Latest / The DEX Show: A Show for IT Change Makers / Vibe Coding, Dune & Staying Human: Sam Gantner Unfiltered

Transcript

Vibe Coding, Dune & Staying Human: Sam Gantner Unfiltered · The DEX Show: A Show for IT Change Makers

• 0:02
  Speaker 1: Hey, it change makers, and welcome to the Deck Show
• 0:05
  Speaker 1: with tom Ac Grath. Let's get into it.
• 0:12
  Speaker 2: Today's guest Everybody, one of our favorite returning guests, and
• 0:15
  Speaker 2: I have to say often contributes some of our best
• 0:19
  Speaker 2: performing episodes of the year. We usually get him on
• 0:22
  Speaker 2: once or twice a year to give us the lowdown
• 0:25
  Speaker 2: on the state of the industry, if you will, the
• 0:27
  Speaker 2: state of the Deck's industry. He is Sam Gantner, chief
• 0:33
  Speaker 2: product officer at Next Think. Sam.
• 0:34
  Speaker 3: Welcome back to the show. How you do Hey, Dom,
• 0:37
  Speaker 3: I doing fantastic. Thank you so much for having me.
• 0:40
  Speaker 3: It's always a pleasure to be here with you. Always
• 0:42
  Speaker 3: a pleasure to have you on.
• 0:43
  Speaker 2: And I was about to qualify it because we were
• 0:45
  Speaker 2: doing something a little bit different today, right because since
• 0:50
  Speaker 2: you last came on the show. I don't want to
• 0:52
  Speaker 2: trivialize it, but you've become something of a source of
• 0:55
  Speaker 2: LinkedIn AI hot takes. Mate. You've been You've been born
• 0:59
  Speaker 2: again as a poster and I'm loving it. Man. How
• 1:03
  Speaker 2: you finding? What inspired you? Why have you suddenly why
• 1:06
  Speaker 2: have you suddenly fled up with interesting and engaging opinions online?
• 1:11
  Speaker 3: Listen, Listen. I want to become an influencer. Now it's
• 1:16
  Speaker 3: been I've been experimenting with that a little bit, right,
• 1:18
  Speaker 3: And and the reason that jove this is the following.
• 1:22
  Speaker 3: There's two things where I spend a lot of time thinking.
• 1:25
  Speaker 3: One is how do we build great software products? And
• 1:29
  Speaker 3: what is behind some of the art of building great
• 1:33
  Speaker 3: software right? And two obviously related to my job. But
• 1:38
  Speaker 3: next think is how is it t evolving to become
• 1:41
  Speaker 3: much more strategic for the for the industry. And what
• 1:45
  Speaker 3: I noticed is that I've been sharing some of my
• 1:47
  Speaker 3: thoughts around those two topics in various forms. For instance,
• 1:51
  Speaker 3: the eight stuff I'm sharing get nexting events, master experience
• 1:55
  Speaker 3: and experience events, talks with customer CIOs, and then the
• 2:01
  Speaker 3: more work related how to craft software it's more around conferences.
• 2:06
  Speaker 3: Have been a couple of times at the PFL or
• 2:08
  Speaker 3: one on one with other people in the industry, and
• 2:11
  Speaker 3: I realized that sometimes some of this content can be
• 2:14
  Speaker 3: relevant for people, right, And so I thought, you know,
• 2:18
  Speaker 3: is there a way for me to spread this a
• 2:20
  Speaker 3: little bit broader and maybe make sure that perhaps someone
• 2:24
  Speaker 3: out there can take advantage of that. And so I've
• 2:26
  Speaker 3: been trying to experiment with LinkedIn, and I must say
• 2:29
  Speaker 3: it's kind of fun, right, it seems to be working,
• 2:32
  Speaker 3: so you know, it's not a lot of effort because
• 2:34
  Speaker 3: those are things that I know already and it's fun.
• 2:38
  Speaker 2: To do it. Your postings not getting in the way
• 2:40
  Speaker 2: of your day job.
• 2:41
  Speaker 3: You're sure, no, for sure.
• 2:43
  Speaker 2: Yeah, no, yeah, it's been known to happen.
• 2:45
  Speaker 3: It's been known to happen. But I'm trying to keep
• 2:48
  Speaker 3: a lot of control on that. A couple of posts,
• 2:50
  Speaker 3: I must admit I've been more of a runt. I
• 2:53
  Speaker 3: did one on Garment that went super viral, and you
• 2:56
  Speaker 3: know that's sometimes companies.
• 2:58
  Speaker 2: When one of the I'm sure I know the post,
• 3:01
  Speaker 2: but say again, you didn't want to.
• 3:02
  Speaker 3: A Garment one, you know, the Garment smart watch and
• 3:06
  Speaker 3: missed that one. Oh you missed that one. No, it's
• 3:09
  Speaker 3: the one that ran about what was right about well,
• 3:11
  Speaker 3: their choice of colors for one of the charts that
• 3:14
  Speaker 3: is completely readable, not saying person on the planet is
• 3:17
  Speaker 3: able to understand the chart and it's just a crazy
• 3:21
  Speaker 3: product decision, right. So at the same time, it was
• 3:23
  Speaker 3: something that was bothering me as a user but also
• 3:25
  Speaker 3: as a professional building software is like, why would you
• 3:29
  Speaker 3: make such a bad choice?
• 3:30
  Speaker 2: Right?
• 3:30
  Speaker 3: And they had to share it with the planet.
• 3:33
  Speaker 2: Nice. Well, I'm gonna I'm gonna admit something, right, and
• 3:36
  Speaker 2: I like, I think from what I be able to
• 3:38
  Speaker 2: read from your post is and it stands to reason right,
• 3:41
  Speaker 2: But you're you don't show any undo loyalty to a
• 3:46
  Speaker 2: certain platform.
• 3:47
  Speaker 3: Right, like why should you?
• 3:48
  Speaker 2: Why should you? But I'm like the opposite, So for it.
• 3:50
  Speaker 2: Just to give illustrate this, I have had Spotify since
• 3:53
  Speaker 2: I think the year it came out of the first cup.
• 3:55
  Speaker 2: I was like, I love this product so much and
• 3:58
  Speaker 2: I've never broken it. And I definitely even when like
• 4:00
  Speaker 2: Apple was over ranged with Apple Music, it was a
• 4:03
  Speaker 2: bit cheaper. There was records on it, but you couldn't
• 4:06
  Speaker 2: get on Spotify. I was like, no, if these guys
• 4:08
  Speaker 2: delivered me this product, I feel attached to it. But
• 4:12
  Speaker 2: I think you make you make a very good argument
• 4:14
  Speaker 2: that this would be very foolhardy with when it comes
• 4:18
  Speaker 2: to l elms and when it comes to AI. And
• 4:19
  Speaker 2: we're going to get into a little detail about why
• 4:21
  Speaker 2: that is. But first of all, give us the overview
• 4:24
  Speaker 2: of what your AI stack looks like personally and all professionally.
• 4:29
  Speaker 3: I guess you know, yeah, absolutely, look I've been experimenting
• 4:32
  Speaker 3: with various solutions, and towards the end of the year,
• 4:35
  Speaker 3: I decided that they would try one AI each month,
• 4:41
  Speaker 3: you know, and go through all of them, just to
• 4:43
  Speaker 3: see a little bit the variety. And so I used
• 4:46
  Speaker 3: to be very focused on OpenAI. I spent some time
• 4:50
  Speaker 3: with Gemini, which was actually great, but they really didn't
• 4:54
  Speaker 3: like their privacy policy, so I moved away and next
• 4:58
  Speaker 3: in line for me was closed. And the problem is
• 5:01
  Speaker 3: that it just became too good for me to stop
• 5:04
  Speaker 3: and go to the next one. So I kind of
• 5:07
  Speaker 3: stopped that cloude and gotten deeper and deeper in clode.
• 5:11
  Speaker 3: But there's something that is happening now with the lems
• 5:14
  Speaker 3: that is changing the game quite a bit, which they're
• 5:18
  Speaker 3: adding memory, and so memory is making them way stickier
• 5:22
  Speaker 3: than in the past because now they really start to
• 5:25
  Speaker 3: understand you, they remember your past conversations, and so the price,
• 5:29
  Speaker 3: the cost of switching to a different model or to
• 5:32
  Speaker 3: a different provider at least, is becoming much much more expensive. Right,
• 5:36
  Speaker 3: So I think we're going to see people starting to
• 5:39
  Speaker 3: stick a little bit more into some of the things
• 5:42
  Speaker 3: that they're doing. And then on the coding front, I
• 5:45
  Speaker 3: went one hundred percent on cloud code. You know oppus.
• 5:48
  Speaker 3: It's just ridiculous and it's been so exciting for me
• 5:52
  Speaker 3: because you know, I stopped coding years ago because simply,
• 5:55
  Speaker 3: you know, the level of contribution that a single person
• 5:58
  Speaker 3: was able to bring was too small, couldn't make progress
• 6:01
  Speaker 3: fast enough. I didn't have enough hours in the day
• 6:03
  Speaker 3: to make something meaningful. And now with clod code, you know,
• 6:06
  Speaker 3: in a few hours you can just build something end
• 6:09
  Speaker 3: to end, and it's just incredible, this sentiment of being
• 6:11
  Speaker 3: able to once again produce work in software. It's personally,
• 6:16
  Speaker 3: it's just amazing.
• 6:17
  Speaker 2: I think that's actually in a ferness. That's one of
• 6:20
  Speaker 2: the nice things about your candy, you know, just offering
• 6:22
  Speaker 2: hot takes and runs. You are you are showing like
• 6:24
  Speaker 2: personal innovation with with coding and such like, which I
• 6:29
  Speaker 2: think which is really nice to see. It's really interesting
• 6:31
  Speaker 2: to see on your feed. What's something you've built recently
• 6:35
  Speaker 2: that sort of really changed the way you work.
• 6:39
  Speaker 3: Look, I've been building a few tools to simplify things, right,
• 6:42
  Speaker 3: I give you a couple of example. I build a
• 6:44
  Speaker 3: tool to help me understand and better allocate are in
• 6:49
  Speaker 3: the investment. That's one of the key things that you
• 6:51
  Speaker 3: need to understand, right, so, how much effort, how much
• 6:53
  Speaker 3: of your available capacity are you putting on different products
• 6:56
  Speaker 3: and projects? And it's not so easy. So I build
• 7:00
  Speaker 3: a tool that really helps me visualize where that spend
• 7:03
  Speaker 3: goes and how to better allocate to really align with
• 7:06
  Speaker 3: the strategic.
• 7:06
  Speaker 2: Things that we want to do.
• 7:08
  Speaker 3: Or more recently, you know, like every company at the
• 7:10
  Speaker 3: beginning of the year with the a compensation review to
• 7:13
  Speaker 3: provide salary increases and things like that, and never have
• 7:16
  Speaker 3: a tool to help me structure that in a very
• 7:19
  Speaker 3: fair and correct and performance oriented way, which was absolutely great.
• 7:24
  Speaker 3: But what I realized is that at the end of
• 7:27
  Speaker 3: the day is not really about the tool. I think
• 7:29
  Speaker 3: it's about how you're thinking changes. To understand the real
• 7:34
  Speaker 3: potential of GENEI and how to use it, you need
• 7:39
  Speaker 3: to be very familiar in how it operates. And the
• 7:42
  Speaker 3: best way to become familiar and how it operates is
• 7:45
  Speaker 3: to use it very very often. This is why, you know,
• 7:48
  Speaker 3: I've been advocating that the R and D leaders that
• 7:51
  Speaker 3: are truly able to drive three x plus you know,
• 7:56
  Speaker 3: outcomes for their engineering teams are those leaders that went
• 8:00
  Speaker 3: back to coding right using LMS This is why the
• 8:05
  Speaker 3: pms that are building the best argentic products that I've
• 8:08
  Speaker 3: ever seen are those that are using LMS to prototype,
• 8:13
  Speaker 3: because fundamentally they understand how GENI works, what are the limits,
• 8:18
  Speaker 3: what are the opportunities, and then they're able to apply
• 8:21
  Speaker 3: that to their job. So I would say this is
• 8:23
  Speaker 3: really the major difference, not really the tool that I build,
• 8:26
  Speaker 3: but how building those tools have helped me understand at
• 8:30
  Speaker 3: a deep level how agents operate.
• 8:33
  Speaker 2: Per Simply, if you're in a leadership position in product development,
• 8:37
  Speaker 2: you're able to run your own prototypes. Is something you're
• 8:40
  Speaker 2: able to experiment of a top level. It's something you
• 8:43
  Speaker 2: couldn't have done with anything like this is two years
• 8:45
  Speaker 2: ago exactly one hundred percent, and just use it. You know,
• 8:48
  Speaker 2: any leader out there, I think you know.
• 8:51
  Speaker 3: General purpose people just doing HR management are losing their
• 8:57
  Speaker 3: relevance more and more. If you want to be a
• 9:00
  Speaker 3: great leader no matter where, but especially in our ind
• 9:03
  Speaker 3: you need to be hence on.
• 9:04
  Speaker 2: Again, that's an interesting point. Okay, So more generally, do
• 9:09
  Speaker 2: we see on the screen before us the two types
• 9:12
  Speaker 2: of AI users via the ones because I take because
• 9:16
  Speaker 2: it's not obviously my my tendency towards extreme product loyalty
• 9:21
  Speaker 2: is only exacerbated when you're in conversation with something in
• 9:24
  Speaker 2: day out and you and you can't help anthropomorphis. Then
• 9:27
  Speaker 2: you can start to feel guilty about the idea of
• 9:29
  Speaker 2: turning away God, but you're getta. You're gonna end up
• 9:31
  Speaker 2: with two types of two types of users, people who
• 9:34
  Speaker 2: who stick with one and people who approach the whole,
• 9:38
  Speaker 2: the whole sort of raft of options or something like
• 9:41
  Speaker 2: a toolkit. And what does that tell us do you
• 9:45
  Speaker 2: think those those respective approaches, or rather the latter approach,
• 9:49
  Speaker 2: what does that tell us about what a I actually
• 9:52
  Speaker 2: is in terms of a product or something that's much
• 9:54
  Speaker 2: closer to infrastructure itself.
• 9:58
  Speaker 3: Well, it's it's kind of fe right when you look
• 10:01
  Speaker 3: at how people are adopting I, I kind of break
• 10:04
  Speaker 3: them down in three categories. The first cuttery is people
• 10:08
  Speaker 3: that don't get it yet, right, the people are refusing
• 10:11
  Speaker 3: to get into I. And it's surprising, you know, even
• 10:14
  Speaker 3: across the people that I know, there's quite a few
• 10:18
  Speaker 3: run right. Then you get the first type of users,
• 10:22
  Speaker 3: which are those that are really jumped into AI to
• 10:27
  Speaker 3: use them as kind of assistance, sidekicks and so on.
• 10:30
  Speaker 3: And so forth. And then you have those that are
• 10:32
  Speaker 3: really starting to get into the mood of AI, which
• 10:36
  Speaker 3: are starting to delegate some of the work to agents.
• 10:40
  Speaker 3: And back to your question when you know, is I
• 10:44
  Speaker 3: a product or infrastructure? Is actually both? And here's the analogy.
• 10:50
  Speaker 3: So in the SaaS word, you add the Asia AWS
• 10:54
  Speaker 3: Google offering infrastructure, right, and then a number of companies
• 10:59
  Speaker 3: building software products on top of that infrastructure. That's how
• 11:02
  Speaker 3: we're used to think about that in the AI world.
• 11:07
  Speaker 3: The infrastructure is the fundational models, So Open, EI, Entropic,
• 11:12
  Speaker 3: Google are building those models, and then on top of
• 11:16
  Speaker 3: this infrastructure you build more specialized agents that are your products.
• 11:22
  Speaker 3: Right now, it just happens that the infrastructure players, so
• 11:26
  Speaker 3: Open Ai, Entropic and so on and so forth, are
• 11:28
  Speaker 3: also building agents on top. So are also building products
• 11:31
  Speaker 3: on top of their own fundational models, right. But so
• 11:36
  Speaker 3: in this perspective, it's really both. How is that influencing
• 11:41
  Speaker 3: your approach to future product design?
• 11:43
  Speaker 2: For next thing? The reality of emerging industry?
• 11:48
  Speaker 3: Yes, we're really following the maturity curve of how AI
• 11:51
  Speaker 3: is going from assisting people to doing work on behalf
• 11:56
  Speaker 3: of people. Right. Give you a couple of example look
• 11:59
  Speaker 3: at the assist even the naming of assists itself. When
• 12:02
  Speaker 3: we started working on Assist back in twenty twenty two,
• 12:06
  Speaker 3: the idea was to provide the kind of a sidekick
• 12:09
  Speaker 3: that would help I T teams navigate next thing and
• 12:13
  Speaker 3: getting value out of next thing much quicker, with much
• 12:16
  Speaker 3: less effort. Now, over time, that value proposition has completely changed,
• 12:22
  Speaker 3: and it's been evolving into how do we build autonomous
• 12:28
  Speaker 3: agents to which it people are able to delegate part
• 12:32
  Speaker 3: of the work. That's a completely different value proposition, right.
• 12:36
  Speaker 3: It's not about how do we let them use next
• 12:39
  Speaker 3: thing better? Is how do we dramatically shift the amount
• 12:44
  Speaker 3: of manual tasks that they need to do not only
• 12:46
  Speaker 3: next thing, but globally for their work. And Spark is
• 12:50
  Speaker 3: a similar philosophy. Right now, it's Spark. We just launched
• 12:54
  Speaker 3: it like two months ago. We're able to replace a
• 12:58
  Speaker 3: lot of that manual service. That's work that used to
• 13:02
  Speaker 3: be so slow, by the way, for people experiencing a
• 13:05
  Speaker 3: problem that they want a solution much more quickly, right,
• 13:07
  Speaker 3: and so Spark is able to produce that at levels
• 13:11
  Speaker 3: of quality that are higher than what we were able
• 13:14
  Speaker 3: to do with manual human workflows before so more broadly,
• 13:19
  Speaker 3: the concept for us, for every new product, every new
• 13:22
  Speaker 3: thing that we want to bring into next thing is
• 13:26
  Speaker 3: is becoming let's do it AI first, and then we
• 13:30
  Speaker 3: add a user interface only when it's really needed.
• 13:33
  Speaker 2: Right.
• 13:34
  Speaker 3: But now the whole thinking has become AI agent first, right.
• 13:40
  Speaker 2: I always I missed the opportunity to say, to say,
• 13:42
  Speaker 2: friend of the show when you mentioned the cyst, because
• 13:44
  Speaker 2: of course it was a prior guest prior I guess.
• 13:47
  Speaker 2: But you talk a lot in your posts about a
• 13:50
  Speaker 2: viv orchestration of different AI agents, right, and speaking of which,
• 13:55
  Speaker 2: what you would just saying You've also got a lot
• 13:57
  Speaker 2: of experience managing human teams interested with the analogous to
• 14:02
  Speaker 2: one another, and where the analogy breaks down in your experience.
• 14:08
  Speaker 3: That's a good question, and I love it. Well, managing
• 14:11
  Speaker 3: agents is much simpler and yet much more complex at
• 14:16
  Speaker 3: the same time. Let's me laborate why why is it simpler? Well,
• 14:20
  Speaker 3: agents have no personality, no ego, no emotions, no politics.
• 14:26
  Speaker 3: They don't get tired, they don't need time out right, So,
• 14:32
  Speaker 3: by the way, all of the things that make managing
• 14:35
  Speaker 3: humans pleasant and great and fun and the interesting time. Yeah,
• 14:43
  Speaker 3: I mean you don't get any of that with agents, right,
• 14:46
  Speaker 3: So it's honestly it's much more boring to manage to
• 14:50
  Speaker 3: manage agents, but simpler from that perspective, you don't have
• 14:54
  Speaker 3: the you don't have the human emotion, right, and at
• 14:58
  Speaker 3: the end of the day, that's what we like anyway.
• 15:01
  Speaker 3: Why is it more complex to manage agents? But it's
• 15:04
  Speaker 3: more complex because you know, agents are very very smart,
• 15:09
  Speaker 3: but unbelievably limited when it comes to judgment. Simply put,
• 15:15
  Speaker 3: you know, a human, a reasonable human, knows how to
• 15:19
  Speaker 3: stop and to ask a question right.
• 15:24
  Speaker 2: Now.
• 15:24
  Speaker 3: Agents they don't do that. If you look at the
• 15:26
  Speaker 3: behavior when they're wrong, they tend to go down that
• 15:31
  Speaker 3: wrong path deeper and deeper and deeper and deeper. And
• 15:35
  Speaker 3: they're not gonna stop, right, They're going to try to
• 15:38
  Speaker 3: figure out a way down the wrong path without asking
• 15:40
  Speaker 3: too many questions. And so this is why a lot
• 15:44
  Speaker 3: of the efforts when you want to manage agents, it's
• 15:47
  Speaker 3: not really in defining the task, which you still need
• 15:51
  Speaker 3: to do fairly precisely, but it's more around how do
• 15:55
  Speaker 3: you set up the guardrails so that they don't go
• 15:58
  Speaker 3: down this path, they don't start to locinate all.
• 16:01
  Speaker 2: Over the place.
• 16:03
  Speaker 3: Also, the other thing is that agents don't really have
• 16:05
  Speaker 3: a lot of common sense, right, and so you need
• 16:10
  Speaker 3: to spell out everything that they need to do. And
• 16:13
  Speaker 3: it's getting better with memory, right because they're starting to
• 16:16
  Speaker 3: learn from some of these things from the past experiences.
• 16:19
  Speaker 3: But today you still need to put a lot of
• 16:21
  Speaker 3: effort on that. So I think that managing agencies less
• 16:25
  Speaker 3: like managing people and much more like designing and architecting systems.
• 16:33
  Speaker 3: And this is why when you look at that from
• 16:35
  Speaker 3: that perspective, people that are really good at managing agents
• 16:40
  Speaker 3: are people that are used to be good or are
• 16:43
  Speaker 3: good at software architecture. And this is why experience engineers
• 16:49
  Speaker 3: are really great out that at managing agents, because they've
• 16:53
  Speaker 3: been trained and they've been practicing that art of orchestrating
• 16:58
  Speaker 3: systems for a long time.
• 17:01
  Speaker 2: Speaking of us, let's come back to your question of
• 17:03
  Speaker 2: code a little bit. If you develop as a still
• 17:07
  Speaker 2: writing code, you've already lost more generally, what does a
• 17:11
  Speaker 2: bad engineer look like? Do you think in twenty twenty.
• 17:14
  Speaker 3: Six, Well, as we classify that, you know, things are
• 17:18
  Speaker 3: changing very fast, right, So let me start by introducing
• 17:22
  Speaker 3: what it means to adopt the I for an engineer.
• 17:25
  Speaker 3: And when you look at that, there are three main
• 17:27
  Speaker 3: waves of AIR adoption. The first one, and it's kind
• 17:31
  Speaker 3: of over now, is once again where AI was there
• 17:35
  Speaker 3: to assist engineers. And this is really the beginning of
• 17:38
  Speaker 3: a cursor for instance, where they introduced this concept of
• 17:41
  Speaker 3: tab autocomplete and making that experience a meaning where cursor
• 17:46
  Speaker 3: is able to help you write some of the code.
• 17:50
  Speaker 3: So that brought some i don't know, ten percent acceleration,
• 17:54
  Speaker 3: twenty percent acceleration, but it was useful but somewhat marginal.
• 17:58
  Speaker 2: Right.
• 17:58
  Speaker 3: Then the second wave is where AI starts to write
• 18:02
  Speaker 3: all of the code for engineers. And so here you
• 18:06
  Speaker 3: have the paradigm of clode code that was introduced where
• 18:09
  Speaker 3: you have multiple AED sessions running in parallel and then
• 18:12
  Speaker 3: you have a human that is courting it. In these
• 18:14
  Speaker 3: sessions have one tap procession, and you know, they work
• 18:19
  Speaker 3: around the fact that the agent is relatively slow by
• 18:22
  Speaker 3: having multiple of this of this one's writing code at
• 18:26
  Speaker 3: the same time. And here, by the way, it's getting
• 18:29
  Speaker 3: pretty complex for engineers because to be efficient as that,
• 18:33
  Speaker 3: since the engineer is doing the manual orchestration, it's like
• 18:36
  Speaker 3: playing multiple games of chess at the same time, right,
• 18:40
  Speaker 3: you need to multitask and that's very, very very hard
• 18:42
  Speaker 3: to do. And and there you could kind of see
• 18:44
  Speaker 3: the skill level of somebody. You know, uh, high skill level,
• 18:49
  Speaker 3: maybe they can manage five of the time. Lower skill level,
• 18:51
  Speaker 3: they can manage two three.
• 18:52
  Speaker 2: Like those great chess players. You know, the best chess
• 18:56
  Speaker 2: players always line line them up that they come down
• 18:58
  Speaker 2: with chess club. They get five, six of you in
• 19:00
  Speaker 2: a row and exactly precisely.
• 19:04
  Speaker 3: And then now you have the third wave that is
• 19:07
  Speaker 3: coming in which you have a fully autonomous agents that
• 19:11
  Speaker 3: can go on for four hours and and almost build
• 19:15
  Speaker 3: a complete product from scratch, right, And to be honest,
• 19:18
  Speaker 3: that's still mostly used for internal tooling and prototyping rather
• 19:22
  Speaker 3: than production code. But the direction is clear. We're gonna
• 19:26
  Speaker 3: head down that way down the line. So what's a
• 19:29
  Speaker 3: bad engineer. I don't know if I if I can
• 19:32
  Speaker 3: qualify that, But to me, a bad engineer, now look
• 19:38
  Speaker 3: for me, a bad engineer is somebody was not yet
• 19:43
  Speaker 3: realized that alone they can do the work that used
• 19:48
  Speaker 3: to be done by a team of five, right, or
• 19:51
  Speaker 3: either maybe they realized, but they don't want to take
• 19:53
  Speaker 3: accountability for that because honestly, it's very hard. And what
• 19:58
  Speaker 3: is paradoxical in all of that is that basically AI
• 20:04
  Speaker 3: is bringing humanity back into the equation. We had a
• 20:07
  Speaker 3: long time in which in software engineering we stopped thinking
• 20:12
  Speaker 3: about people and we were only focusing on teams.
• 20:16
  Speaker 2: Right.
• 20:16
  Speaker 3: The individual kind of lost its importance and everything was
• 20:20
  Speaker 3: about what is the team able to accomplish, which was
• 20:24
  Speaker 3: absolutely great, right, it was the right thing to do,
• 20:26
  Speaker 3: But it was also making it possible somewhat to hide
• 20:30
  Speaker 3: within that context. And now it's changing that because individuals
• 20:35
  Speaker 3: are coming back at the center of attention, but also
• 20:38
  Speaker 3: they face much more pressure because now if they can deliver,
• 20:42
  Speaker 3: is much more visible for them. And I think this
• 20:44
  Speaker 3: is one of the reasons why some people are uncomfortable
• 20:48
  Speaker 3: with this new reality, because it's obviously pretty tough.
• 20:52
  Speaker 2: If you think about that, what do you think about
• 20:55
  Speaker 2: like learning code nowadays? Is it like learning handwriting? Is
• 20:59
  Speaker 2: it is it which which, to be fair, like my
• 21:01
  Speaker 2: kids still do right, like our children still do school,
• 21:04
  Speaker 2: they do they write with pens and pencils until they
• 21:08
  Speaker 2: twelve years old, and they never ever look at them
• 21:10
  Speaker 2: again other than signing a document. And I can't make
• 21:12
  Speaker 2: up my mind whether that's a kind of a quirk
• 21:16
  Speaker 2: and should be pushed past or whether it's a kind
• 21:19
  Speaker 2: of a decent foundation in and of itself. And where
• 21:21
  Speaker 2: would you stand analogously to learning code for engineers? You
• 21:24
  Speaker 2: know how necessary?
• 21:27
  Speaker 3: Wow, you make me feel obsolete. You know, I'm still
• 21:30
  Speaker 3: taking some of my notes with a with a pen
• 21:32
  Speaker 3: and a piece of paper.
• 21:35
  Speaker 2: So we don't me too on occasion.
• 21:37
  Speaker 3: Right, yeah, but they get your point. Listen as you say,
• 21:41
  Speaker 3: my kids are still learning how to handwright, and I
• 21:45
  Speaker 3: believe that it matters for engineers. It matters enormously. Not
• 21:50
  Speaker 3: really because of the fact that they're going to have
• 21:53
  Speaker 3: to write code in the future they want, but as
• 21:56
  Speaker 3: a way of thinking, Yeah, right, I was answering before
• 21:59
  Speaker 3: engineers are the best at orchestrating agents because it's like
• 22:03
  Speaker 3: architecting systems, And so the question becomes, how do you
• 22:06
  Speaker 3: become how do you become great at architecting systems? What
• 22:10
  Speaker 3: do you need to learn in order to be able
• 22:12
  Speaker 3: to be proficient at that? And coding is the foundation
• 22:16
  Speaker 3: of that because it brings you that ability to think logically,
• 22:21
  Speaker 3: to break down a complex task into match smaller units, right,
• 22:26
  Speaker 3: And that is really what Orchestrating is all about how
• 22:29
  Speaker 3: do you take something that is very big that you
• 22:33
  Speaker 3: wouldn't be able to solve as a whole, and break
• 22:36
  Speaker 3: it down in different units that are able that you're
• 22:40
  Speaker 3: able to understand. And code is the same, right, you
• 22:43
  Speaker 3: take down a complex program and you start to define, Okay,
• 22:45
  Speaker 3: these are the modules that I need into that, and
• 22:47
  Speaker 3: then within modules, these are the classes that I need
• 22:50
  Speaker 3: for that, and then within those classes, those are the functions.
• 22:53
  Speaker 3: And this is really the process that you need for
• 22:55
  Speaker 3: orchestrating agents. So I believe that in the future, you know,
• 22:59
  Speaker 3: if myke it's one to get into this space, that's
• 23:03
  Speaker 3: certain point. I would definitely recommend for them to learn coding. Now.
• 23:08
  Speaker 3: They won't need to do it to the level of
• 23:10
  Speaker 3: proficiency that they're that we used to teach people ten
• 23:12
  Speaker 3: years ago, but at least the foundations I think they
• 23:15
  Speaker 3: can really help shape your way of thinking in a
• 23:18
  Speaker 3: very useful.
• 23:19
  Speaker 2: Way in any space. As well as things become more automated,
• 23:24
  Speaker 2: as it were, foundational knowledge will remain really important, almost
• 23:28
  Speaker 2: just like symbolically that our professionals retain some awareness of
• 23:33
  Speaker 2: even how the technology is functioning, you know, like absolutely
• 23:36
  Speaker 2: be it a pilot or whatnot. But okay, so this
• 23:41
  Speaker 2: is an interesting question to me based upon what you
• 23:45
  Speaker 2: were saying about infrastructure. You're building your own tools and
• 23:49
  Speaker 2: you're not waiting around for products to be delivered to
• 23:51
  Speaker 2: you yourself. Are we heading for a world where the
• 23:55
  Speaker 2: best users outgrows SaaS entirely? And what is the future
• 23:59
  Speaker 2: of SaaS in this emerging equosis?
• 24:03
  Speaker 3: Look for me, it's kind of unlikely, right. I think
• 24:07
  Speaker 3: a lot of engineers are building their own tools to
• 24:10
  Speaker 3: accelerate things because the truth is, by the way, a
• 24:14
  Speaker 3: lot of engineering time never went into writing code. It
• 24:18
  Speaker 3: went into everything that was associated to writing code and
• 24:21
  Speaker 3: to get in product code to production, right. And so
• 24:25
  Speaker 3: to go from a vibe coded thing that kind of
• 24:28
  Speaker 3: works to something that is secure, scalable, production grade, that
• 24:34
  Speaker 3: can be maintained, that can be used by others, there
• 24:38
  Speaker 3: is a huge enormous gap, right, And most people won't
• 24:45
  Speaker 3: spend the effort to close the day apps fact, right,
• 24:48
  Speaker 3: and especially in the large enterprises even more important, right,
• 24:51
  Speaker 3: because you know what if this person leaves no the
• 24:55
  Speaker 3: vibe coded a tool that a bunch of people are using,
• 24:57
  Speaker 3: and then they just leave, what happens?
• 25:00
  Speaker 2: Imagine exactly.
• 25:01
  Speaker 3: Yeah, it's a mess, right, or what if they didn't
• 25:04
  Speaker 3: don't secure it and then the data goes everywhere? What
• 25:07
  Speaker 3: about GDPRS and privacy and all of these things. I
• 25:11
  Speaker 3: think it doesn't scape. I believe that the best users
• 25:16
  Speaker 3: will become very proficient in delegating work to AI solutions
• 25:20
  Speaker 3: that are provided by the market, and so at the
• 25:24
  Speaker 3: same time, software companies have the opportunity to build much
• 25:29
  Speaker 3: better products right. Also, by the way, this comes with
• 25:33
  Speaker 3: a side effect, which is the temptation for software companies
• 25:35
  Speaker 3: to build much more right, and so you would likely
• 25:40
  Speaker 3: see some of the products out there explode with a
• 25:44
  Speaker 3: huge number of badly thought features. And you know you
• 25:51
  Speaker 3: mentioned before that I'm that I I'm okay in naming names.
• 25:55
  Speaker 3: Look at Microsoft Copilot right, it's it's kind of terrible.
• 26:02
  Speaker 3: And now there is a lot of a lot of
• 26:04
  Speaker 3: pressure Microsoft to fix things. But it felt like somebody
• 26:08
  Speaker 3: that Microsoft told everybody put a copilot icon and copilot
• 26:11
  Speaker 3: functions everywhere, and people didn't really think about what outcome
• 26:15
  Speaker 3: do we want to let people achieve? How do we
• 26:18
  Speaker 3: make it easier for people to do things? But they
• 26:20
  Speaker 3: just add the future of a future over future over feature,
• 26:23
  Speaker 3: and that's definitely not what we need to do, right
• 26:25
  Speaker 3: So I think as software professionals we need to think
• 26:28
  Speaker 3: about Yeah, right now, we can produce so much more,
• 26:32
  Speaker 3: but what still matters the most is what is the
• 26:35
  Speaker 3: problem that we want to solve and how do we
• 26:38
  Speaker 3: do this in a way that will actually simplify people's
• 26:42
  Speaker 3: lives and not making it more complex.
• 26:46
  Speaker 2: Another thing I learned about itself from your posting is
• 26:49
  Speaker 2: a really lot of sci fi, am You're I did, yeah,
• 26:55
  Speaker 2: And I think you even alluded to it right, like
• 27:00
  Speaker 2: to chat below the post, which is is so much
• 27:03
  Speaker 2: sci fi now seems a little bit redundant because it
• 27:07
  Speaker 2: wasn't able to kind of concretely anticipate AI. Some of
• 27:10
  Speaker 2: it looks quite quaint and outmoded, even if it's at
• 27:13
  Speaker 2: a thousand years in the future, you know, So, do
• 27:16
  Speaker 2: you feel like, you know, elaborate on that a little
• 27:19
  Speaker 2: bit more? And if you were to write your own
• 27:22
  Speaker 2: sci fi presently, what kind of future would you imagine?
• 27:26
  Speaker 3: H Look, that's a tricky one. The thing is so
• 27:30
  Speaker 3: fi was never really about predicting the near future, right,
• 27:34
  Speaker 3: most of it, at least the thing that made it
• 27:38
  Speaker 3: exciting or things like faster than light to travel or
• 27:42
  Speaker 3: contact with alien civilizations and to that kind of scope.
• 27:48
  Speaker 3: By the way, we're no work closer. Right as far
• 27:51
  Speaker 3: as I know, we haven't made any major leaps.
• 27:54
  Speaker 2: In broken disclosure, announcements, made busy and maybe another way
• 28:02
  Speaker 2: of sci fi is getting a little nerror that could
• 28:05
  Speaker 2: be absolutely.
• 28:07
  Speaker 3: Yes, So sci fi what it doesn't do is to
• 28:15
  Speaker 3: reflect the speed of AI. Right, So if you look
• 28:19
  Speaker 3: at most of the narrative, it's still narrative that is
• 28:22
  Speaker 3: centered around people or aliens, but still you know, thinking
• 28:26
  Speaker 3: biological beings, because those are the things that matter for us, right,
• 28:32
  Speaker 3: those are the stories that we want to hear. We
• 28:34
  Speaker 3: want to hear the stories, the struggles, the adventures of people.
• 28:40
  Speaker 3: And that brings to me a bigger reflection, which is
• 28:45
  Speaker 3: where are we going to be a thousand years from now?
• 28:49
  Speaker 3: Look at how much AI has evolved in the past
• 28:52
  Speaker 3: three years. It's completely mind blowing, and so it's a
• 28:57
  Speaker 3: huge question mark for me. I really have no idea
• 29:01
  Speaker 3: where the world is gonna be.
• 29:04
  Speaker 2: What do your go to AI movies like, if you
• 29:07
  Speaker 2: have to name one or two whatever, whatever one's for you.
• 29:12
  Speaker 3: Look, Probably one of my favorite books ever that now
• 29:16
  Speaker 3: is a movie is June. Have you read it?
• 29:21
  Speaker 2: Yeah, Joe, I've got a bit. I never read. I
• 29:24
  Speaker 2: never read the novels. I love the movies. I've watched
• 29:28
  Speaker 2: them both about half a dozen times. Yeah, read the
• 29:31
  Speaker 2: novels or even match. Everyone tells me, you know, I
• 29:34
  Speaker 2: grew up with so many people who read them. I
• 29:36
  Speaker 2: had so many accounts going back years and years and
• 29:38
  Speaker 2: every and everyone says, oh, the first one's amazing, the
• 29:40
  Speaker 2: second one was okay, and then it goes really really bad,
• 29:42
  Speaker 2: really quick. I hear that's so often.
• 29:46
  Speaker 3: Another one, another one that is now a TV show
• 29:49
  Speaker 3: is a Foundation by Isaac Casimov. So that's another great,
• 29:53
  Speaker 3: great one.
• 29:55
  Speaker 2: When was the last time? Off topic a little bit,
• 29:57
  Speaker 2: But when was the last time you watch the first
• 30:01
  Speaker 2: Matrix movie? Oh?
• 30:05
  Speaker 3: Probably ten years ago.
• 30:06
  Speaker 2: I haven't.
• 30:07
  Speaker 3: Actually, maybe I should do that.
• 30:09
  Speaker 2: You've got to rewatch it post you know, lllms. You
• 30:13
  Speaker 2: know you've got to pose AI explosion. You've got to
• 30:16
  Speaker 2: rewatch it. It just looks it's even more uncannily eerily prescient.
• 30:21
  Speaker 3: I have not thought about that.
• 30:23
  Speaker 2: Yeah, the backdrop of the twenty first century so far,
• 30:26
  Speaker 2: it gives you, It gives you chills and it's still
• 30:28
  Speaker 2: a great movie. But that's a great idea. You have
• 30:30
  Speaker 2: rewatched it for a while, I would, And I've got
• 30:33
  Speaker 2: a great question off the back of its view stands
• 30:35
  Speaker 2: to lead us towards the end of the conversation. You
• 30:40
  Speaker 2: go to try me as it, you know, particularly overly
• 30:42
  Speaker 2: sentimental about the past. But if you could preserve three things,
• 30:48
  Speaker 2: I'll take any number of things, to be honest with
• 30:50
  Speaker 2: from AI driven disruption, what would they be?
• 30:55
  Speaker 3: Well, we try to preserve humans around there? Okay, just one,
• 31:01
  Speaker 3: just one thing us. Look, let's go. You know, I
• 31:10
  Speaker 3: completely change everything in three years, right, what's gonna happen
• 31:13
  Speaker 3: one thousand years from now? And I think I will
• 31:17
  Speaker 3: obviously completely redefine society what really it will redefine what
• 31:23
  Speaker 3: work means and today truth is, you know, a part
• 31:28
  Speaker 3: of the meaning of our lives is around work, not
• 31:32
  Speaker 3: for everybody, but for some of us, right, And so
• 31:36
  Speaker 3: I think fundamentally I will redefine the whole concept of
• 31:41
  Speaker 3: why we matter and what our lives matter. And I
• 31:45
  Speaker 3: have no absolutely no idea what the answer of that
• 31:49
  Speaker 3: is going to be. So at the same time, it's
• 31:52
  Speaker 3: a very exciting future when I'm embracing but it's also extremely,
• 31:58
  Speaker 3: extremely scary, and so if you gave me absolute power,
• 32:03
  Speaker 3: I might even decide to stop AI and to kill
• 32:07
  Speaker 3: a eye, which is really weird for somebody that. Yeah, no,
• 32:12
  Speaker 3: it's it's crazy enough about.
• 32:13
  Speaker 2: The Gum books to know what the Butlerry and Jahadde is.
• 32:17
  Speaker 3: I don't know what that is.
• 32:18
  Speaker 2: Isn't that Isn't that in the Dune universe where they
• 32:21
  Speaker 2: ceased AI, they had to they had to turn it
• 32:24
  Speaker 2: off effectively?
• 32:25
  Speaker 3: Yeah, yeah, exactly, No, absolutely, absolutely absolutely, which is, by
• 32:28
  Speaker 3: the way, it's the premise of some of these books,
• 32:30
  Speaker 3: because otherwise, you know, the narrative wouldn't work. But I wonder,
• 32:34
  Speaker 3: you know, maybe maybe we should just turn it off.
• 32:37
  Speaker 2: It's started here, the revolution started here. But before we
• 32:41
  Speaker 2: let you give it Sam, it would be a shame
• 32:43
  Speaker 2: not to do one fulsome next thing. Plug what's something
• 32:47
  Speaker 2: your team's building right now?
• 32:49
  Speaker 3: But really excite? Uh, so many things. It's like choosing
• 32:54
  Speaker 3: between your kids. You can't. But recently we put a
• 32:58
  Speaker 3: lot of attention around Spark with the launch that just
• 33:01
  Speaker 3: came and this ununbelievable product. So I will not go
• 33:04
  Speaker 3: in in the details of that because I think it's
• 33:07
  Speaker 3: you know, I said that so many times, but what
• 33:10
  Speaker 3: is coming right now in terms of agents as part
• 33:14
  Speaker 3: of workspace, So in the broader assist concept, that is
• 33:19
  Speaker 3: unbelievably exciting and it will really really find the way
• 33:24
  Speaker 3: of the broader IT ecosystem will operate not only the
• 33:27
  Speaker 3: support organization, but how everybody in it is doing their work.
• 33:32
  Speaker 3: Rather than keep doing things by hand, they're going to
• 33:35
  Speaker 3: be able to use workspace every time they need to
• 33:38
  Speaker 3: do something, create an agent for that, and then automate
• 33:41
  Speaker 3: that work. So we're putting a lot of advancies into
• 33:44
  Speaker 3: that part, and I think what we're gonna see coming
• 33:46
  Speaker 3: in the product in the next three to six months
• 33:49
  Speaker 3: will be really changing the game for itighty operations, more
• 33:54
  Speaker 3: broad Land in general. So I'm really really excited about that.
• 33:58
  Speaker 2: Awesome, So I'm gonna to have you back on keep
• 34:01
  Speaker 2: on posting. We're enjoying them there.
• 34:04
  Speaker 3: Thank you so much.
• 34:05
  Speaker 2: Cheers.
• 34:08
  Speaker 1: That's it for this week's Deck show. We'll be back
• 34:10
  Speaker 1: next week, and if you don't already, be sure to
• 34:13
  Speaker 1: follow the show so you don't miss an episode until
• 34:16
  Speaker 1: next time.