How to Be Indispensable in a Post-AI World - Micheal Carducci

This talk addresses the challenges and opportunities for developers in the post-AI era, led by speaker Michael Carducci, who reflects on his extensive coding career. He shares personal anecdotes, including a transformative moment from his early career when he transitioned from coding to architecture. Carducci emphasizes the importance of human judgment and unique value over mere coding skills, as AI increasingly automates code generation. He illustrates how developers can leverage their diverse experiences and skills, particularly soft skills gained from non-technical pursuits, like magic, to enhance their roles and remain indispensable. The session concludes with practical advice on navigating the changing job market by focusing on problem framing and architectural thinking, urging attendees to recognize the value they bring beyond just writing code.

One of the nice things about being independent is that it means that there doesn't always have to be two 8:00's in a day. I'm a firm believer there should never I'm kind of iffy on two 9:00's in a day, but today is different because I am so excited to spend this morning with you and talk about a topic that is very near and dear to my heart.

Something that we're all dealing with. Something that we're all struggling with. Something that we all must navigate. Put very simply, how do we remain indispensable in the post-AI era? Now, for those of us who I have not had the opportunity to meet personally, my name is Michael Carducci. As Kate mentioned, I do a lot of things. A lot of those things revolve around code. I've been writing

code since the late 1980s. And I am horrified to look around the room and realize that there are many of you who weren't even alive back then. So, thank you for making me feel old. But with age comes wisdom. I've spent a lot of time on my craft as a coder. And you know, things sure are looking bleak for developers these days. Are you hearing the talking

heads? They keep predicting that we are going to go extinct like the dinosaurs. I went through this a while back. I learned that my coding skills didn't matter anymore. Maybe it sounds familiar. I'd spent years, decades polishing my craft. I was good at my job. I was good at what I did. I was really good. All those years, I wasn't just doing the same things over and

over and over again. I was not that developer that had been around for 10 years, but really just had 1 year of experience that they repeated 9 more times. Every year I grew. I spent years mastering my craft. And then one day I realized were obsolete. Practically overnight. One day I wake up and they said, "Michael, thank you so much. We don't need your code anymore." So,

that was 15 years ago. Well before generative AI. So, what happened? I became an architect. I got a promotion. That's what happened. My job changed from writing the code to making sure that the code that was produced coalesced into a cohesive whole. It was my job to make sure that the system was greater than the sum of its parts. The code didn't come from me anymore. At

that point, it came from other people. Implementation teams. Today more and more of that code is coming from machines. But the reframe is this. I didn't become obsolete. I moved up the value chain. What happened was my employer realized that the most valuable thing that I brought to the table, the most valuable thing that I could be focusing on wasn't the And that wasn't a sudden shift.

That was always my job. See, they realized that I should focus 100% of my time on something more important. The job hasn't changed at all. Even when my job was to push the buttons and produce the code, I was never getting paid for lines of I was getting paid for my judgement. The code was just how I expressed my judgement. You see, the code that we produced

was never the value. The code was just the mechanism that we used to shift value. And our uniquely human ability to think. That's how we decide what value is worth shipping. What value is going to pay dividends in the long term instead of being debt that we'd have to pay down. So, let's look at what's actually happening. Because if you're nervous about this, if you have anxiety

about your future, the fear is real. But the diagnosis might be wrong. Like, I spend some time on LinkedIn now and again. For better or for worse, against my better judgement, it's kind of a noisy place. But occasionally nuggets of insight slip through. And I saw a post from a gentleman who had a conversation with a young developer a couple years in industry. And he he relayed

that conversation. He talked about what this young developer had said. He said, "The skill that that developer had spent years developing was now commoditized. The thing that he was good at, that made him feel special, no longer unique. Anybody could do it. Now your boss can write code a feature. And after that, he said, "That realization empty. Not scared, not angry, And that's a real, valid feeling.

Because if we're honest for a second, if we're truly honest, watching your craft that was unique to you, that you'd spent all of this time mastering, suddenly get automated away, that doesn't just threaten your paycheck. It threatens the entire story that we tell ourselves about who we are. our identity as a human being. Well, I want to reframe that a little bit. And I know this is

challenging because nobody ever built a road map for that. But here's the thing. have always had so much more to offer than the just code. I figured that out 15 years ago, and we're figuring this out right now. And I want to prove it to you. Let me tell you the story of my first job in tech. As Kate mentioned, I am a professional magician. I traveled

the world doing shows. I've won international level awards. It's been my other career. And in fact, for a long time I thought to myself, my ultimate goal was to be a full-time professional magician. And I've done that. It was fun. That's only a part of me. The code is just a part of me. I like putting all the parts of me together. And putting all the pieces

together to be able to offer something to the world, that is everything I bring to the table. And that's why I'm so delighted to be here with you. But back in the day, back in the late 1990s, early 2000s, I had gotten my big break in show business. I'd been a magician doing little shows here and there, little house parties, a couple birthday parties. I don't like

to talk about those. I'm very bad at birthday parties for little children. But I'd done whatever gigs I could get. And finally, my big break came. I was hired as the house magician at a restaurant called TGI Fridays. Cuz dreams really do come true. And it was amazing. It was the first time I had a regular job. They said, "Your job, every time you come in, is

to do magic for people." So, I'd been doing this for quite some time. And I was beginning to realize that the resident magician at TGI Fridays didn't really pay the kind of money that would support the lifestyle that I was hoping to become accustomed to. I had a conversation with my mother. And I said, "You know maybe it's time I get a real job." Oh, she was

delighted to hear that. You have parents like that, too, you know. Look at the AI. You should have been a doctor. I can hear them right now. She got up, gave me a big bear hug, and she said, "Honey, I'm so happy to hear you say that." And then she said, "You know, there's a man who attends my church. You should meet him. He runs a software

company." So, I go, I get online that night. This is before Google. So, I was on AltaVista or Lycos or one of those searching for him. We hadn't started verbing the nouns yet. So, I didn't Google him. I didn't AltaVista him. I just searched. And I found his company website. They looked like a pretty thriving gung-ho software development consultancy. They'll build whatever you need them to build.

It looked like a cool place for me to start my career as a professional software engineer rather than just a hobbyist. And then I kept clicking. And I found a website that was buried in the depths of his server. It was his personal homepage. You see, back in the day before we had social media where you could just upload a picture and type some details in, and

suddenly you have a profile. In my day, you had to handcraft that HTML. You were picking your tags from the field by hand like the pioneers, assembling them in a text editor, FTPing them to a server. That was your social media profile. We built it all by hand. And he had done this as an exercise to teach himself HTML because at that point he was pretty sure

this World Wide Web thing had some legs, and it might stick around for a little while. But he didn't just half half-heartedly do this. He really put a lot of effort into creating this profile with photos and biographies and stories about his life and about his family, about his wife, how they met. His whole life story was in here. And I read all of it. And then

I went to bed. I woke up the next morning morning. Certain values of morning. It wasn't It wasn't 9:00 or anything like I wasn't due to go to TGI Fridays until 2:00 p.m. So, I got up around noonish. Gathered up my deck of cards and everything else. Went to the restaurant. Going around table to table doing magic. And I walked up to one of these tables. I

introduced myself. It was a larger table, a big family. I said, "Good afternoon. My name is Michael Carducci." This is my regrettable opening line back in those days. I'd walk up and I'd say, "Good My name is Michael Carducci. I am the wandering magician. I was wondering if you'd like to see a little bit of magic." Now, that was a terrible line. I regret it to this

day. But, here we are. Martin the head of the table, was very polite about all of this. He said, "Well, good afternoon, Michael. My name is Martin. This is my wife, Grace. My daughter, Hannah. My daughter, Vanessa. My son, Michael. And my son, Jason." And as he was rattling off those names, bells were ringing in my mind. They were all so familiar. And then it dawned on

me. That was the guy. That was the guy that I needed to approach about a job and I had just spent that evening learning everything about him. I didn't say a word. I did my first magic trick. The whole family was floored. Martin, an engineer, thinks. EVERYBODY ELSE SAID, "WHOA!" MARTIN did Huh. That's usually the best reaction you can get from engineers, by the way. Regular people

scream, jump around, say, "Whoa!" I've seen I've had somebody do a backflip as a reaction to a magic trick I've done. Engineers, the best you get huh. It'll be years to figure out that that sound huh is the sound of something breaking. And I looked at him. I said, "Well, that's a typical right-brain reaction right there." It tells me a lot about you. A guy like you

probably studied, I don't know, natural sciences or something, maybe somewhere like Oxford or Cambridge. Like, I knew where he went to school and what his degree was in. But I'm playing with him. I'm milking this for everything it's worth. And for a moment, he was shook. And then suddenly, the engineer kicked back in. He says, "Yeah? What did I study?" And he spat the challenge at me

like he was saying, "Checkmate, magic man." pause and I hem and I haw and I look at his shoes and I look at his hands and make a real show of this. Say, a guy like you, I don't think you studied something like botany. To you, botany's just applied biology. And in fact, to you, biology's just applied chemistry. In fact, I bet you look at the world

like chemistry is just applied physics. That's it right there. You have a master's degree in physics from the University of Oxford. And now he was broken. His wife, she you're very perceptive." I said, "Grace, your accent." She said, "Yes." I said, "You're from the Philippines." She said, "Yes." I said, "But specifically." I was really pushing it at this point. I don't know how they caught didn't cotton

on to what I was doing. I said, "But specifically, you're not from Manila." She says, "No, I'm not from Manila." I said, "I want to say Davao City." And she her face betrays her. There's a little grin because she knows that I'm wrong. But I'm not done. I said, "But I don't think you're from Davao City. Maybe the Davao region." And then I named the village where

she grew up. I stayed at that table for over an hour. Their food arrived. They did not care. This was the greatest magic show they would ever see in their life. Some random guy at TGI Fridays who knows everything about them. It was inexplicable. The food showed up. It got cold. They didn't care. Three times the general manager of the restaurant walked by and looked at me.

I didn't care. I was not going to leave until I deployed every single fact at my disposal. And when I'd finally run out of information, I said, "Well, I should probably let you get to your meals and I should get to the other tables. But before I go, and this is exactly what I said to them." I said, "I happen to be a keen amateur programmer. And

if you ever had an opening for a junior developer or even an intern, I would love to have a conversation with you." And this is what happened. He stood up. He looked me dead in the eye and said, "If you're half as good at that as you are at this, you're hired." And he hired me on the spot. I didn't want to in that moment explain how

I did everything that I did. I wanted him to sit in the astonishment for a little bit. And then I started working for him, and everything was going so well, I didn't want to kill the illusion. And then enough time passed that I realized it's too late to explain this to him. I've been in this world for too long. So, I've never actually told him the rest

of this story. So, I don't know, Martin, if you're watching this, sorry. Thank you. But that's how I got started with my job, my first job in tech. And there are actually some really important lessons that we need to learn here. And one of the important lessons is you should always make sure when you're starting a keynote that your laptop is plugged in. a very alarming warning

right now that is informing me that in fact my battery is about to die. But the the key takeaway here is there's not only is it just kind of interesting but how I navigated this entire process. How I got this job, because there are lessons that we can take away from this whole thing. See, how I got this job, and this is important today when the job

market is a little iffy. It wasn't keywords. I didn't wrestle with an ATS. I made a genuine human connection. And that's one of the secrets of navigating this job market. Essentially, I leveraged something else that I brought to the table. Now, the takeaway for navigating this job market isn't be me and go study magic for 40 years. use what you have. And in fact, the job market,

I'm not sure what it's like here in the United States. It's been a little iffy historically. I I run a YouTube channel and I try to make interesting videos that are fun to watch, but if you pay attention, there's little nuggets of wisdom that you can take away and will make you better. And we did an episode trying to offer a field manual for developers to navigate

a broken hiring And that was one of the big lessons. We interviewed a recruiter. It was one of the big lessons. Just make that genuine human connection. Because if everything is automated this everything is trying to scale, do the things that don't scale. And I took an inventory of all the skills that I have and I would encourage you to do this too because you have them.

You've just never connected them. I started to connect the skills that I learned doing magic. One of the things I realized is being a performer made it easy for me to be client-facing. I had soft skills to be able to read the room. But as I began to accept that magic wasn't just over here and coding over here, but they're all connected and all knowledge is connected

and everything you've learned, everywhere you've been, everything you've been through is connected and helps shape who you are. It helps shape the unique set of skills that only you possess. that my other career gave me more. See, magicians have this concept of second sight. This is the phenomenon of being able to see without eyes. I have a friend of mine, he's he and his wife do a

second sight act and they are genuinely and without hyperbole the best at it in the world. And the way they do their act, his wife stands on stage blindfolded. There are no electronics. They are scanned. People will bring RF readers. They'll look in her ears and her hair and everywhere else. There's nothing. Just their raw talent and abilities. And he'll walk out in the room and you

can just offer him anything. Anything in your pocket, anything in your bag. If you pull out a credit card, he'll just hold it and she'll say, "Oh, it's a MasterCard. It expires in November of 2030. The last four digits of this." If you bring out your driver's license, she'd say, "I think your date of birth is in March." And she'd tell you your date of birth. It's

very incredible what they do. I feel I would be remiss if there wasn't a little bit of magic in a Carducci keynote. Would you like a quick demonstration of second sight? Oh, thank you. One person clapping. Well, we're going to do it anyway. And Scott doesn't know this yet, but last year I gave a keynote and we opened with a magic trick. You'll notice that Scott over

here is looking slightly befuddled because I set him up to fail while everything that I was doing was working perfectly. I feel that it would only be fair to turn the tables or to allow Scott to turn the tables on me. And so, even though he saw a run through of this keynote last night, this slide was not in there. Would you join me in giving a

warm round of applause to welcome to the stage Mr. Scott Davis. I'm not sure we are on. Oh, you got one. >> I need better friends. That's all I'm going to say. Scott, so delighted. So, we're going to do a simple example of second sight. I've got a deck of cards. They're all there. They're all different. I'll even shuffle them. Because I feel like showing off, I'll

shuffle them without looking. Scott, your job is simple. Going to run my thumb my finger down the side of the deck. Your job is to say stop whenever you feel like stopping. Stop. Right there, you got it? Yes, sir. All right. Now, you're not looking at the you're not looking at this one, right? You you're thinking of one in the middle? Okay, great. Wait, wait, wait. What?

You think it's all in the steel trap. We can drop those. Now, Scott, I'm going to try to see through your mind. In fact, I'm going to try to see through your mind's eye. I want you to imagine the card is hanging in the air in between us. It's facing you. Okay. Okay, you can see the card? >> Yes. I want you to picture the bright vivid

colors on that card. Got it. Okay, that actually tells me something. It tells me it's a brighter card. It's not a black card. It's not a club or a spade. It's a heart or a diamond. Let's see what else we can do. I want you to picture the numbers in the corner. Picture the numbers in the corners here and here. You can see them here and here?

Yeah, yeah. >> Okay, and you know what? He did not hesitate there, which tells me that it's not a picture card. It's a number card. So, it's not an ace, a king, a queen, or a jack. It's one of the number cards. Okay, picture the spots. Count them in your mind. Don't don't move your mouth Okay, okay, okay, okay, okay. Around the middle. It's not a high,

it's not a low, it's in the middle-ish. Going to say it's a six. Is that correct? Oh my goodness. Yes. >> All right, one more. We need the suit. We know it's a red card, but it could be it's a six of either hearts or diamonds. Just say the suit in your mind over and over again. Just keep saying it over and over again. In your mind,

just say it. Diamonds, diamonds, diamonds, diamonds. Six of diamonds. Now, before you go Wait, there's more? Haven't I been through enough already? You know, I I I feel that some of you might be lying awake thinking about this magic trick tonight and thinking, well, you know what? There are 52 cards in a deck, there are 52 weeks in a year. Maybe what I did was I've been

doing this every week for the last 12 months. And I'm lucky it's just my week. Do we want to go bigger? Of course. All right. Instead of thinking of a card I want you just to think of a word, a random word in the English language. Well, actually before we do that, let's mix it up a little bit. Because depending on who you ask, there is between

two and 800,000 words in the But if I said think of a random word, you're not going to think of one of 800,000. It really But it really depends on your particular perspicacity. the other thing we have to recognize is humans are generally bad at being random. So I want you to think of a word that only you would think of in the English language, something that

means something deeply personal to you. Think of it, don't say it. And we're not going to use that word because there's an even smaller number of words that might be significant to a person and an even smaller set still that might be significant to Scott. So we're going to use that as our seed in the entropy pool. Scott, do you have your phone? Yes. Pull it out.

Open up a browser. Go to Google or Bing or DuckDuckGo or wherever it is that you go to Google Cuz we're going to use that word to get a random to to get a article that only you would look at at Wikipedia. But first I want to find out go to Google find out how many words and articles are in Wikipedia. How many words? >> How many

words and articles are in Wikipedia? You want me to Yeah, just search that. DuckDuckGo, good man. There goes that privacy. Cuz we're going to do this under test conditions. Almost, sir. No pressure. Talk amongst yourselves. All right. What do we got? Uh Aha. Yeah, there's first result. Okay, so this is saying an average article has 1,690. doing the maths. 1,690 words in an average article and that's

times 7,172,368 articles. And that means in English Wikipedia alone, there are 12.12 billion words. We're going to get one of those at random. Go to Wikipedia. Hit search. Bring up your article. So type in your personal word. The word I was thinking of? Yes. In the search there. Don't let me see it, but you're on the Yeah, yeah, yeah, yeah. Yeah. So you're in Wikipedia, you're searching

to go to that article. And just let me know when you've got that up. scroll somewhere random in the middle. >> And find me a good chunky word. Something with some heft to it. Seven, eight, not an easy word like the or an. Something like seven, eight, or more letters. Doesn't have to be the first one you see, but just find a good chunky word somewhere in

the middle. Let me know when you've got it. I've got it. Okay. And you can put your phone away. We're done. Just just You didn't write it down. You didn't whisper in anybody's ear. And even if somebody's looking at your shoulder, they might have picked a different word. Okay? Okay. So, let's do the same thing. Think of the first letter of that word. Got it? Okay, look

at me. A B C D E F G H I J K L M N O P Q R S T U V W X Y Z. I got two strong hits, one around R S T, and one around the letter C. That was the strongest. Did your random middle of the article word start with C? Uh no, sir. Oh, your other word started with C. Oh,

that's why. Okay, what did your with? Do you want me to tell you? >> Yeah, just the first letter. R. Okay, that was why I got the second hit on R S T. Okay, imagine the word is hanging in the air. Okay. Now, shuffle the letters up in So, now they're all mixed up. Yes. And I'm looking at them backwards and mixed Okay, I see some familiar

shapes. Looks like a C, right? >> Yeah. >> There's an M in there. >> Yeah. There's a B D D. Can I Can I buy a vowel? An E, there's a couple E's in there. Uh I'm going to just go for broke. I'm going to say something like recommended, recommendation, something like that. Yes. Yes, that's correct. Without eyes. And just for funsies I suddenly got a hunger.

Was your secret personal word chicken wings? Yes, sir, it was. Give Scott a round of applause. Scott, thank you so Dang it. Next year, he's going to get one over on Anyway. So, being able to see things without seeing them, even though he just looked at a word in the middle of the article, he picked one. And we were able to pull that out. Magic actually gave

me something else because a lot of magic relies on deception. And when you are trained deeply in the deceptive arts, you gain the ability to see through This gives us something called first sight, which is way more valuable. See, first sight is the ability to see things as they really are, to see through the Not what you hope to see, not what you expect to see, not

what everybody else sees. You see things as they really are. And that's an enormous asset. That really did something for my judgement. Now, remember, you have different tools. I just want to show you that those different tools can come from unexpected places. Not saying be me, I'm saying be you. See, first sight clarifies my judgement. You have your own assets. And the people who are getting through

this shift right now, as we shift into the post-AI era, they all share one habit. They're focusing on their unique value proposition. If something is left field as magic can make me a better coder, the left field things that you do, that you've never connected before, give you something you that you can bring to the table, new tools in your toolbox, different mental models, different perspectives. That's

the key thing. That's why the people thriving are focusing on their unique value proposition. And this allows us to stop pretending that everything is fine and start being honest about what's actually changed. Now, one of the big things is the economics, and this is where the headlines are getting it wrong. counterintuitively, last I checked, software developer openings are beginning to rise. We hear about the layoffs. We

don't hear so much about this. You might be saying, wait, what? This is actually a counterintuitive principle known as Jevons paradox. Which basically states when something gets cheaper, we don't use less of it. We use more of So when code gets cheaper, we don't write less of it. We use it for everything. And suddenly, companies are able to tackle projects that they've never been able to afford

before. Teams are leveraging technologies that they didn't have the expertise for But somebody needs to know what to build and what not to build. From a technical standpoint and all the challenges that are showing up, we've seen this movie before. People are on LinkedIn bragging, Claude code writes everything for me. Cool. That was always the easy part. That was always the easy part. Code was never the

core bottleneck. It was just the most visible one. Now that's changed. Now we're seeing new bottlenecks that we've never noticed They're becoming visible. I gave a talk earlier this week called the gold flow architecture and it's all about that. It's all about how we speed up code, everything else slows down. How do we navigate that? That is something truly valuable that you can bring back to your

organization right now. So what are the real bottlenecks? One of the big ones is figuring out what to build. We've got all this capacity. It's like, oh, we've never been rigorous about ideation. Another bottleneck, just being able to convince the stakeholders that the right that this is the right thing to build. That's a human skill. Deciding what not Now everybody's getting really excited right now about writing

big gigantic specs for their coding agents. Anybody who's got a little bit of gray hair like me remembers the last time we went down that road in the 1990s. And we realized that it was really problematic for a number of reasons. And a bunch of smart folks got together in some cabin somewhere and hammered out a little thing called the Agile Manifesto. And that was a response

to the challenges with that approach. New ways to navigate this, to move swiftly, and to adapt quickly. We've got better tools. We need to apply the wisdom that already exists in our field. So, we figured this one out 25 years ago, and the problem hasn't changed because building software is more than just deciding what it should do. One of the big ones that is coming right into

the forefront is we need to be more vigilant about what it shouldn't do. We're learning there suddenly more and more rounds of requirements and requirement refinements. Legal, security, compliance. These things are mattering more than ever. And if we're not applying this and bringing our judgment and our experience and our expertise, then all we're really doing with LLMs is we are bringing a high-performance automated CVE pipeline into

our production environment. So, let's reframe the current reality. Code was only ever about 10% of our job, but the reason this shift feels so weird is because it took 90% of our time. AI's eating that 10%. Boy, when you look at it like that, Dario here is almost right. AI will replace 90% of the stuff that we're doing, not 90% of developers. And that hands us capacity

back for that 90% So the big question is what are we going to do with that 90%? I read a very influential book in my life called deep work. And it talks about deep work being the most important work that individuals can be doing. It's the stuff that's hard to replicate, requires deep concentration, it pushes our cognitive abilities to the limit. And he talked about what's in

the way of deep work, what he called shallow work, the non-cognitively demanding logistical style tasks that we tend to do distract distracted. This is the stuff the AI is going to do brilliantly with. It means we get to shift to do the most exciting stuff the deep work. And for this to flow we have to understand the system and not just the software system. Let's look at

some of the lessons we learned. We learned a lesson in manufacturing back in the 1980s. This isn't the first time we've seen automation come into a career. In the 1980s we started having industrial robots that could do certain parts of the work. The robot appears. And the thing is the factories added robots to a station here or a station there. That station got faster but the system

got slower. We didn't eliminate the bottleneck, we moved it. So if code generation is 10 times faster now but our review capacity hasn't changed or testing capacity or design capacity or ideation capacity all we're doing is swamping the new bottleneck. We don't want to do that, we want better flow. And that's one of the lessons that we learned that you can take back from this conference that

an isolated acceleration always makes the system slower. But here's the secret. None of the problems in AI are truly new. They're just louder. I was I shared a drive for about an hour with a friend of mine who was the director of software engineering for a company that you might have heard of. And he was talking about all these new challenges that he's facing in his role

and his teams are facing. I turned to him and said, "These aren't new. These are the same problems that have existed in software forever. You're just noticing them now. Welcome to the show." And we talked about a lot of this stuff. They're not new, just louder. Technical debt? We named that in 1992. The theory of constraints? How isolated accelerations actually slow everything down? We figured that out

in 1984. What about the accidental complexity that we're that we're introducing with AI? Fred Brooks warned us about that in 1986. We're finding the limitations of natural language programming and everybody's trying to do better prompts and better prompts and better prompts and suddenly we actually have engineers spending more times on prompts than it would take them to write the code. Well, we saw that one coming in

1975. Here's the beautiful thing. We've already solved these problems. Solved them. But unfortunately, we weren't ever feeling the pain like we are now. And a lot of us forgot the solutions. But that's okay, we get to remember. Now we get to focus on the harder problems. Now we can start to deliver durable value. Let me tell you the rest of the story from my first job. When

I started, I was expecting to go to an office and work with real people. I went to the address. It was Martin's house. His wife Grace greeted me. She took me to the office. Martin wasn't He was paying the bills with a consulting engagement. There was just a note on the keyboard. We had an idea. And I was so eager to do a good job. And nobody

was telling me what to code. So, I went to figure it out. I started learning about the domain. I started watching how our customers actually worked. All software, whether it's an app or an API, needs to make somebody's life better somehow. Software always serves people, humans in some way, in some capacity. So, I learned about the people. I learned about the domain. And a couple of years

into that job, he says, "Michael, you're no longer my junior developer. You're my 50/50 partner." And we were partners after that. And then I definitely could tell him the story, you know, about how we met. So, that's the key. That's what you're doing, and that's why I'm so thrilled that you are here this week, at this event, at this sacred space with all of these minds and

all of these minds, because even you are bringing magic and wisdom and knowledge and experience to everybody else in this room. It's all of us together. That's what we're doing. We're architecting ourselves right now, because when code is cheap, what's priceless? Things like problem framing. This is Kent Beck, creator of XP and test-driven development and JUnit and a bunch of other important things. He calls AI the

genie. He says the genie focuses on plausible completion of tasks. You're the one who knows what might not work. That's your job now. Kent Beck nails it. You think, the AI completes. You question. AI optimizes for completion. You optimize for the right problem. And like Dijkstra taught us, it's more than a better prompt. It's better thinking. Natural language is ambiguous, contextual, and full of unstated assumptions. And

everybody else right now is building production systems on exactly that. The precision didn't disappear. It was just deferred. And deferral has a cost. Your greatest asset is investing in architectural thinking. This is your mental model. Your mental model is the real program. This is Peter Naur. I think I'm saying that right, probably not. He wrote a very important paper that's more relevant than ever called programming as

theory building. And he says, "The program is not the code on the disk. The real program is the mental model held by the people who built it." You can't delegate that. And when you're close to the code, whether you press the buttons yourself or not, you inhabit it. You know how the argument proceeds. You know where the tension is. You know where it doesn't quite hold together.

When AI writes the code, and humans are out of the loop entirely, which is what some people are trying to getting mixed results. Nobody inhabits The mental model is even more important than ever. This is architectural You need the theory even more when you're not writing the code. Now, the thing is we all have what I'd call a lower bound. Right? This is the level where you

stop caring about anything below the abstraction. Right? You start stop caring about the mechanics. You just trust the Like in Scott's keynote, Grace Hopper was debugging a system by looking at the machinery, and she found a literal bug breaking a relay, a moth, the first computer bug. And today, we don't verify voltage gates when we're debugging a React app. See, AI, when we use it wisely, allows

us to push that bound higher. Now, the danger is having no lower bound at all, where you're just delegating everything and you're not sinking. And when that happens, you're not at a higher altitude, you're just drifting. And these are the people who AI is going to replace. Not you. We're all architects now. So, what is architecture? It turns out decades and decades and decades ago, somebody figured

out fundamentally that architecture is constraints. Two somebodies, in fact, Perry and Wolf in Foundations for the Study of Software Architecture. That work was built on by Roy Fielding this 2000 doctoral dissertation, Architectural Styles and the Design of Network-Based Architectures. Both of them referenced Fred Brooks in 1975 in The Mythical Man-Month. In 2016, there was a great paper called Software Architecture Constraint Re-Design Design by uh Re- Software

Architecture Design by Constraint, Reused by Composition. It's in my book, Mastering Software Architecture, building on that lineage. And fundamentally, architecture's all about constraints. You're constraining the degrees of freedom to induce desirable outcomes. This is exactly what we need to be doing with AI agents. Architecture's not lines and boxes. We're constraining the agents, the same way as an architect I would constrain implementation teams. I'm sorry, no, it's

not a free for all. You can't have a 30,000-line god class, let alone five. You can't make everything global. We're going to constrain that a class has a single purpose for existence. We're going to constrain that classes are open for extension, closed for modification. That induces understandability, maintainability, testability, extensibility, things like That's what architecture is. So, we're deciding which constraints matter most for this system at this

moment given these stakes. This is the context that doesn't fit in the context window and judgment that the AI can't replicate. This is the stuff the AI can't see. AI can generate code that compiles and runs, but it doesn't understand architectural thinking. Only learn patterns devoid of the context of your system and your organization and your team structures and how people communicate and where the real bottlenecks

are that are outside of the machine, the reality of that software exists within. So, the code might work even when the engineering fails. That's why you're priceless. That's why you're irreplaceable. You can vibe code a feature, but you can't vibe code a complete production system. Responsibility doesn't disappear. It concentrates. Yeah, it concentrates. When code production accelerates, accountability concentrates onto fewer people. Those people need judgment, architectural thinking

cuz AI doesn't get paged. AI doesn't have to explain failures to executives. It doesn't rebuild trust after an incident. Somebody still owns the consequences. And that person will always matter. Those are all the things that AI can't Is it you? I think it can be. I think you can be a force multiplier. Cuz everybody wants to be a 10x Or in age of AI, I guess a

500x So, you have people try to be the 1x developer that does the exact right thing one time at the exact right moment. They pick the right problem at the right time for the right reasons. And they stop worrying about vanity metrics like personal velocity. And they start thinking about the real metrics of delivering value, being a force multiplier. Because the engineers who multiply the team's output

are more brilliant and more valuable than solo artists who make everybody else slower. And trust me, I've worked with them. And I've been them. And I still feel bad about it. Force multipliers write documentation that nobody hates reading. They ask the questions that unblock five people. They return those reviews in 20 minutes instead of 2 days. They share their context because a senior engineer who can't transfer

is a liability. House lights on? Now we are illuminated. I I'm going to just assume that we did that on purpose and it was all about how this is the moment of illumination in the talk. So, let's uh talk about some ancient Cuz this stuff we're unearthing it. We're reviving it. And we're rediscovering the wisdom of the things that we never took the time to understand because

we never felt the pain until now. Test-driven development. I'm not talking about unit tests. In fact, if you go read Kent Beck's book Test-Driven Development by Example, do you know how many times he mentions unit tests? Once. One time in the entire book. Test-Driven development isn't about unit test, it's not about checking some code coverage box somewhere in a report. It's a way of thinking, it's a

way of working. It's a discipline. And it's even more valuable than ever. Not because the methodology changed, but because the failure mode changed. When human wrote humans wrote the code, we could skip tests and it just meant slower feedback. But when AI's writing the code, skipping tests means losing control. Going back to that idea of constraints. Other ideas that were pioneered from Kent Beck, working in small

batches, pair programming. I pair program with my agent. And I don't mean coding with an audience while it codes and I watch. That's not pair programming. That's just um I don't know what that is. I don't have a clever aphorism for that yet. Uh I do strong-style pair programming, driver-navigator. So the driver's the one programming, the navigator's the one typing. So I am the Or sorry, the

driver isn't typing, the navigator is programming. And so I am navigating my agent, and then it's doing the work. And then every now and again it gets stuck in that loop, and we've all been in it, where it's like, "Okay, all done." You say, "It doesn't even build." "Oh, great catch. Now you're thinking like an engineer." It's just like, "Don't do It's definitely fixed this time. Did

you not notice the build failed again? "Oh, I see what I did." And then you do rounds and rounds and rounds and rounds of that. It's like, "Oh, it's definitely fixed this time." "Are you sure?" "Oh, good catch." And then it suggests the thing that it did four iterations ago, and you're in an infinite loop now. That's when we switch, and I switch my copilot into ask

mode, and it's navigating, and I'm driving. And we switch it around. Continuous integration. See, the teams that are succeeding, that are picking all this up the fastest, they're the ones who already did these things because these habits map directly onto keeping the agent reliable. Small context, tiny tasks, quick feedback loops. You're generating 10 code 10 times faster. If you're generating 10 times more code, guess what? You

need to integrate 10 times more often, not less. No 100,000 line of lines of code diffs. We learned decades ago that that was risky. That was insanely risky, and it's even more risky now because we don't know what's in there. That is inherently unknowable. Smaller batches The companies that are panicking about AI are the ones that never really understood software engineering in the first place. Right? The

problems aren't new. We just need somebody who remembers the answers. And that someone is you. See, AI doesn't accelerate, it amplifies. AI doesn't replace, it reveals. If you add AI to clear thinking and sound engineering, you get leverage. If you add AI to a messy mind, you get acceleration of noise. AI doesn't create structure. It's just going to amplify whatever's There are things that AI does better

than me. That's exciting. It means I can be better. Especially when you realize that there are things that you do better than AI. We're not going to take humans out of the loop. The loop needs both parts, the silicon and the soul. AI handles my weakness, I handles AI's weakness. So practically, what do you do? Know your domain, not just a framework. Build those mental models before

you generate the code. Constrain your agents the same way you constrain a junior developer. Test what matters, not just what's easy to test. Invest in flow, not speed. And most importantly, don't confuse business with progress. It's a bad metric. It's a vanity metric. Don't confuse code with engineering. Your thinking is your greatest weapon. You can actually think. AI cannot. I know they call it reasoning mode. It's

just predicting tokens in a very clever and a very expensive way. It's not thinking. It's not intelligence. That doesn't mean it's not useful. Doesn't mean it's not cool. But it means the world still needs you. If you step up. If you don't let the AI think for you. Remember this, tools that are designed to make it so that you don't have to And that's what they're trying

to sell us, tools to delegate thinking. Hang on to that. All right, we've got to wrap up. We've got about 5 minutes left. To quote the great Dr. Venkat Subramaniam on this moment in time where we are, he said, "Skill up." Don't sear down. my job stopped being to write code. And I started instead making sure it all held together. code comes from a different source. The

job hasn't changed. I saw a great post on LinkedIn recently. Somebody was brainstorming a list of what should we call the opposite of vibe coding. We have vibe coding. That one took off like a storm, but what about the opposite of vibe coding? What do we call it? Crap You know, sense coding? crap coding? Uh they're they're brainstorming these ideas. What do you think? And then somebody

showed up in the comments and it had like a hundred times more likes and reacts than the original post. It was how about software engineering? There's nothing new under the sun, my Yours hasn't either. You just got You just didn't know it yet. I'm going to leave you with some words from a very good friend of mine who isn't here this year, but I'm hoping to drag

him here next year. This is Brian Sletten. He said, "If you think AI is going to replace your job, you're not paying attention. If you think AI won't affect your job, you're not paying attention." I want you to remember this as you go through the rest of the day and go back into the world with the knowledge and wisdom you've gained this week and the friendships and

the connections that you have made. Your future's bright if you're paying attention. By the way, PS PS shameless plug. If you like what I do, I like to tell stories. I like to share some little insights here and there. And both Kate and I have been working on very special. We've got a catalog of videos we've already already created. We've got more that are coming. You might

like this. This is my YouTube channel Dev Random TV. Uh we've already got some great videos in the catalog to watch. We've got more coming. I'd love to see you in the comments. I respond to every single comment. So, uh check out some of the videos. Check out the new stuff that's coming. Drop a comment. And I'd love to keep that conversation going. Just knowing the conversations

I have, I have a feeling you would dig this. So, check it out. Thank you so much.