Searching for Truth in the Age of Maybe - Scott Davis

This talk explores the intersection of truth and the uncertainties inherent in artificial intelligence, using the speaker's personal anecdotes and experiences with Google Gemini as a framework. The speaker begins by sharing a clever birth announcement, which leads to a discussion about the deterministic nature of computers versus the generative, sometimes fallible outputs of AI. They delve into the challenges and nuances of using AI in software development, particularly around the idea of establishing which aspects of a project require strict adherence to truth versus those that can accept a more flexible approach. The speaker provides a case study involving the creation of an accessible website, illustrating the collaborative nature of working with AI tools. By leveraging methods such as Retrieval-Augmented Generation (RAG) and Notebook LM, developers can enhance the accuracy and functionality of their projects while recognizing the complexities that arise from the 'maybe' of AI outputs.

Today we're going to talk about truth and maybe in equal parts. So to begin, I'd like to present you with a birth announcement. A birth announcement that reads on the top 1 + 1 = 3. Computer is outwitted. Barbie + Ted = Scott Howard Davis. This is the birth announcement that my parents sent out when I was born. And so I thought I would start with this

because as you're thinking about why would Scott be showing me his birth announcement, we can ask AI for that, couldn't we? We can take any image and upload it to AI and ask it to explain it to us. I'm using Google Gemini in this presentation, but really the AI engine that you're using, the LLM I'm using, I don't think is the important part of the conversation. But

when I asked Gemini, "Help me understand the joke on this birth announcement." Gemini says, "This is a wonderfully clever birth announcement, especially given the medium. It is printed on an old IBM punch card, which is central to the joke." My parents met at IBM. They were working at IBM when they sent out this But the setup of this joke is what I'd like us to consider. We

know that computers are built on strict deterministic logic. To a computer, the mathematical sequence of 1 will invariably and always equal two. So, if you have a calculator that says to you 1 + 1 = 3, what do you suppose you should do with that calculator? Throw it away, of course, right? That's a pretty significant bug right there. Perhaps not a bug that we could pull out

of the calculator and tape into a logbook with a handwritten note under it that says the first actual case of a bug being found. But if we ask Gemini about this as well, why is that joke funny? And Gemini says that image shows one of the most famous artifacts in computer science. That is the bug that Grace Hopper literally found in her computer back in 47. This

idea is not new, but we're experiencing it in new ways in this age of AI, aren't we? This idea of bugs, this idea of logic, this idea of determinism, and generative processes. So, what we're going to talk about today is how do we balance this? How do we deal with the age of truth and the age of all at once in your day-to-day job? Because we like

to think that what we do is based on logic, is based on truth. There's something deeply satisfying about holding that card, and that's the physical manifestation of truth. Quite literally, you can hold that card and shine a light through it. And only the truth is what comes through in the holes of that card. That's what made that moth, that bug so significant. If you can imagine a

moth being pressed inside of two different punch cards as it's being run through the system, is literally blocking the truth. how do we rectify that in this age of AI, in this age of generative engines, hallucinations? Certainly, if your calculator tells you 1 + 1 = 3, you need to throw it away. If you threw away your generative AI engine every time it hallucinated, you would very

quickly run out of generative AI engines, wouldn't you? asking Gemini once again, are punch cards a reasonable metaphor? And Gemini says, "Yes, in fact, punch cards are the perfect end point for this keynote cuz they represent two extreme opposite ends of the spectrum." Do we only look for truth? Do we only look for maybe baby? We have people who will tell you both, right? We have people

that'll tell you, "Oh, we can never use AI. It hallucinates." They'll tell people, "Oh, you can't use that old deterministic software. Go home, grandpa. It's late." Yeah? have I set up a false dichotomy for you? Have I set up an argument that can't be answered? I hope not. Otherwise, this is going to be a very short keynote. Yeah? So, in fact, it is the conversation's worth having

right now. In fact, this might be arguably the most critical conversation that's happening in software architecture right now. This is why you come to conferences like this. To think these big thoughts. You don't have story cards waiting for you here. You don't have the pressure to ship code. This is your chance to think and consider and evaluate what the architectural options you have as a software engineer.

nuance is perhaps the message of the day here. How do we, in fact, embrace both? And how do we use this as an engine to evaluate our architectural decisions? How do we use this to actively decide which parts of our system require truth and which part of the system can be delegated to maybe. rather than giving you a bunch of theory, I thought we could explore a

case study, a real-world usage of this technology. So, let's begin. This is how I began with Gemini. I started a new chat and I said, "Excuse me. May you Can you help me? Can you help me build a static website? And it's important that this website is purely standards-based. Only HTML, only CSS, no JavaScript frameworks. Oh, and it also must meet WCAG 2.1 AA compliance. Doesn't that

just roll off the tongue like Shakespearean prose? WCAG 2.1 AA compliance. Ah. It's wonderful. WCAG is the Web Content Accessibility Guidelines. So, I'm saying, not only does this website I want to create with you, it needs to be standards-based. It also needs to be fully accessible. So, after that pleasant introduction to Gemini, I said, "Okay, let's get down to work here. I'd like to build a home

page for this website first. There's this idea when we're using AI that in one prompt, I could say, 'All right, go off and build this entire website for me. I'll sit back here and drink my chai while you're working.'" If only that was the case. In fact, what you're going to find is pair programming. This is pair programming with someone doing iterative development, changing one thing and

making sure it was correct, and then changing another and making sure it's correct, and again and again and again. So, I'm not asking it to build an entire website for me. I'm asking it to just build a simple home page, if you would. And the domain name for this website is simpleaccess.io. Simpleaccess.io. That website is live and published right now, if you want to take a look

at it. We'll revisit this website quite a bit in this presentation. But, in addition to saying I'd like to create a website and I'd like to create a home page, I also need to let it know what's on my mind. I say the theme of this website is that accessibility isn't extra work or something extra that you have to include. In fact, accessibility is already right here.

Right now. So, I created that homepage. I published it and I said, "Oh, wonderful. Okay, Gemini, please create a QR code for this website, simpleaxis.io." It was a beautiful-looking QR code, wasn't it? Have any of you tried it? Does this QR code work? It does not. It does not. So, oh, rats. All right. So, let me try again. Maybe I didn't craft my prompt correctly. I just

asked it to create a QR code, not a valid QR code. So, I create I asked it again, "Will you website?" I actually like this graphic quite a bit until I pulled out my iPhone and tried to visit it and it still didn't work. That's when I said, "Are you sure that you know how to create QR codes? Because I've asked you twice and it hasn't worked."

But, in this prompt, hidden in this was me supplying success criteria as well. Inside of this prompt, I said, "Are you sure you can create a valid QR code because the last two didn't allow me to quickly visit this website on my iPhone." What was Gemini's response to that? create a working QR code. Oh, yeah, no, no, no, I can't do that. But, what I can do

is create an image of what it would look like if you were holding your iPhone out and you snapped a valid QR code and went to the website. Is that good enough?" I said, "Perhaps for now." But, I'm a professional software engineer, right? Like many of you. What do we do in the face of adversity? We Google it. We Google it. So, I asked Google, I said,

"Hey, Gemini create a valid QR code? And so, do you see the irony here? I'm asking AI if AI can AI all the things. Yeah. But, it said, "Okay, yeah, yeah, yeah. So, that standard Gemini chat that you were in, um yeah, if you ask the standard chat to make a QR code for this URL, um you know, it might give you some Python code or more

likely it's going to give you a bucket of pretty pixels and nothing else." But, then it went on to say, "Oh, you know what? In Google Chrome, you know, if you just kind of visit that website and right click, yeah, you can create a QR code that way." I said, "No, why didn't I do that?" I visited simpleaccess.io, I right clicked, and there it is, create QR

code for this page. This is what it generated. This is a working QR code. If you'd like to take a picture of this now, this is your chance to visit simpleaccess.io. It's going to figure prominently throughout this keynote. where Gemini fell short, Google Chrome came to the rescue, and it said, "Here's your QR code." And notice that I can copy it, I can download it as a

ping, I can do what I need to from here. So, here's my first tip of the day for If Google can't create a QR code, Google it, and then follow the Google instructions to create a QR code in Google. Any questions? I'm not making fun of Google. I'm quite impressed with their technology stack, but this is the state of the maturity in our industry right now. And

so, we need to have as much resilience and grace for the AI tools we're working with as we might with a fellow co-worker. A co-worker that we trust. A coworker that we enjoy pair programming with. we've got our MVP now, don't we? I've published a single page filled with broken links. If they're all working now, but this initial page filled with broken links, but it's published and

I have a QR code. So, let's figure out what our next steps are. Maybe we were going to lean more into the maybe of the situation than the So, I started with this. I said, "Okay, if you would, please, create a compelling hero image for this Here's the text of this." And so, once again, every time I'm prompting my pair programmer friend, Gemini, I'm setting the stage.

I'm setting the context. I'm saying, "Here's what I think should be evocative in this image." And this is the image Gemini created. Again, as nice as it would be that we could be one and done on all of our prompts, that's not how it worked. I said, "You know, actually, this is kind of more of what I had in mind. I'm thinking about bold colors. I'm thinking

about simple evocative icons." In fact, this is a slide from one of the previous keynotes I've given right here at Gids. And notice I said, "Here's my thinking. These icons, they could be used throughout the website for, I don't know, sidebar callouts or easily identifiable themed discussions." And so, Gemini thought for a bit and came up with this. I said, "Ah, that's much better. We're getting there.

I like that. tiny asterisk text at the bottom might be hard to read. That tiny asterisk text says, "Use these icons throughout your sidebar callouts and easily identifiable themed discussions." So, in fact, what Gemini did was just turn around and take the text that I had supplied it in its prompt and incorporated it into the image, but we are getting better. As Gemini are working together, as

our chat history gets longer and longer together, Gemini is learning what's important to me, and I'm learning what works with Gemini. I find that if I say, "Hey, create me a hero image." It's going to create a hero image with exactly that degree of precision. I wasn't keeping up my end of the bargain. My end of the bargain is saying, "No, what I had in mind is

this and colors and thoughts and icons and things like that." And it did get better. And so, the conversation continues. Like, "All right, you know, the whole theme is digital accessibility is sight, sound, and touch. And this is hearing. So, if you could change hearing from sound. And yeah, you can leave off smell and taste. There aren't any W3C specs for those. Yeah. And oh, yeah, well,

you know, be more creative. I was trying to inspire you, not tell you to do this exactly." And it came up with this. I said, "Wait a second. That's where we started the conversation. I literally say, 'I'm not sure why you would show me this image again when I already told you this wasn't what I was looking for.'" Have you ever taken a cross tone with your

AI? It says, "Oh, I a thousand apologies, sir. I don't know what I was thinking. I must have lost my mind for a moment." It says, "My intention was to follow your specific text, but now I realize that you're seeking a cleaner, more direct, likely more abstract hero." Here's what changed in our conversation, though. What changed in our conversation is Gemini started prompting Gemini says, "Based on

everything we've been saying, I can generate a new image that removes the detailed people and the the clutter that you didn't like. Um it's going to focus on sight, sound, and touch. It's going to be behind a clean, modern background." And so, my prompt was yes, please. Proceed. If I can prompt Gemini, Gemini should be able to prompt me as well, shouldn't it? And we're getting closer

to my vision. And so, then we do further refinement. We say, "All right, I like this much better. Thank you. But, let's refine it a little bit more. Let's not use any text other than our bold text. Let's get rid of all the extraneous background information." And it says, "Oh, like this?" Yes, like that. Thank you. And so, after a number of steps together, in fact, what

we ended up saying then is, "Yes. That image that we worked on all this time, would you go ahead and add it to the hero section?" That was all I had to prompt it. I didn't have to say, "Oh, go to this specific HTML ID. Oh, go to line 397 in the code. Oh, go to this subroutine" or anything like that. I'm like, "Yeah, we've been talking

about the hero image. Go ahead and add this to the hero section." And if you visit simple.io, excuse me, simpleaccess.io right now, you begin seeing all of this. And this is the culmination of the collaboration that we were doing together. What you're seeing right there is me saying I've got this idea that the web is sight, sound, and touch. And then Gemini created a bunch of text

to support that idea. I said I've got this idea for this kind of hero image, and Gemini said, "How about this? And how about this? And oh, now I get what you're saying. How about And this is what we see. So, there's this popular notion around generative AI. I call it the "Hey dude, write my term paper for me" scenario. That's the hype that we're getting around

AI, isn't it? You can type one tiny command, and it will generate an entire website or an entire code base. And in fact, the work I was doing with generative AI felt just like the work I'd already been doing for the last 30 years. Working with a trusted colleague, taking tiny bites, tiny steps one time at a at one step at a time, iteratively building something to

the point where we both agree that this is what we had if that was the end of the conversation, you'd be saying, "Okay, great, Scott. You can create a website using Gemini." But Simple Access IO says the web is sight, sound, and touch. And in fact, Gemini agrees. Gemini says, "You don't just have to use text prompts to interact with me." Gemini says, "You can upload images."

Images of punch cards. What's funny about this punch card? Images of moths taped in logbooks. Images of ideas of what I'm hoping to work towards. You can, of course, upload code as well. And oh, yeah, notebook LM. Put a pin in that. We're going to re-examine that in a moment. This is called foreshadowing, by the way. But what we find is that Gemini is a multi-sensory experience

as well in terms of input. These are all the things you can do. And then if I click on tools right next to that, we're going to see that Gemini is multi-sensory in its output as well. You've already seen me creating images in Gemini. If there was a create QR code option, I would have selected it. But Canvas is quite a curious addition to our pair programming

regime. If I turn on Canvas, then in fact, this is a closer representation to what my screen looks like as we're pair programming. I have my prompts off to the left, and I didn't share my entire prompt with you. It was four paragraphs of me saying, "All right, here's my URL, and here's my website, and here's my theme that I'm looking for, and I want to talk

about HTML elements and CSS, how accessibility is just baked in. You don't need to download everything else. Accessibility is just baked into HTML. I wanted to be able to say, "Screen readers are free on all the platforms we have. If you have an iPhone, you have VoiceOver for your screen reader. If you have an Android, you have TalkBack. If you have a Windows laptop, you have Narrator."

So, I wanted to give information about the free screen readers that you had available to you. In Microsoft Office, you have free accessibility checks. The accessibility assistant. Any Word document you're creating, down in the lower left-hand corner, it'll say "accessibility, good to go." Or "accessibility investigate." And that's true in Word, in Excel, in PowerPoint. I wanted to have a whole section of the website to talk about

the accessibility of our Office documents, PDFs. Woo. PDFs are hard mode when it comes to digital accessibility, but I wanted to be able to talk about this well. And so that was my prompt. And it loaded its first draft in the canvas. This is an interactive canvas over here. I can click on all the icons. I can scroll up and down. This was its first draft, and

this is what it looks like in preview mode. But if you look in the upper right-hand corner, you'll see that you can also begin seeing the source I said, "Oh, go on." Now, I have a way, perhaps a higher fidelity way, to pair program with Gemini. I can go into the code and start making changes, and Gemini can see what I'm seeing. Or I can prompt Gemini

to make changes to the code, and I can watch the changes happen in the canvas in real time. This was an incredible developer experience for me. I mean, just look at how pretty that source code is. I didn't do any of this. The simple website I specifically tried to generate through prompts almost exclusively. And so as much as I would love to take credit for this well-commented

code, this well-formatted code, this well-indented code, that in fact is all as we're collaborating together. Did you know that you can create variables in CSS? That's what Gemini started with. It's saying, here's the variable for your primary color. Here's your variable for your text color. Here's your default font set in your line spacing. And these are all variables that I can change once and the changes will

cascade throughout the rest of my website. I could wish I could take credit for writing this code cuz it's pretty. What I can do though is take credit for the prompt I used. I was looking at the source code as Gemini was writing and I said, okay, I what I'd like you to do is please use SMACSS to organize our CSS. And I really want to let

you know that view source is going to be very important. So, I want to make sure that the source code is as beautiful to look at as the rendered product. Now, notice I said, please use SMACSS. And Gemini said, oh, you mean scalable and modular architecture for CSS. I'm on it. This is the next question. This is the next step in our journey. How did Gemini know

what SMACSS is? Remember that foreshadowing we talked about earlier? How do you pound include truths with your AI agent? How does Gemini know what SMACSS is? And the answer is Let's explore that idea now. When I start a new chat with Gemini, there it is, a blank slate, tabula rasa, just waiting for us to figure out what adventure we're going to go on next. But if you

look to the left of this screen, you'll see I've got a list of notebooks there. That's notebook LM, and in fact, I've got a notebook for simple access. So, when I start this session with Gemini, if I reach out and click on simple access, all of a sudden, Gemini says, "I know kung fu." "I know all the sources you've stored in notebook LM. I've got six sources

I'll start drawing from. I have now a memory of all the chats you and I have had, and they're listed in chronological order. So, as I'm getting ready to start, I could click on a specific conversation and continue that thread, or if I don't click on any conversations, Gemini remembers everything that we've talked about. That's what I'd expect from a partner who I've pair programmed with. That's

what I expect from my generative AI, as well. This is what notebook LM looks like. As you see over on the left, this is where you begin defining all your sources. And so, if you look over there, you'll see that I bring in the latest 2026 CSS specifications. A little bit farther down, I bring in the living spec, the HTML5 spec. Oh, and yeah, the home for

scalable Adding sources is incredibly easy. You can upload files. I end up including websites quite a bit. I included links, URLs to the living HTML spec, and to SMACSS. I won't lie to you, though. Sometimes I like just copying text and pasting it in as a source of truth, as well. "Who said this?" And I said, "I did." It says, "Okay, all right, we'll include that as

a source as well. Thank you. the chat, though, as our source of truth, is what I think is the real killer feature here. Keeping that chat history as our source of truth means that Gemini forgets less what we've already covered. It remembers more of the important things that I said. And it gives us a higher fidelity pair programming experience. But how do I prove that this website

that Gemini generated is truly accessible, is truly standards-based, is all the things we were hoping it would be? Turns out I built another website to describe that one. allyview.com, a111yview.com. I could have spelled out accessibility view, but that's a lot of typing, isn't it? So, in fact, accessibility is a quite long word that begins with A and has a letter 11 letters in between and ends with

Y. If you'd like to visit allyview, you're more than welcome to. I can be trained as well. I didn't ask Gemini to generate this QR code. I just published the page and right-clicked on it and choose generate QR code. And in this webpage, I'm proving my work. I'm proving our work. I'm not proving Gemini's work any more than Gemini's proving our work. I am proving our work

together through codes, through engineering rigor, through tests. And so, there are HTML reports we can pull down at liveview.com. There's JSON that we can pull down there, comma-separated values that we can pull But, in fact, I was shocked the first time I started running Lighthouse audits on the code that Gemini was generating. Pleasantly surprised. Because as I was running these tests, I was almost exclusively getting a

hundreds across the board. And that's when you rub your eyes and run the test again saying, "Surely, something has gone wrong here." And they said, "No, it's a hundreds across the board, and stop calling me Shirley." 99 out of 100 on accessibility. And it's because this is the work that we did together. It didn't happen automatically. This wasn't a one-prompt-and-done kind of thing. This was the iterative

process that we went through together. As it generated a web page, I'd right-click and run a Lighthouse report, and I'd say, "Huh, the color contrast is a little bit off on this one." And it's like, "You're absolutely right. I'll make sure that that meets WCAG 2.1 AA standards." I'll say, "Ah, you're using buttons for navigation and links for in-page changes. Let's reverse the semantics of those. Links

are for visiting new pages, buttons are for actions within the page." And Gemini said, "Oh, that's an excellent point. I'll remember that from now on." What we just experienced was RAG in action. So many acronyms nowadays, right? RAG isn't the newest, but it is one of the most interesting. I asked Gemini, "How would you explain RAG in an easy-to-understand way?" And it says, "Oh, you mean retrieval

augmented generation?" I said, "Oh, yeah, yeah, that's what I'm talking about. Yeah." They said, "This analogy I really like is this open book test versus a closed book test analogy. It's one of my favorite analogies as well." So, I said, "Yeah, continue. Let's talk about this." This, of course, is Gemini explaining Gemini to us. Gemini says, "Imagine a brilliant student that's read a million books, read all

the books in the library, and they have to take a test, but they have to answer all the questions on this test from memory." They said, "Yeah, it could work, but it does get stuck in the past. If I've added new code to this website, I can't expect the training data the LLM caught that. But, more importantly, oh, it bluffs, it hallucinates. Think about that. If you're

taking a test and you're being quizzed across a million different books, you might bluff as well, wouldn't you? So, this is what standard out of the box Gemini gives you or ChatGPT or Copilot or Claude. RAG turns that closed book test into an open book test. RAG says, "All right, you're going to ask a question, and immediately I'm going to go and retrieve the information that you've

given me. I'm going to retrieve the documents that you stored in Notebook LM. I'm going to retrieve your source of And I'm going to use that truth to augment what I generate based on facts, based on truth, not vague memories of something I read once. I like demystifying magic. And I'm not saying that isn't truly magical interacting with a AI agent like I've been doing. That's the

magical part in fact. It can't do it alone. It's the shared partnership between the two of us. It's asking me to provide a source of truth so it can begin generating the artifacts I ask it for based on our shared understanding. The retrieval engine, the R of rag, is And the standard LLM is the maybe. And so if you have a maybe and a truth and you

want to use them together rag is the architecture that forces the to play by the rules of the truth. I think that's lovely. I love asking Gemini about Gemini. I say, "So what do you think? Is this a reasonable example? Should we continue talking about this?" It says, "Yeah, absolutely." It really should. For an audience of software architects and developers >> [snorts] >> it serves as a

perfect case study for how to build systems in the age of maybe. Huh. So we continue to play with this idea of We continue to play with this idea of the maybe. And it's so easy to get bogged down in the infrastructure of vectorizing your data and agentic this and that and the other. this is an incredibly important exercise for every developer, every architect what parts of

your architecture require the truth? And what parts of your architecture depend on the maybe? So, this idea of wrapping in the truth is in fact where we are right now. And there are a number of different tools. I've shown you one involving notebook LM and Gemini, but we know that these are architectural patterns that can be implemented in whatever particular engine you're using, whatever particular framework you

prefer. where does this leave us? Where's this leave us? Of course I asked Gemini. Why wouldn't I at this point? Gemini, get long. Gemini says, "You know what? To wrap up a keynote that's been this conceptual, what can you do to bring this back to Earth? What can you do to make sure that you walk away with tangible outcomes, not just big ideas?" I even want to

leave you with a call to action. Three big thoughts for us to leave with. You are going to hear people telling you that AI will solve all of our problems, that AI is the only thing you should be focusing on as a software developer, the rest of software engineering is obsolete and dead, and you'd be better off leaving it behind. You also have people telling you, "Ah,

that AI is just a flash in the pan, right? It's a fad. That AI is never going to catch on. I've been working in this industry for 30 years. I didn't need no AI to do my job. I don't know why you do. Those are the polar opposites. Those are the extremes of opinion that we face every day. As software engineers, you're given decisions to make. You're

asked to choose what programming language you're going to use. You're asked to choose a platform. You're asked to choose a framework. And so, couching these ideas, once again, in and maybe. Not truth or Not choosing one and abandoning the other, but in fact, finding a way to reconcile both of these. How do we incorporate both the truth and maybe in our architecture? In fact, how do we

wrap in the truth? And this idea of the multi-sensory web, AI isn't just a parlor trick for generating code. If you are only using text to interface with your AI engine of choice, you're only using one of all the senses Gemini represents the true promise of multi-sensory architecture. Some people I know use a microphone and talk to their AI agents. Perhaps not in a crowded auditorium like

this, but if you're at home, can you imagine just having a microphone up in every conversation that you saw me typing with Gemini, I'd speak it instead. Rather than saying, "I've got this idea for bold colors and icons to describe site, sound, and touch." I was able to upload an image showing Gemini what I meant. Did you know that you can include YouTube videos in Notebook LM?

I'm not going to lie. I have included YouTube videos in my Notebook LM. I'm saying this YouTube video is my source of truth. It inspired me to have this conversation with you, Gemini. I hope you can watch this video and get inspired by it as well. But, finally, this is your call to action. This is your Monday morning directive. When you go back to the office on

Monday, I want you to be aware of people who are trying to force to be something it isn't. I want you to think about the people who are trying to convince you that if you just work hard enough, AI can in fact be made deterministic. It can't. It won't. It's the opposite of the truth. But, by the same token, that doesn't mean you should write everything in

COBOL on punch cards as well. In fact, what this means is we need to embrace all aspects of this architecture. We need to have all the tools available to us, not just one or not just some. So, you remember that punch card that we started with? That simple punch card that outwitted the computer. Because the simple messy human reality software has been written by simple messy humans

for as long as software has been written. And how has AI been trained to write your software now? On the entire history of simple human-generated code. So, this terrifying messy, beautiful time that we're living in right now is the age of maybe. Did you enjoy yourself? I did as well. Thank you so much for your time. >> [music]