, you have ML Ops, you have all of these other things, and then you have this whole new QA process.

Because it's not enough to QA the software, you have to QA the data coming in. Did the model drift because the model drifted because you're retraining it all the time, or did the model drift because the data input changed? And so the implication of this, and there's, there were, you know, four or five other layers to that.

The implication of this is the CTO who's leading the technology, their job has suddenly expanded. It's changed.

[00:07:26] Peter Bell: I think it's a really interesting question. And organizationally, this is going to be implemented in a lot of different ways in a lot of different companies, because it's the right answer is not really about the titles. It's really about the humans. To take our naive example that we've been dealing with for a long time and I know you're well aware of - think of the CTO versus the CPO versus the CPTO. At a super, super high level, we're moving towards a world where it is not uncommon for both product and technology to roll up to a single person. Because at the end of the day, the job of the technology is to under, is to leverage product management best practices to solve business challenges for your customers, right?

So, uh, that said at some companies, it's perfectly appropriate to have a CTO with a VP of Product. At other companies, a CPO will have a VP of Engineering. Historically, the challenge was that if you had a CTO with a VP of Product, it was more likely to, you are more likely to do an excellent job of building the wrong thing.

If you had a CPO with a Vice President of Engineering, it was harder to attract the best engineering talent. Cause they're like, wow, we don't even have a C level seat at the table. We're probably going to do a crappy job of building wonderful things because we're never going to get the support and time to pay down technical debt and all the rest.

But again, it actually comes to the individual. If you've got a CPO with good empathy for the challenges of engineering, it really doesn't matter what those organizational titles are. Honestly, for me, generally the worst of all possible worlds is a CPO and a CTO that don't get along extremely well, because then basically you're telling the CEO that his job is to stop focusing on the company and the board and just deal with the bickering between the two C level positions trying to decide what to build, which is never a good strategy.

So to roll that back then, it's interesting. I run conferences historically for senior engineering leaders. So CTOs, directors, and VPs of software and platform engineering. We just started over the last year or so, including CDOs, chief data officers and VPs of AI, ML, data analytics into that community, as well as like senior data and software architects and engineers - so staff, principal, and distinguished engineers and architects. The reason we've broadened that is I can't tell the difference between data and software anymore. If I'm a CDO running a data engineering team, if I'm building a data platform, I've got a team of software engineers writing code who need to be thinking about all the same things that my software engineers do in terms of observability, reliability, resilience, engineering, all of these are challenges we need to face. So I don't - so the short answer is, I don't know that for every company, it will be that the CTO will be responsible for ML ops, AI ops, gen AI ops, but I do think that the CTO and the team that they work with are absolutely going to have to deal much more with data as a first class citizen, because it's the output of a lot of their systems. It's the input to a lot of their systems. And it's going to become increasingly critical to be able to deliver reliable data products at scale as part of running a software engineering org.

[00:10:56] David Subar: Okay, so let's, let's pretend for a minute. I'm a CTO in a company where I have all of that,

[00:11:04] Peter Bell: Mm

[00:11:04] David Subar: right? I have data, data ops, ML ops, you know, data science, all of those things. And this all happened to me in the last nine months. And by the way, the world's changing every three months because we're learning how to do Gen AI.

How do I keep up? What do I do?

[00:11:24] Peter Bell: I mean, honestly. I'm not sure that you will. I think I'm going to suggest some things that you have to do. First step is start using it yourself. It turns out that there are use cases as an executive for generative AI. And it's interesting, we did a survey at our summits back in May and I asked a room full of, you know, 150 people like, okay how impactful do you think generative AI and large language models are going to be on the process of software development?

I then asked, okay, how much time a week do you personally spend engaging with OpenAI, BARD, whatever tools they are? And there was a very strong correlation between the people who understood how impactful this is eventually going to be, and the people who are getting hands on experience of doing it themselves.

So step one, play with it. If you're not sure how to play with it - ask it. Like, this is the bizarre thing, because I was speaking with Mike Buford, the CTO at Greenhouse, and one of his great examples was that he says ChatGPT is the best tutor he ever had. Because he would literally just ask it, okay, ask me a set of questions to determine my competence and understanding of generative AI.

And then for anything I get wrong, ask me additional questions to determine what my misunderstanding was that caused me to get that wrong. And then provide me with some more information. And then ask me a new set of questions. So basically it was ad hoc mastery based learning of the domain just by engaging with thoughtful prompts.

[00:12:55] David Subar: Let me just pause you for a second because that was a great interview. You did a great podcast. You did. I'll put it in the show notes. I listened to it. I emailed you about it. I suggest everyone listen to it. Keep going. Sorry.

[00:13:06] Peter Bell: Well, and then the next thing I also want to say, just another interview, I interviewed Matt Welsh, the CEO of Fixie. ai. He's got a strong background at scale up startups. He was at, I think, Apple and another big tech firm. He was a professor of computer science. And we did this kind of foundational interview and it was a lot of fun where it was just like, okay.

What are all the reasons that people are saying this whole generative AI thing's never gonna work? What are the legal implications of potentially having code generated by Gen AI and the provenance of that? What are, what do you say to people who says it's just auto complete?

If there's anyone listening who's like, it's just auto complete - A. As Matt said, you're right, it is just autocomplete, but B, that fundamentally misses the point of emergent behavior. And I'm just going to take one moment to kind of set the scene for anyone who's like, "Why am I listening to a podcast on Gen AI? It's never really going to amount to anything because it's always going to hallucinate." The short answer is you're wrong. This is going - I did another interview with Albert Wenger, the managing partner at Union Square Ventures in New York City. So a very well established VC firm. And he said two things are equally true at the moment.

Firstly, there is an unbelievable, outrageous, and completely indefensible hype bubble surrounding AI. And secondly, the move to a Gen AI world from a non Gen AI world is going to be more of a transition than moving from an agrarian to an industrial society.

[00:14:43] David Subar: That's interesting. That's a bold statement.

[00:14:45] Peter Bell: It is, and I think.

[00:14:47] David Subar: Satya said something very similar. Satya, the CEO of Microsoft, said something very similar.

[00:14:50] Peter Bell: A lot of smart people are saying that. So I generally start listening, generally, like the immediate thing is, oh, come on, it's just auto complete. But when you start hearing statements like that from people who have deeply evaluated, who you respect their previous decisions, I think it's going to be transformative.

So, given that it is going to be transformative, we should probably take it seriously. And the, the auto complete thing, yeah, the thing about auto complete is emergent behavior. Which is at a very simple level, you have a very small model and you ask it to do some arbitrary human task and it sucks.

You have a slightly bigger model, still sucks. Slightly bigger model, still sucks. Slightly bigger model, suddenly it's human level performance. And there's no a priori way to determine what stuff you can and can't hit that way. So there's no reason to definitively be able to say that there is anything that a human does that with a sufficiently large enough, well formed enough and well tuned enough Gen AI model, that it would not be able to replicate or improve on that behavior.

[00:15:53] David Subar: I'm going to make a different argument for you.

[00:15:56] Peter Bell: Go for it.

[00:15:58] David Subar: It's not going to replicate human behavior. It'll be as good as us in some things, not as good as us in others, and better than us in some. And it's hard to predict a priori, which of those are going to be which. But we know that even if it's good enough as humans, good as humans rather, in some areas, being able to do that at scale has great leverage, and we're going to find out where that is.

[00:16:27] Peter Bell: Absolutely. The cost - I mean, you have to look at costs. Like one of the first things you need to do when you're thinking about AI ops is look at the cost of running the operations, the cost of training the models, the cost of running queries, and also the latency, just how long it takes from typing something into the box to getting it back.

That's why Google hasn't replaced its current search interface you know, end to end with, with generative AI. One of the reasons is that there are both costs and latency issues around the systems, at least now. But roll it all the way back to how you get up to speed. Firstly, play with it yourself.

And then secondly. You need to have a person who is responsible within your engineering org for nothing, but figuring out, and these are probably two different people actually within your larger org. One person in the product org needs to figure out from a product perspective, what does this mean? For example, if you have a lot of detailed reporting and administrative screens, do you actually even need them? I don't really want a reporting screen with 47 variables. I need to drop down and click on, and cut and paste. It's just the only way we figured out to query complicated data in a manageable fashion. You can replace a lot of that with a textual model, right?

Where you just, you just provide prompts. And I think it's going to be really interesting to see how many elements of existing pieces of software disappear or are radically transformed by it. Second thing to think about is what are your competitors potentially going to do or what are disruptors going to do in your industry?

And then the third thing to think about is what features should you be prioritizing and shipping? I did an interview with Matan, one of the co-founders of a small company called Spinach. io. Usually we're dealing with larger companies, but he gave a great example of how they went from, "Oh! What is the generative AI thing?" to actually shipping working features in a very small amount of time, leveraging prompt engineering, and then kind of filling in the details after he'd spiked it with vector databases and all the other tech you need to make it scale.

So. On the product side of the house, you need to look for - what is, what are the disruption threats to, is your category going to disappear? Are there certain categories of reporting tools that simply won't be necessary once conversational interfaces are good enough? What are your competitors going to do that could disrupt you and what can you do to disrupt your competitors and add more value to your customers?

You also need somebody on the engineering side of the house saying, so how does this impact how we write software? It's clearly going to be more than just moving from VI to an IDE with like code complete. It's going to be more than that once we figure out how to use it. Is it going to be a 50 percent increase in productivity or a 20X increase in productivity?

We have no idea, but you need to have people running spikes to determine approaches to implement the technologies now within your engineering org, and also running thought experiments around how the structure and nature of your engineering organization might change over the next five years.

[00:19:44] David Subar: I'm going to, I'm going to add a couple of things to what you said. First about going from data screens to textual screens. I'm going to argue that's the Claude Shannon problem, right? Claude Shannon, father of information theory. The difference between data and information, information is that which reduces uncertainty.

We're giving people data now. We're inundating with data and they have to reason about it. By abstracting out into what Gen AI can do, which won't necessarily just be text, it could be graphical, Even if it hallucinates, it's going to allow someone navigating the information and then digging into the data better.

That's A. B is, I agree with you on engineering, but I think that there's a third point as well, which is - what is Gen AI, what is deep learning good at? If you just have a matching problem, where you match this known piece of data to this known piece of data as a relational query, we can do that great today.

We've been doing that, you know, since SQL came out in 1976, I'm making that data up. I have no idea. What we haven't been good yet is, what happens when things don't exactly match? I think these data records and these data records are the same customer, but they logged in with different authentication credentials, and so I can't do exact matching.

I can't do good clustering, for instance. Deep learning is good at things that nearly match. What's the next few words in this sentence, says

[00:21:29] Peter Bell: Probabalistic.

[00:21:30] David Subar: Chat GPT. Right, exactly. And I think that is a class of engineering problems and a class of products that we're just starting to imagine now.

[00:21:41] Peter Bell: Absolutely. And I think you really nailed it with just starting to imagine, which is nobody knows. Like, I have now interviewed probably 20 people who I would consider to be, you know, deeply immersed in generative AI, and I feel like in some ways I have less certainty about what two years from today is going to look like than I had before I started the process.

There's no question that this is going to be transformative, and it's going to be transformative. It's the nature of complex adaptive systems, right? You have simple systems, super easy to reason about. Complicated systems, there's a lot going on, but once you hit a certain, a certain level of complication, they actually become complex systems where it is not deterministic from a given input what the output's going to be.

Kind of like the butterfly wing problem - butterfly flaps its wings and suddenly you get a tornado someplace in the North Atlantic and so the challenge I think with this, or the approach to this is the OODA loop, right? You know, is it basically you need to take these agile approaches where you spike things, where you try things, where you determine whether or not you're getting the output you want, and then you adapt your process depending

[00:22:53] David Subar: So so those listening, those listening don't know,

[00:22:57] Peter Bell: Yep. So it's observe, orient, decide, and act.

[00:23:01] David Subar: That's right.

[00:23:02] Peter Bell: And there are a number of other similar frameworks, but at the end of the day, with a complex adaptive system, you don't definitive, you may think, you know, what will happen when you do X, but unfortunately, it's going to cause your competitors to do Y, which is going to cause the customers to do Z, which is going to cause the government to decide to do F, which is then going to mean that the second and third order effects were not what you expected.

So it's important to take small steps and to iterate quickly, towards your desired outcome.

[00:23:31] David Subar: Yep. Okay. So we dug into a bunch of different things around Gen AI. There's other things that are changing in CTO's world. You talked about the, what I'll call the seasonal aspect of the macroeconomic effects, trying to be more efficient with capital, the kind of things, what other things are changing that are important.

And then I'm going to go back to that follow up question about how would you stay up to speed with these other things that aren't Gen AI?

[00:23:56] Peter Bell: So it's a great question. So some of the other things that are changing, obviously we're still trying to navigate the workplace. I think that it's funny cause I was, I was a digital nomad back, you know. Back in, I think 2006, I had clients in the States and I decided to spend the winter in Sydney because I was done with having winters.

I wanted to have three summers in a row. I built a remote first online instructional team for Flatiron School of Bootcamp back in 2016, built up a team of 15 and a management team. And a lot of people say "you can't build and manage a remote team" or "you can't X with a remote team. You can't build rapport."

You can't brainstorm effectively. And it's just not true. Anybody who thinks that you - if there's anything you think you can't do purely remotely, you're probably just not doing it right. Turns out that management is a lot harder for a remote team. All this stuff you get for free. If I'm running a team in an office and, you know, Mark's crying by the water cooler, I'm like, I should probably go have a chat with Mark cause something's clearly not right.

Whereas if he's just crying, you know, with his camera off on zoom, you're not necessarily going to pick that up. So you absolutely need to engage the whole human. You need to have autonomy, mastery, and purpose. Your job with your weekly or bi weekly one on ones is to build a strong relationship with your team and to teach them to do that all the way down and through your org so that people are aligned and passionate.

And this stuff I'm seeing about productivity being lower remotely, it's simply not true if you do it right. I think there's a wonderful place for bringing humans together for onsites, for in person experiences. And there are certain things that we certainly can't replicate the speed and ease of making human connections with others in person is why I run in person conferences, not just online ones. But we are still figuring out this whole hybrid world.

There are fundamental challenges with the model of coming back into the office two or three days a week, given that we now have the expectations that I can work in San Francisco, but live in a yoga retreat in Santa Barbara, or in a horse farm in Montana. And if you want to capture that top 10 percent of the market that wants those lifestyle choices, you're going to have to figure out how to do remote right.

So one thing we need to get up to speed with is actually understanding how you use tools like Miro or Miro Online to, to do brainstorming, to do shared envisioning, and how you can effectively leverage in person time in a way that's supportive of what's probably becoming a global workforce.

[00:26:35] David Subar: Interesting. Yeah my position is slightly different than yours. The question is how many of the top 10 percent are there? And if you can't recruit someone that want to live on the horse farm in Montana, are there enough other top 10%? I don't know the answer to that, but perhaps. And another thing I argue is - the bandwidth of a whiteboard, and while you're not allowed to do this in a work situation, the bandwidth of a hug is higher than an "attaboy" on a Zoom.

[00:27:08] Peter Bell: It absolutely is, which is why you need to be more intentional and invest the time in the 10 minutes before stand up when everyone on the team, the fact that you do explicitly do. Now, we've gone past the Zoom drinks once a week, like everyone burned out on that about six months into the pandemic. But for example, what you do is in your Slack channel, you allow a communities of interest around cooking, bicycling, whatever it is.

And then what you do is for anything that can aggregate a certain number of employees, you give them funds to allow them to organize experiences online that bring connections. And what that does is it's interesting. It not only makes people to be more connected to the company, but it also creates those loose connections.

Like, oh, I was just talking about baking with John in finance. So, you know, if we're, you know, if the CFO is like not getting back to us, don't worry, I'll ping John. So those kind of water cooler conversations, those informal conversations and networks can absolutely be replicated online. All that said, I agree with you a hundred percent.

Nothing beats a hug or like that. There is absolutely higher bandwidth communication in person. And yet it's possible to build deep trust. I have people that I worked with for four or five years that I never met in person. And I feel as connected to those as people who I worked in an office with for years.

[00:28:32] David Subar: Well we, as you know, we have a lot of different clients around the world, and some we coach, some, we actually do interim working. And several of our clients have taken to the path of, either or both, bringing people all together once a quarter, everyone flies into Boston or DC or LA or, and or, the manager flies to various cities over time to meet with folks face to face. Even if there's only two people in the city, not everyone necessarily gathers in the city, but the manager becomes the nomad.

[00:29:07] Peter Bell: Absolutely. And both of those, and again, it's like, you know, should you have a CPO or a CTO with a VP or the other? It depends upon the humans. If the manager's got like three young kids, probably not going to be a great lifestyle choice. If they're either pre kids, post kids, or don't want kids, maybe they don't, if they've got a life partner, maybe they don't like them too much, having them on the road like 200 days a year might actually be the best way to go. So that will attract some talent and, and repel others, which is absolutely valid.

[00:29:37] David Subar: Right. Okay. What are the other things that are changing that are important to talk about?

[00:29:41] Peter Bell: Connected to that globalization, right? I mean, why are we - once we figured out that we could do most of our work over Zoom and over the interwebs, do we absolutely need people living in the country that we live in to do all of that work? And so I think over the next couple, you know, I don't know if it's the next 20 to 50 years, we're slowly going to see a trend towards income being less dependent upon the country you happen to be born in without you necessarily having to move out. There's still huge opportunities to get a quality of talent that you could not otherwise afford by being thoughtful about the countries that you hire from. And that's continually changing, right? I had a team in Ukraine and then I'm like, wow, post pandemic, I couldn't even afford Ukraine.

So I'm like, okay, so how about North Africa? How about Morocco? How about Nigeria? There are many sources of talent around the world. And as Andela, one of our partners says,

[00:30:47] David Subar: I was going to talk about - that's exactly what I was going to say.

[00:30:49] Peter Bell: Is like talent is equally distributed, right? It turns out that not every smart person in the world was born in America. Who knew?

[00:30:58] David Subar: Yeah. I was going to say also, the CEO of Andela, we both know, says exactly that. He believes intelligence is equally spread throughout the globe, and I think he's right.

[00:31:07] Peter Bell: Absolutely. And I think we're going to see that increasingly. Historically, there have been a couple of confounders for that. One confounder was potentially the education system was not equally good for children all around the world. And secondly, access to both technology and the bandwidth to connect with people was not equally distributed.

That aspect, the first challenge is going to be solved by generative AI. Once we start wrapping the right prompts and interfaces around it, you basically have an infinitely patient tutor that will work with every child in the world, one on one, to bring them up. And there's this two sigma problem, which is basically there is a substantial deviation between the results you get if you're mentored one on one versus having classroom or group instruction.

Suddenly nobody's going to need group instruction as the primary way to learn things because they can have mastery based learning automated through large language models.

[00:32:05] David Subar: Well, Sal Khan is working a lot on this updating Khan Academy. My kids, when they were in high school, did classroom flipping. They would do the lectures at home on video, and then the teachers would answer questions. This is going to flip classroom flipping yet again. Now question is, having access to the machines where people can do this, but, yes.

[00:32:31] Peter Bell: Yeah. There are still, there are still going to be continue to be equity issues. It turns out that it's good to be married to rich people in a first world country or to be born to rich people in a first world country. That will probably continue to be true for the foreseeable future.

[00:32:47] David Subar: Okay. Other changes we should chat about? Because I want to get to, I want to get to groups of folks and stuff like that. But other, other changes?

[00:32:55] Peter Bell: There are some fundamental changes going on. The world of software engineering in some ways is broken. I remember back in the day, you know, walking. Walking 12 miles to school every day through the snow. I remember when, if I wanted to ship software, I needed to FTP a couple of files I'd written in classic ASP or CodeFusion or PHP or Perl back in the day, FTP them to a server.

[00:33:25] David Subar: Was CodeFusion before or after abacuses?

[00:33:29] Peter Bell: All about the same. I think mid nineties was really the renaissance when everyone was like, wow, everyone can program these web things. And what's interesting about it is now the level. On the one hand, we can do amazing things. We are being challenged to build things which in the physical world would just be too complicated to actually build.

If you start to look at the complexity of even things like the Windows operating system. I would challenge somebody to find something with the same degree of complexity that has been built by humans in the physical world. And so because of that, we continue to build higher and higher levels of tooling and abstraction.

But unfortunately what that means is, let's say you just want to be a software developer. Well, you're gonna want to pick up JavaScript for the front end. Of course, you need to know React. You're probably gonna wanna understand GraphQL, you might wanna consider TypeScript so that you can control thetyping so that it's easier for more people to engage with your software and to understand what the intent is of the variables and the objects. Then on the backend, you're probably going to pick up some arbitrary language that you could make it the same JavaScript, but you still have to learn a whole set of other things with Node.

But then we need observability because you've got lots of little microservice talking to each other. You've got logging, incident response, feature flagging. And the levels of complexity and sophistication, even before you start adding ML and AI ops to the table is just overwhelming. So I think we have this challenge of exploding complexity, partially vendor driven, but also those vendors are solving real problems and we need to figure out how we continue to allow people to reason about the systems that they're building, at least at a high level.

So I think that tooling is going to continue to change the structure of engineering orgs. I think everyone with more than 50 to a hundred engineers is going to end up with some kind of platform team, whether it's just a rebanding of the operations department or whether they're also responsible for developer onboarding, developer happiness, developer experience and productivity.

And I think all of the, one of the big things you need to do is actually sit down with the vendors and learn from them. They have stuff to teach us about how they're helping other companies to build better software.

[00:35:48] David Subar: Okay. Okay, so all this stuff is happening. You and I regularly talk once a month, but we don't broadcast them. Aside from hacking into our usual conversations, how do people create community? You and I created community with each other one on one, right?

But

[00:36:10] Peter Bell: So so there are lots of scales and ways to engage with community, but I mean, at the very least find an existing broad community, something like the RAND Slack. It doesn't matter whether you're a team lead or a CTO at a three person company or 300,000. It's a way that you can engage with other engineering leaders and see what they're thinking about various topics.

The challenge with that, of course, is it's online only. So you don't get the extra bump of connecting to somebody quickly and deeply by meeting in person. And it's a very broad church. So there's an awful lot of team leads and EMs, engineering managers talking about, how do I run a performance review?

And how should I do my one on ones? Which may be less appealing and compelling if you are 30 years in as a CTO running a team of 4,000. And so then increasingly what you need to do is find specific communities. We're building community for geeks who lead at scale - directors, VPs, CTOs, CDOs running engineering orgs and data orgs at larger companies.

And we're trying to create an online community and in person experiences where people can connect. But definitely, I would say, one of the things that I see consistently is that there are two types of CTO. There's the one who can get their next job and there's the one who can't. I would suggest it's worth being in that former category just out of self interest.

Connect with enough other humans who have overlapping interests, who are dealing with similar challenges, who are engaging at scale so that the next time you're looking for a job or for a solution, Oh, I need to hire a head of data science and I know nothing about what questions to ask them. Maybe there's another CTO who did their PhD in data science and would absolutely help you out with three or four questions you should use for first round filtering of the candidates for the VP position you're trying to fill. So I think it's incredibly important to find ways to connect with other humans who do similar things at a similar scale.

The one other thing. We used to have startup CTOs running teams of 20 in the same room as enterprise CTOs running teams of hundreds of thousands. Turns out the only thing they got to talk about is the weather. Because they may have the same job title, they do not do the same things or focus on the same challenges.

Macro level, it's the same problems. Deliver software, attract talent. The ways that they're doing it and the challenges they're having are fundamentally different. So find people as similar as possible in role and scale and connect with them in whatever ways make sense. Ideally, meet in person and then kind of sustain that relationship online.

[00:38:53] David Subar: Okay, my friend, you're burying the lead. Here's what I wanted you to say, I'll keep this on the video, is there are communities, you talked about RANDs, and that was very generous of you, there's CTO Slackers, another online community, that's, you know, but meeting face to face, there's only a few places to do that.

One of them, and I don't mean to make this an advertisement for Geeks Who Lead, but I think it adds value here.

[00:39:23] Peter Bell: Right.

[00:39:24] David Subar: Talk about your conference. I'll talk about LACTO Forum, the nonprofit that I'm involved in, but.

[00:39:29] Peter Bell: Yep. So, so there are two broad ways of doing the in person thing. Firstly, find a local CTO club. There's a New York CTO club, Chicago CTO Group, Denver CTO Club. There's Atlanta, Seattle, San Francisco, the LA CTO Forum. And then, across Europe, you know, there's AlphaList, there's TechRocks.

There are lots of these groups that engage. And if you can connect with a monthly meetup, in person CTO group, you should definitely do that. It's a great way of building that relationship and building those connections over time. The one challenge with all of those orgs is they are somewhat broader in like you'll probably find CIOs and CISOs there, which is both good: "hey, I got a security question. Let's ask the CISO". And bad: some of the people you're chatting with simply won't have the insights to the problems that you are facing. And so what we try to do is twice a year in San Francisco and New York is aggregate people with very similar challenges and issues and create an environment where they can learn from vendors who absolutely have things to teach us, but they can also learn from case studies from each other's and from structured table conversations where they build deep relationships with their peers so that they can continue those relationships online throughout the year.

[00:40:44] David Subar: Yep. And, I've been to the summits.

[00:40:48] Peter Bell: And presented at, thank you so much, David.

[00:40:51] David Subar: No problem. And they're very valuable in the things that - the serendipity that I have found. Either someone on stage saying something that I hadn't expected or someone at the table saying something that I found suprising, has been very helpful to me.

[00:41:09] Peter Bell: With that said - we've actually completely restructured everything. It used to be 20 minute talks. I remember a good friend, who came up after the pandemic, said "Peter, the only thing I want to say, your talks used to be shorter". And I had to say to him, it was the ex CTO of InVision, I had to say "Bjorn, I have some bad news for you. Your attention span's shorter, the talks are the same length they always were. So we've got all the talks down to 10 minutes. And we spend for every one minute we spend listening to someone, we spend a minute having informal conversations. And then we have a number of structured breakouts, where we will take three 10 minute talks and then spend 20, 30 minutes talking with the other participants at the table about how does that apply to your org? And do you have experience doing that? And what would you do if you were trying to deal with legal who was maybe uncomfortable with incorporating Gen AI produced code within your code base?

So, yeah, we've had, because that serendipity, what you really want to do, we're all, most of us CTOs are still geeks at heart. We kind of like, don't really want to speak to humans, but once we got a topic to talk about, once there's a, somebody teased it up with, well, what do you think about building a globalized workforce, pros and cons, let's have a, you know, "he said, she said", and then provide an environment for everyone to discuss that. That's when actually the deep connections occur because people now have a reason to engage. And in doing so, they build community and connection with peers who can help them long into the future.

[00:42:39] David Subar: Well, so I'm going to wrap, but before I wrap, I want to say something about Peter. People who are listening might notice the deep knowledge that Peter has about technology, technology development, org design, these kind of things. Peter leads Geeks Who Lead, so... although I know the inside of his organization has a lot of deep technology to run it, he's not today a coder. He's not today a CTO. And so one of the reasons I respect Peter, his ability to see deeply and widely about a number of things that attract, that impact our industry, that impact how we lead, that make a difference and also vision to the future. So that's one of the reasons that I always enjoy interacting with Peter. It's one of the reasons I wanted to talk to him in this forum. So thank you. Thank you very much, Peter. If people want to get ahold of you or learn more about Geeks Who Lead, how do we do that?

[00:43:43] Peter Bell: There's geekswholead.com. Just go there. If you want to send me an email personally, it's peter@geekswholead.com. I do my best to respond to anyone who drops me a line.

[00:43:55] David Subar: Excellent. Thank you very much.

[00:43:58] Peter Bell: David, thank you so much for the chance to chat. It's always a pleasure and this time we get to share it with the world.