Host Intro (Gene Kim)

All right. So just to set the stage for today, I found myself sort of reflecting a lot this morning thinking about everything I had learned yesterday. And I want to maybe put some things in a frame about what I think is interesting and important and why.

So some of you may know that I worked with Dr. Nicole Forsgren and Jez Humble on what became eventually known as the DORA metrics, right? The four-tuple: deployment frequency, deployment lead time, change success rate, and MTTR. And yeah, so that was a cross-population study that spanned 36,000 respondents over six years. And I think it was just a really— the big insight here was that high performers exist, and they outperform their non-high-performing peers by two to three orders of magnitude. And so that's across every dimension. So one to three orders of magnitude across all four dimensions, right?

And so just to put that in perspective, when I was working with Dr. Steven Spear — famous for decoding the DNA of the Toyota Production System — we worked on the Wiring the Winning Organization book together. In manufacturing, right, researchers were marveling at the fact that Toyota could create one-half the input to create twice the number of outputs. So that's like a 4X increase in efficiency, effectiveness, whatever. And so when he saw these stats, it was as much of a revelation to him as it was to us.

And what's interesting is now in this AI community, right now some people are reporting 10X gains, right? Some people are proud to report 20% gains, right? And it's like, I don't think anyone's lying, right. But what can explain the different effects? How are some people gaining these outsized gains while other people are eking out? In my opinion, if you're getting low double-digit gains, we have to be hunting in the wrong part of the forest. That's almost not — I'm not saying it's not worth reporting, but that is not the rewards we're looking for here.

And that's just the beginning, right? Now, here are some kind of... I loved how so many people said, "Wow, no one really does know anything," right? Now, there are so many open questions that seem almost contradictory — puzzling, baffling, right. Who adopts AI more, juniors or seniors? Don't know. Do we do more TDD or less? It was amazing to hear Kent Beck last week say, "I don't know." Right? And if Kent Beck doesn't know, right, anyone who had certainty around this question maybe should revisit why you feel so certain. More capabilities to customers, even if they can't absorb? Don't know. Do we keep deploying more code to customers even though, one, they don't feel totally in control, and it's more than what the customer can absorb? Do you keep doing it? Don't know. Is OpenClaw safe enough to use judiciously, or is it way too dangerous? Don't know. In fact, as we were getting ready, I was like, "Kiloclaw?" I was like, "Oh, that sounds interesting."

Yeah, so I just love that there's so much variance in what the answers to these questions are. And it reminded me of some of the slides I showed yesterday around the Solvay Conference, around how it really transformed quantum mechanics in the early 20th century.

And so I just want to geek out for a second and describe this really interesting scene. Apparently in 1801, these physicists thought physics was mostly a solved problem. And by the way, how do I know this? I've gotten in the habit of going to sleep listening to this YouTube channel called "Going to Sleep on Physics" or something like that, and it's all about quantum mechanics. And it's just, for whatever reason, I love this. Anyway, so it's all sort of fixated on how really confusing quantum mechanics is. So it starts in 1801. Thomas Young does the double-slit experiment and shows that, wow, light is both a particle and a wave, which is really confusing, right? And there's apparently the uniqueness of the observer that sort of turns a wave into a particle. So very confusing.

And it brings up the question of, if a particle's a wave, then all waves need a medium to be transmitted through, like sound gets transmitted through air. And so Michelson-Morley, they do this fabulously expensive experiment, right, where their goal is to measure the speed of light — do it perpendicular and parallel to the speed of the Earth. And right, they're going to figure out what is the speed of light, and they'll capture this thing called ether because everyone knows that it has to be there, right? Because if light is a wave, there is a medium through which it gets transmitted. And it was probably one of the most famous null experiments ever, right? They could not — the speed of light was the same no matter what direction it was going.

So very puzzling. Alexander Graham Bell helped fund it. It involved creating these incredibly precise optical instruments floating in a pool of mercury. They're trying to isolate all the measurements from anything that can disturb it.

And so the way the physics community responded, for I guess 50 years when this came out, was puzzlement, confusion, excitement, and a lot of smart people saying, "I've never seen this before." And it turns out 18 years after this, a guy named Fitzgerald — which I sort of remember from a physics textbook — said maybe objects physically shrink when they move. And this is a really strange idea, right? That seems preposterous. And then Einstein writes this paper in 1905 as a patent clerk, right, and actually says, "All right, the speed of light is relative." Right? That length contraction is real.

And yeah, just a little footnote — apparently, Lord Kelvin, I just learned this morning, said physics was a solved problem. There's just these two little anomalies that we need to explain. He called them two clouds. One of them turned out to be the theory of relativity, and the other one turned out to be quantum mechanics, right? So it's like these were not small things. They actually upended our understanding of physics.

Anyway, so I had mentioned the Solvay Conferences yesterday, because the way this happened — this kind of discovery and your understanding of the way the world works — they published in journals, they had these conferences where they debated, they wrote letters to each other. And that's how the understanding of the modern universe was formed. And I wrote down here: they could write each other letters, they could have an annual conference, they could argue over dinner, go home, think for a year, and come back. Because the speed of light is not going to change. It will wait for them.

And in contrast, Scott Prugh said, "Does anyone want to talk about myths? Because I'm having a board meeting later today, and I really need to come up with answers." I don't think these physicists ever had, like, we have this thing where there's this monster being unearthed that's going to wreck all the critical infrastructure on the planet, and we need an answer today. So the tempo of learning is massively higher now.

So I'd mentioned the monthly AI forum. The goal was, after last conference in September, it was like, all right, we're going to assemble the most switched-on people using AI and vibe coding, and the goal is to share experience reports. And not do it on a yearly cadence, but on a monthly cadence. And if you don't know about this — a bunch of you asked me about it yesterday — Anne Perry is going to put a message in Slack with more information about how to get it.

And then the other thing I'll mention, just in the spirit of sharing. So earlier this week, we had 50 really switched-on people, and we're just trying to puzzle this out. And I was in a really amazing group with Matt Jones, Andy Bean, with Ryan Martens — he was the founder of Rally Software, he's back from retirement — and Joe Beutler from OpenAI. And I think it was really dazzling because we were trying to explain what would be a theory that could explain why do some organizations get 20% or nothing while some are getting outsized gains.

And the one phrase that kind of stuck with me was, does Conway's Law still matter? It's like, oh, maybe it doesn't. But after going to sleep that night, I woke up like, no, you don't get to choose whether Conway's Law matters. That's like the law of gravity. It chooses for you.

And so for me, the kind of epiphany — or the theory that I think we know what we need to know already — is that we think AI is this sort of technical transformation. That's true. But the reason why everyone's talking about bottlenecks and why are you getting these weird pairings of teams and people is that it changes the entire communication topology of the organization.

And so there are some teams and groups that always used to have to talk — like dev and ops, reviewers and developers, dev teams and auditors — and now with LLMs, like a developer and a Terraform expert, they should never have to talk again. These LLMs are a sufficient tool where they can gain independence of action and get sufficiently decoupled. But then there are all these other conditions where teams that never had to talk directly with each other are having to talk all the time. Like customers and engineers, or the sales team and engineers bypassing product, or CIO and frontline operators, like we saw with Matt Jones.

And so I think the reason why this theme of the CEO matters — the only point that can create those cross-functional connections is the CEO. With DevOps, it was the head of technology. Now we have to go all the way up. And I think that's going to be a big challenge, and I think it does sort of substantiate this claim: man, this is really going to be the biggest leadership challenge in the last 100 years. And I see Ryan Martens back there nodding. So I think that is a challenge in front of us.

So we started to catalog all these bottlenecks. And so I guess these are the questions we're going to seek to gain clarity on over the next day. And I'll have some questions and thoughts on, like, okay, how do we move this forward? And we don't get to have 35 years like the physicists did 100 years ago. We have to figure it out in weeks, months, or quarters because decisions are being made right now on your behalf that you have to make decisions on.

All right. So with that, let's go to Brian to help us open up the day. Thank you.

Brian

Day two of the AI Summit here today. I think we have an incredible lineup again today. Yesterday was awesome to experience all the different great speakers. And I'll tell you, a new experience for me yesterday. So as many of you know, we were working on that game yesterday with Brian. And him and I were spending time in that back room doing some pair vibe coding, which is a new thing for me, which was awesome. We had multiple computers running. We were doing voice to Claude. We had multi-agents running at the same time — front-end, back-end, deploying to the servers, doing live QA-ing. It was an awesome new experience for me, something that I'd love to explore more, what that looks like.

I've been an advocate for pair programming for a lot of years — the dojo style, all these things. It's a whole new world with this, and I think it's something really cool to explore. So something to think about there.

Before Gene kicks us off today with the speakers, let me get you squared away with just a couple of things. Schedule-wise, expect great sessions today throughout, running till 4:30, so stick around if you can till the very end. We have some exceptional speakers all the way up to the very end today.

If you haven't already — there's the QR code if you want to pick up the schedule. Nothing has changed since yesterday, but it's there if you want it. And also make sure that you're providing your feedback. Again, very valuable to the speakers, very valuable to the team at IT Revolution to make sure that we're keeping these sessions good quality and what you would expect to hear.

We have a networking break at 10:55 today, so if you haven't had a chance to go over and talk to our great sponsors who made this happen in short order, it's an excellent time to do that. Also during lunch, we have demos, sessions, roundtables, discussions like that. So please head over there. Lunch will be at 12:40.

And so yeah, that's it for me other than — oh, sorry, one more slide. Yeah, there we go. The sponsors that I mentioned. Sorry, didn't put that up, but you all see them. We're in the one room, pretty easy to get there, no conference hall. Expect that there. So with that, I will hand it back over to Gene, who will bring up our first speaker. Thank you.