Closing Remarks — Day 2

Host Intro (Gene Kim)

All right. So to finish this out, here is what I'm hoping you found the last two days to be. Just to put some context: for anything I spend time on, I really want it to fulfill a couple things. One is, is it high network? In other words, are you hanging out with the best in the game? Two, is it high learning? Meaning that you're learning something that's actually important to you. And three, is it high impact? It's not just interesting trivia. No, this is interesting stuff that is actually important to things that you are working on. So by a show of applause, to what extent in the last few days have you felt these three things?

Okay, fantastic. That means that we have done our job.

If I could see a slide next on the benchmark monitor, it'd be great. So let me attempt to try to maybe describe why and maybe what we're trying to do here, especially when you talk about some of these initiatives.

So I love this thing called scenius that came from Brian Eno. He's a musician, a songwriter, producer. He was the genius behind the sound behind Talking Heads, U2, and so much more. And scenius is described as this: despite heroic mythology, lone geniuses do not drive most scientific, cultural, business, or policy advances. Instead, they emerge from a scene — in other words, an exceptionally productive community of practice that develops novel epistemic norms. Major genius may involve genius, but the genius really is supported by a scenius. Epistemic means to create knowledge.

Number two is features — there are three of them. One is there's mutual appreciation. Risky moves are applauded by the group, subtly appreciated. Friendly competition goes to shy. Scenius can be thought of as the best of peer pressure. I think you saw that throughout the two days.

The other thing that I love is number two: rapid exchange of tools and techniques. As soon as something is invented, it is then flaunted, but then shared. Ideas flow quickly through the scene because they're flowing inside of a common language and sensibility. And I think you saw that again over the last two days. We saw practices, but also theory, presentations referring to each other. I think that is absolutely another amazing hallmark of a productive scene.

The third one is network effects of success. When a record is broken, a hit happens, a breakthrough erupts, the success is claimed by the entire scene. I sort of got goosebumps there. The artist is celebrated, but the entire scene that surrounds the artist celebrates with them, and this empowers the scene to further success.

And so I think, just by a round of applause, to what extent do you feel like those characteristics have shown up here over the last two days?

So let me just show you a couple of things that you might find useful. I'm just starting to talk about the projects to friends and family. But I've been working on the third volume of The Phoenix Project. So The Phoenix Project came out in 2013, The Unicorn Project came out in 2019. In the early stages of volume three, The Pandora Project, there are two notable features: Brent hates vibe coders. He hears about vibe coding, he's completely appalled by sloppiness. He'll just leave Zoom calls. He gets bright red. He rolls his eyeballs so hard that you can hear it. It's actually starting to become a problem. And Sarah is now at a private equity firm, and she wants to extract some value out of the company — and maybe also some revenge.

And so as part of the writing process, there is a lot of research that I've had to do. Even with my own books, I put them all into a repo, started putting them in, analyzing with Prolog. And here are some of the questions. One of them is: okay, who are the top characters in The Phoenix Project and The Unicorn Project? How many times do they appear, and in what chapters? What are the top characters that cross over to The Unicorn Project? And how many times does the bottleneck move in The Goal? And it's just amazing to be able to put things into this LLM wiki so that I can quickly answer these questions and do essentially database queries for all these things, because everything's represented as a Prolog entity.

And by the way, you might be thinking, "Oh my gosh, I heard Joe or Topo say something amazing — what was it?" And just to be able to retrieve — we have all the Slack, seven years of Slack history, all archived. Now you can do things like I did. Sometime about eight years ago, we were talking about aviation safety, and I mentioned how I love watching metro pilots, and someone said, "Oh, but if you really love this stuff, you'd love this channel." And I just couldn't remember it. So I put it into an agent. What was that channel? Turns out it's a Blanco Lirio channel. And so just the ability to retrieve not just this week's Slack events, but all of Slack going back — I think seven years.

And so I had mentioned the notion that peer pressure goads the shy. So when Philip Gollner said that he wanted to give every attendee their own chief of staff, we were like, "Gosh, wouldn't it be amazing to take some of these tools I had built to help write The Pandora Project and be able to give it to the chief of staff?" And the whole idea: we put the entire IT Revolution body of knowledge — including every talk, every video, every forum paper, all the guidance papers, and so forth — into an MCP, so that anybody can do what I did as I'm working on The Pandora Project.

And basically we put everything into it. And it's been so freaking useful. 64,000 Slack messages, every talk put into this LLM wiki inspired by Andrej Karpathy, and all of the Idealcast episodes — all the things that I've sometimes wanted to find but it's just too hard. Finally, this is something that I can retrieve instantly.

So all of the things — if you were using a chief of staff agent, it knew how to use this IT Rev MCP. And I meant to show this during the build thing, but I forgot. So should we just cue this video that the team generated about what you can use in the chief of staff agent? It's 45 seconds.

(video plays)

So if that is of any interest, just follow the instructions. The instructions will be in the Slack channel. If you find it even half as valuable as I've been finding it over the last two months, I'll be super, super happy. All right.

So it is free to be a beta tester, as noted.

So here is — I have "thank you," and then I had every speaker end with a slide that says, "Here's the help I'm looking for." I'll describe the help that I'm looking for. So first off, before we do that, I want to thank GitLab, Andy Ramirez, and all our fabulous sponsors who made this happen. So by a round of applause, can you just show your appreciation for everyone that made this happen?

I also want to thank the entire IT Revolution team, especially Marguerite Kim, my boss and wife, and the amazing staff at the Carolina Theatre who have made it possible. Most of the videos are available now for you to share with your colleagues and friends. And all the amazing speakers — and also all of you. Speakers hanging out together is fun, but it's all about you who have attended. So thank you to all of you.

So, here's the help that I'm looking for. Over the last two years, the goal has been to assemble the most switched-on people using AI and agentic coding, who are using it to transform their organization, to move the needle on the objectives that really matter. And you've been hearing two days of fantastic experience reports from people who have been doing that. And yet, I think what eludes so much — and you heard this from Gallimore and Topo — is how we measure value is elusive.

And so I had mentioned the DORA report. This is the famous four metrics of how we measure throughput, reliability, stability. But I have to share with some people that what I find very frustrating is that what we don't know how to do yet is measure the value that this is creating. We're not looking for the 20% increase in throughput that so many studies are showing. We know there's something bigger, as Topo and Jeff said in their reports — we know it, we feel it, but how do we quantify that?

And so one of the things I'm super excited about: I've been working with a lot of technology leaders who have been doing these incredible things with these switched-on teams, generating 100X the output, doing things in weeks that normally would have taken years. I was just hanging out with five teams at Adidas that did this, and I think we have a hypothesis. And so we're in the early stages of a research project to see if we can actually show measurements that can finally completely express the awe that we have in terms of how much value we can actually deliver.

And so we're in the early stages, working with some cracked PhDs on this thing called Intuit GitLab. And so the goal is to actually show in numbers things that we can show to maybe CFOs or CEOs about the value we're genuinely creating. And actually touch on the value thing, which is something that we deliberately sidestepped in the DORA days, just because value is so elusive. If you've read Zen and the Art of Motorcycle Maintenance, you know how difficult it is — it's the road to insanity. But I think we actually have a — the signal is so strong that I'm so confident we're going to find it.

So if you are interested in taking part in that early research, where we'd ask for anonymized data from your Git repos and maybe some surgical surveys, I would love your help. So that's the first bit of help I'm looking for.

Number two: Self. A lot of you have talked about, "Hey, I have an experience report." A gentleman from MassMutual said, "Hey, would you be interested in an experience report of what it was like to roll out coding agents to hundreds of thousands of engineers and share lessons learned?" Yes. Go to Self, sign up, and we'll be conversing there. Do you have questions that you think belong on the list of questions that deserve better answers? Yes. Go to Self and write them there, and we'll discuss it there.

The discussion forum is invite only, meaning that we know you, and if we don't know you, you know someone that we know. And so that way we can have a conversation that we can hold under Chatham House Rules, that we can actually share with each other not worrying about whether they'll be leaked all over the internet. If you want to see Randy Shoup and me sort of helping catalog all the practices that we discussed the last two days, you got it. Go to Self. We'll be doing the work in front of all of you there, helping build a wiki of everything that we've learned and trying to make sense of all that's happening.

And so, just to put this into context — I had mentioned many of you — I've had a lot of fun in my career, but nothing can hold a candle to how much fun I'm having now. And this is the reason. There's so many problems to solve, and there's no better group of people to be solving them with than you all.

Andy, how am I doing?

Andy Ramirez

Gene, I have a little bit of news. As of today, we've actually hit a major milestone — we're now over 100 founding Self members.

Gene Kim

All right! Woo! So congratulations on that. Between us, we have the friends, the funding, the insanity — I mean, tenacity — to do it. I think we can get this thing to a thousand. So thank you all. I'm really, really excited for this. I think this is going to be an amazing community.

Andy Ramirez

To the folks that came today, I really appreciate it. It was a great conversation. I promise to keep certain things private. Everything you said will be on my pod only. But otherwise, really, really excited to get this done with you.

Gene Kim

Awesome. And so to evoke what I said before: when I first met Andy, he said, "I want to get to 1,000 technology leaders in a year." I was like, "That's crazy, but that's my type of crazy." So go to selfcommunity.co, and the adventure continues.

All right, to close up — all the talks, most of the talks are available online. The ones that are missing will be available as soon as they clear corporate communications from the folks. And the instructions are here as well as in the Slack channel. Share them, because that's why we do this.

And last thing — see you hopefully in Portland, Oregon, on May 5th and 6th in 2027. Go Trailblazers. Thank you.