What Is a Sovereign Engineer and Why Do You Need to Become One?

00:00:10: Hello, I'm your host Michael Doudin and here with me today is Russell Miles to discuss developer platforms.

00:00:17: Russ.

00:00:17: Thanks for joining me today.

00:00:19: It's my pleasure to be in my friend

00:00:21: fantastic.

00:00:22: so.

00:00:23: you recently wrote a little book The Sovereign Engineer.

00:00:26: it's yeah, its a great entry into AI literacy for software professionals.

00:00:33: Can he give me just a little bit about the premise of what we were thinking when you are putting that book together?

00:00:38: There was a book that was about two and half years in the making, maybe little less than two-and-a-half but it was... The premise of its changed a lot since I began it.

00:00:51: And everything then went through on my way.

00:00:53: It's different books to begin with from what i ended up With.

00:00:58: Now actually have potentials who publish the book that I was originally writing.

00:01:03: So, maybe putting it out in a couple of weeks as well which is another in the sovereignty idea but its different audience completely.

00:01:13: The second book will be for new engineers and people entering this industry where AI's present.

00:01:21: And what does that mean to you?

00:01:22: As an engineer now?

00:01:24: But yeah thats my original idea In my journey, I realised that was writing a book for folks like myself who are trying to figure out what our identity is changing too.

00:01:40: As we work with AI and what does it mean?

00:01:44: What's the ecosystem around us?

00:01:46: That's where one day very distinctly... ...I was talking to Kevlin Henney on occasion usually quite frequently And one of the things that we sometimes just comes out as conversations is this sort of language, which used to describe things really matters.

00:02:05: One other thing I remember...I think i brought up the term Habitat and said when im building platforms Im trying establish a habitat where I get to thrive make decisions and be creative as I need to be.

00:02:23: The functions and safety nets, guardrails.

00:02:26: And everything else that's around.

00:02:28: the software development lifecycle is also there to support it.

00:02:32: so I say when i build a platform I don't build A product So much or at least like It's a product but I Don't think of it primarily Is That?

00:02:40: I Think Of It As A Product.

00:02:41: That Is A Place.

00:02:42: It'S A Place Where I Can Do Great Work.

00:02:46: What happened in the last few years is, of course we've started to see that place has inhabited not just by my cognition but another form of cognition what the LLMs bring.

00:02:57: And I'm careful here to call it Cognition Not Consciousness.

00:03:01: There's been a lot wonderful press recently about how these things might become conscious.

00:03:06: We don't completely understand human consciousness yet.

00:03:09: So i think suspending any judgement in those areas Is a sensible recourse.

00:03:14: But when it comes to the book, It became this journey of.

00:03:18: what does it mean?

00:03:20: To still be a software engineer in this world where there is other cognition.

00:03:27: In The system?

00:03:29: What has the platform have to feel like?

00:03:31: What are the interactions with their behavioral differences?

00:03:34: and I created This called a ladder but its also you know framework of progression so as establish the platform that contains the right habitat elements and harnesses, in other bits of pieces.

00:03:49: You begin to provide a launchpad or platform if you will, to behave differently?

00:03:56: To work

00:03:57: differently?".

00:03:58: And then the question becomes what is the thing that you need to retain?

00:04:03: What am I adding?

00:04:04: If i'm not writing as much If I'm reading the code even, if i'm struggling to read some of the code that's produced because it may be very distant from my current perspective and reference frames in my head.

00:04:18: What am I doing?

00:04:20: In that arrangement?

00:04:21: what am I bringing into creative endeavour?

00:04:24: And remember this lovely conversation with Alistair Coburn over at Vienna I think maybe a month ago not so long ago when we were talking about books and bits, some pieces.

00:04:35: And he was giving one of his workshops... ...and it was a workshop around I think it's around agility.. ..and then software engineering….

00:04:45: ..and the hexagonal architecture.

00:04:49: Yes!

00:04:50: He had great distillation in almost everything that I'd written in the book….

00:04:53: …and published on the book months before which is decision making Is what we do When we're writing code, were making decisions.

00:05:01: We are making decision at the granular level of the code... ...of the syntax, semantics and architecture.

00:05:08: design If you keep going out will make a whole host of decisions... ...white rapidly as work through development.

00:05:16: A sustainable pace for human is understanding those decisions And making them as well.

00:05:23: because ultimately with accountable individual in all this or the accountable team perhaps.

00:05:30: And so, the book was written.

00:05:33: really with this emerging throughout The Writing of the Book if I'm being honest...of okay what does it mean to be an accountable individual?

00:05:40: What do you have to retain?

00:05:43: the sovereignty in this arrangement, because no one's going to blame the LLMs for that code you've shipped.

00:05:50: It will come back on your team and it is normal.

00:05:54: The way world works.

00:05:58: we have to retain sovereignty over our decision-making process.

00:06:05: so how do then build workflows?

00:06:11: There's

00:06:16: so much of what you just talked about that I want to dive into.

00:06:21: Had a fantastic conversation with Alastair actually, couple weeks ago on the podcast.

00:06:26: So it is lovely have these two conversations dovetailing each other so nicely.

00:06:34: Talked me little bit About accountability thing for moment.

00:06:40: Like you mentioned the entire world is built around human accountability.

00:06:44: We always.

00:06:45: there's always a person responsible for things.

00:06:46: when things goes wrong.

00:06:48: It it's always some either Either some executive or some developer that gets fired, right?

00:06:54: You know There's always humans and look how do you see that changing over time?

00:06:59: Or do you think that rather than trying to change that we need to Be very careful about how would build humans into this process?

00:07:11: This is this is a great question.

00:07:13: Okay, this is the question I think at the true heart of what's changing.

00:07:19: One of the themes in the book is around changing literacy and Literacy is about agency.

00:07:26: every shift in literacy brings A new agency anew changed to craft and a new democracy, or if you like spread and distribution of where that agency can sit.

00:07:41: So we're looking at a world now where people can produce very strong applications in code bases with minimal knowledge on how they got there which is wonderful and challenging to what we do.

00:07:59: so yeah the first thing he's doing standard any change.

00:08:03: literacy.

00:08:04: the question always gravitates to who retains enough agency to be held accountable.

00:08:13: And I think that's going to be a question we as society explore with AI, it is important.

00:08:26: what is their processing approach?

00:08:28: How are they trained, how are they adjusted and tuned.

00:08:33: And then...how do they don't evolve over time?

00:08:39: that's why we get new versions you know.

00:08:43: so I think until with the current technical stack that we tend to work with i think there should be very little threat to the importance of our accountability as humans and an agency.

00:08:59: I don't...I think what we're doing at the moment is, we are figuring that out with seeing how where we go too far and surrender too much for decision making two processes do not understand our embodied existence.

00:09:13: they don't understand our world.

00:09:16: They aren't designed They're literally not embodied in that way.

00:09:20: That cognition doesn't have that way of functioning, whereas we do right?

00:09:25: And so that's our advantage is our accountability and it's our perspective, right?

00:09:30: It comes to the mix now within a future world where if We do manage to create some other things that we rather dramatically think we are Now but I had my skepticism in that regard But if we do start to create intelligences that have an embodied sense of accountability, If they do understand much more about the real world like we do.

00:09:57: And judging that when itself will be quite challenging because We're going back to Thomas Noor's work around what is it?

00:10:03: Like To Be a Human?

00:10:05: What does It like to be?

00:10:06: abat was his particular paper in this regard.

00:10:10: What is it like to be an LLM?

00:10:12: Is one of the exercises I go through when I deliver a workshop on working with agentic engineering, because helps you understand how that cognition... what are the limitations and power that cognition actually is.

00:10:25: If we were to create cognition that was like our own then we could have a sensible conversation about accountability and responsibility.

00:10:34: but until any cognition can And the consequential world that we live within then I don't think We should ever be handing over Accountability and responsibility to anything.

00:10:51: That cannot grog those things because right?

00:10:53: We open up the floodgates too extremely dubious decisions Because we make decisions because of our three D reference framed cognition, and understanding Of consequence in our world, which is why we can be held accountable.

00:11:14: LLMs cannot be so as they currently stand.

00:11:17: the tech stack is not it's not the right direction for that sort of thing and I think

00:11:31: It's interesting that when a significant new technology like this rolls out, it can be very hard for humans to accurately judge the consequences of risk or human element.

00:11:45: We saw this with self-driving cars where you would have some people are driving in car and letting them assist but others sitting there reading while the car drives them, right?

00:11:58: And there's different levels of understanding where technology is in that sort-of example.

00:12:03: I like it because maybe a bit easier for us to grasp what happens under the covers when we're pulling elements into our software.

00:12:13: so some failures are turning too fast over machine cognition that they didn't necessarily understand, and yet some of the people have been very slow to adopt.

00:12:28: It's like well I don't understand how it was trained...I don't know.... I don' t understand what thought process is or how its getting into results..and i really want to understand that.

00:12:37: so then question than I had as our systems get much better at self-reporting.

00:12:48: Here's the material that I'm using to come up with this information.

00:12:51: Here is the steps or the logic that I am applying, here are training going into it?

00:12:58: Do you think we'll end in a place where humans are more likely not only trust machine but start demand sort of reporting each other right if becomes an interesting type exercise and even relevant work?

00:13:16: really doing her

00:13:18: lovely question.

00:13:19: It's a really lovely question.

00:13:22: Do we really do that with other humans?

00:13:24: Is one of the first thoughts that comes to me is, when we approached a human to work with them... We imply that we see the same world and in truth I think most psychologists would say we don't.

00:13:43: Most people researching into neuroscience would say build of the world and I built over.

00:13:51: our experiences for our lifetimes are very different.

00:13:55: Even on basic levels, we may literally read the same headline or view the same picture and see it fundamentally differently.

00:14:07: but that is kind of a slight side note in the debate.

00:14:14: i think the interesting part is around, continue to understand what you're dealing with and what it can see.

00:14:28: In fact the very first step that I recommend people do in a book Is ask the LLM What It Can See?

00:14:35: Its Very Rarely The First Prompt.

00:14:37: Anyone Uses just a notch and go, okay what can you see right now?

00:14:46: before I build any mechanisms?

00:14:47: Before i populate an agents.md with anything?

00:14:51: Before I create co-pilot instructions whatever mechanisms of control surfaces you apply to these your interactions those are things that come later.

00:15:01: let's start with What Can It See?

00:15:04: What Can it Reason About?

00:15:07: And I do think we're going because the LLMs are very good at being able to report on themselves.

00:15:13: I think they don't have human reticence to fictionalize, there's storytelling devices but limited in what their doing so better reporting more truthfully and that respect i think... But we should always hold them accountable for the vendor.

00:15:35: The vendor, if the vendor is going to allow their LMS To tell you how they reasoned about things then You want to make sure that your not seeing Dremptup hallucinated text around these Things as well but assuming They're making a pretty good job.

00:15:50: That you can verify Alongside and in heart of that statement Is part Of the answer why I wanna give you which is because you mentioned trust Which was bigger thing They can account for.

00:16:05: I think they can be.

00:16:07: one of the things that elements and elements are very good at is critiquing themselves.

00:16:11: So, they don't have our blind spot And we have a very human blind spot in that.

00:16:18: We find it Very difficult if not almost impossible to see you the errors in our own work.

00:16:24: Which one of these?

00:16:24: as we have editors?

00:16:26: It's one of their reasons that That we have peer review this.

00:16:30: there's a lot that we cannot see when we have done it.

00:16:33: And if you've ever written a paragraph and use the word for twice, then you literally cannot see it.

00:16:40: When you come back to that paragraph until something puts a little green squiggle underneath it says I think this is wrong.

00:16:47: Right?

00:16:47: LMS don't have that line spot so they can be asked to critique their what they produce which is a helpful mechanism To begin to attack elucination.

00:16:58: You can ask an LLM Tell you what the gaps are and what it just responded with, things like that.

00:17:05: So I can help with the auditing process.

00:17:08: but bigger than that is your point there about trust?

00:17:11: What can we trust in these arrangements of what should be

00:17:14: trusting?".

00:17:16: And i think in those cases because our interaction modes or collaboration with the agents he's moving to treat them as collaborators which is how the Sovereign Engineer moves through towards, then it isn't about simply trust.

00:17:38: It's what are the mechanisms you would use for a human?

00:17:44: If you were collaborating with an engineer who just joined your team?

00:17:48: because every time you launch in session with Clore Code its new engineer joining your team so you set up conditions where they can relearn where it was, and you can use that context in our architectural guidance on everything.

00:18:01: It needs to be a good collaborator!

00:18:04: You'll see these understand your workflows too as well of course once you've set those conditions up.

00:18:09: then the next question becomes if I were human doing this things what would you verify?

00:18:17: What do look for an understanding about what they have done?

00:18:20: What decisions would you like it to tell you that he has made?

00:18:28: There's a lovely agent I built not long ago called the Choice Cartographer.

00:18:33: And what is does, tells me in given session and work or PR even... ...what are choices of decision that can discern from this?

00:18:44: And it tries to tell stories about those decisions.

00:18:47: It tries to give you the whys around them and a little bit more information, because it turns out that's really good way to explore the decision-making process is to understand the narratives around these things.

00:19:00: again I don't want to name drop too much but i stole that from Kevin Hennie as well... ...and The Choice Cartographer used to be called The Henni when he was in my first version of The Agent.

00:19:10: all this reason, you know that for every time... Every usage of a pattern there's a story around why we thought that pattern was applicable right now in every architectural choice and everything else.

00:19:20: We do as well.

00:19:21: so when it comes to building trust with what an agent is doing or who agents are doing?

00:19:29: Or even your agent teams are doing then I think one of the hearts account, well no not countable because you cannot make AI accountable as it currently stands but you can make it responsible for maintaining human understanding.

00:19:47: and this is a phrase I'm borrowing from some wonderful researchers over in Canada who have coined the terms intent debt and cognitive debt.

00:19:57: There's some of the factors that fed into Setting aside those debts for a moment, the phrase helping the human maintain understanding so that it... The human can still be the decision-making party in this.

00:20:16: I think that's the sort of trust we're looking to build.

00:20:19: We are looking to built the trust That your agentic system, the platform around It, the harness around it, habitat you work together is set up In such way that it understands its role to bring you with the process, not be careful of your human limitations because we can't make too many decisions in a short time frame.

00:20:46: We get decision exhaustion and the worst thing about it is that decisions.

00:20:57: and yet our reserves, our reservoir is what I call it in the book.

00:21:01: And then some of the work i've done since that reservoir can be depleted to nothing and you don't know.

00:21:08: so one of things The Agents Can Do Is Not Only Maintain Your Understanding Of What You're Doing So That You Can Still Make Decisions.

00:21:15: It Can Surface To Decisions will detect when I am probably getting near the limits of my ability, because i can't detect them.

00:21:34: So yeah it's a trust in the conditions of the collaboration rather than focusing too much on conditions of trusting everything it produces and that...I think this is also just uh very quickly what-.

00:21:47: I think he's the heart of difference in why seeing what we're trying to achieve?

00:21:51: It's less about handing off and trying to build bigger specifications, batches so that we're handing off more.

00:22:01: And then having the challenge of how do you verify what it produces?

00:22:05: It's trying maintain us in our collaboration such as make good decisions.

00:22:13: I think some respects going down a little bit rabbit hole thinking will specify generate that's it, specify and generate is our job now.

00:22:22: I think we'll still be collaborative partners in the arrangement of decisions might shift in level but i think there will always.

00:22:31: at the moment whilst with the accountable party will be involved in that decision-making process Making our agents responsible for maintaining our understanding of paying down our cognitive debt, if you like is going to be one of the most important factors to bake into those workflows.

00:22:54: Absolutely that makes a lot of sense.

00:22:57: so I know that for me personally

00:23:00: i get

00:23:01: To The Point Of Exhaustion and A Number Of Fronts Like Decision-Making.

00:23:04: you make enough decisions during the day at work.

00:23:06: You can't even decide what to make for dinner, right?

00:23:08: Like these become hard problems when they shouldn't be and sometimes when you're doing reviews like trying to review a really long pull request for example um...you get to a certain point And then you start just kind of skimming instead of reading an understanding.

00:23:26: it's not building the middle model anymore.

00:23:28: It is sometimes hard to realize that we've hit that point.

00:23:34: So it's really valuable to have those checks and balances on yourself as well.

00:23:38: And this whole thing is one of the reasons that I really like, way you've coined a term on habitat engineering in building the Habitat because it puts... Like he said words matter!

00:23:52: It put emphasis where we can live thrive and succeed and there are organisms living in shared space right?

00:24:03: It's very different from some of the other terminology we use in industry.

00:24:08: Yeah, and it was all very much on purpose.

00:24:11: that at heart is habitat thinking And there certainly a history to that term.

00:24:21: I've tapped into Richard Pigabriel work on habitability you know, for example one of the things he talks about in his book Patterns Of Software is how a large hierarchical structure of abstractions.

00:24:38: Is anathema to any habitable system or system where you can explore and bump around?

00:24:45: One thing I talk about when we're doing platform engineering... ...is Can You Make A Platform Discoverable By Humans?

00:24:52: And humans like to bump around and navigate you know, sort of prod and find their ways around things.

00:24:59: We usually we literally use that phrasing I want to find my way around the code base You do because it's the same with a platform.

00:25:07: A lot of platforms are though designed not To be explored Not to be bumped around not for mistakes to Be made as safely As possible so i can back them out right.

00:25:18: And one Of The key features they talk about Sometimes what I'm doing platform engineering is most platform engineering demos.

00:25:26: when someone's got a Vendoring of product to you if it's going to automate your world away.

00:25:30: It's gonna be fabulous.

00:25:32: Is the creating new service feature.

00:25:34: with one button click?

00:25:36: You can create any service and my point And I use this as an exercise at the beginning of platform engineering workshops, okay If do ship that button and no ones pressing it why?

00:25:51: And part of the reason that I'm trying to move people towards is an unsafe button.

00:25:58: It's a button that creates a new service, so it's probably creating all the infrastructure to support you-all the pipelines even registering as a new thing to be maintained in.

00:26:08: however your maintain your tickets.

00:26:10: So there are lot goes with it.

00:26:13: But we've all done this.

00:26:14: We have hit that button and named it wrong.

00:26:17: Oh i need change now And I can't change it because he's now followed this enormous construction job.

00:26:25: that changing one thing is going to mean everything else.

00:26:27: Is then out of step, so you've maybe made it feel unsafe?

00:26:30: You do This and if i get It wrong If I don't want That service anymore is there a way Of getting coming back from it?

00:26:38: Can I can I reverse the decision?

00:26:41: So what I Do is when I'm helping people understand platform engineering I come at it From What are The Decisions that are trying to be made.

00:26:50: And then spread out from that, so I use the OODA loop which is comes out of dogfighting in all places.

00:26:57: but okay what do i need to observe?

00:26:59: What do we need to orient towards?

00:27:01: or what do you need to help make a good decision and perform an action?

00:27:06: Then can I provide away where this is safe as possible to

00:27:11: do?".

00:27:12: That might mean providing reversibility.

00:27:15: it may editability or modification, all of that should be designed into the habitat you provide.

00:27:25: And you'll only do that if you understand that your helping humans fallible human and fallible agents to be able to do safe operations.

00:27:34: when they're discovering what they can do They want a bump around.

00:27:37: figure things out.

00:27:40: It's amazing for me how little of that we apply when building our platforms.

00:27:48: I jokingly in my keynote on platform engineering, put up a button and said do you want to press it?

00:27:57: What about here?

00:27:58: what if i add some information like run pipeline on the button?

00:28:01: Do you wanna press now And everyone's going?

00:28:04: You can feel nerves.

00:28:06: well don't know what is does right problem.

00:28:09: You don't know what it does.

00:28:11: If you can change when its done, there's a whole context to that action and decision we need to explore in order for someone to confidently press the button.

00:28:34: It's about everything else around it, which is the habitat that we exist in.

00:28:42: Almost sounds like you're talking about user experience there for a minute!

00:28:47: And I absolutely am This.

00:28:49: this one of the comments That has come back from The Workshops Is okay.

00:28:55: well i've had two Comments.

00:28:57: This is product management and this is User Experience Design that has an experience.

00:29:08: What is it going to succeed on, being a good product with a good experience?

00:29:14: Everything else is mechanism!

00:29:17: It's okay how do we implement beneath?

00:29:20: but deciding what you want to empower people to be able to do in their habitat Is where the decision making starts for you rather than How Do I Automate A lot of infrastructure.

00:29:33: If you come up from that bottom, then You can have a very elegant orchestrated or automated system.

00:29:41: But are you creating your habitat where engineers creative people?

00:29:48: Are able to Feel empowered and autonomous enough to do their work?

00:29:53: And now how did we do that with the additional players Or habitants of the agents as well?

00:30:02: Yeah, absolutely.

00:30:04: So when you talk about some of the mechanisms on how to build a habitat I know that you spent a lot of time in platform engineering.

00:30:12: Are you thinking in terms platforms?

00:30:14: In terms of processes or procedures and techniques?

00:30:17: are tools like where's kind of the mix for You?

00:30:22: it well in terms of how we approach it Tools an implementation.

00:30:30: choices come quite late in The mix.

00:30:33: I focus primarily on the types of work being done.

00:30:39: So a lot when we explore platform engineering, but even if agents are involved in this... We look at software development life cycle and what activities people have to do?

00:30:56: What other things that cause them pain or should they be doing?

00:31:03: When you catalogue that, You can then begin to go okay what are the platform features That we want to implement?

00:31:12: How should we implement them?

00:31:13: these decisions Can come not too late.

00:31:17: We try to work in an agile fashion But they should be decisions made In service To the experience.

00:31:25: your trying to create one of The things I can use as An example there and We do this as an exercise in the boot camp, where I say okay.

00:31:37: What's the most common implementation choice on platform engineering?

00:31:43: Usually what bubbles out of question is infrastructure automation like Terraform or it's a dashboard?

00:31:52: right we have a back stage and one other variant of different dashboards.

00:31:59: And then I ask a question that is slightly rhetorical, okay?

00:32:05: It's a bit cruel.

00:32:07: How many engineers want to go to a web user interface when they're doing their work?

00:32:15: When do they want to jump there?

00:32:18: When is that important to them?

00:32:20: and it is rhetorically because the answer very few if any wants to jump.

00:32:26: so my point then let us explode this problem.

00:32:29: Let's look at what's happening here.

00:32:32: We're creating dashboards, we're creating reasons to move from one place to another... ...we are saying that this information can't possibly exist where you are.

00:32:44: it is so important!

00:32:45: You've got to go somewhere else.

00:32:47: What we're looking at is.. ..we're making someone make a cognitive leap Okay?

00:32:53: So were introducing cognitive load in just the switch to new location.

00:32:59: Right Does it justify that?

00:33:03: Is there anything happening in your dashboard That absolutely justifies that leap?

00:33:10: and you can also understand from a human perspective, okay When humans move From one perceived location to another Do folks know what happens in the human brain?

00:33:24: What happens in The Human Brain?

00:33:25: And this is by the way not limited To older people.

00:33:28: I should point this out.

00:33:30: Being one of those older people myself, when we move from one room to the next our short-term memory is lost.

00:33:38: it's dropped on purpose because we as embodied creatures moved from one setting to another and were rapidly getting into grips with a new settings.

00:33:48: so make space drop the old setting then go back up to where you went upstairs or while in the kitchen.

00:33:55: It happens.

00:33:57: It's just that when we're younger, we might retain more residue where it goes to the next place I suppose.

00:34:03: But we do the same

00:34:04: with our

00:34:05: engineers.

00:34:06: We go okay.

00:34:07: now you need to jump to SharePoints To find this document and understand what this process may be.

00:34:14: or You have to jump back stage Or press a button over here And every single time we are introducing cognitive burden.

00:34:25: If we want to retain short-term memory, We've got try and hold onto it.

00:34:30: And then remember when come back why were originally there.

00:34:34: So its usually a terrible experience from cognition perspective.

00:34:40: Then I'll say something even more rhetorical which is could we surface that information where they were at the moment?

00:34:51: Up until relatively recently, Ben it's quite hard to do that.

00:34:54: We might have to customise an IDE or you know.

00:34:57: there are lots of ways potentially but they're not easy.

00:35:01: with agents It starts become easier.

00:35:04: so I'll be using our agents again to facilitate human decision making by surfacing things which would typically jump into another room for Can we bring them gently and appropriately where their attention currently is?

00:35:20: And that ignites another set of creativity when we're looking at, okay what could a platform experience be like now?

00:35:28: That we can surface the right information At The Right Time in the right way to support the right decisions and actions where people are.

00:35:38: That becomes really fascinating as A Thing To Explore with With A Room Full Of People From Different Organizations

00:35:44: I Can Imagine!

00:35:48: giving some of your talks and workshops in the near future.

00:35:52: I think MLCon, ijs maybe

00:35:56: yeah

00:35:56: it's a big great opportunity for people

00:35:58: to

00:35:59: catch on.

00:35:59: so some of that.

00:36:01: well absolutely!

00:36:02: And I should point out that Yeah...I'm getting them at.

00:36:04: uh The Workshops at.

00:36:06: I Think There's DevOps, Cons, Jacks and other bits from pieces around Around the world..and I do run my boot camps and workshops a certain way.

00:36:16: So I have a philosophy of learning and training, it follows what i just mentioned about humans learn best when they're bumping around trying to figure things out.

00:36:25: Yes!

00:36:26: When we deliver a bootcamp its not the easiest way to do it.

00:36:29: um...I usually go in with a lot perhaps anything up through one hundred things that we could cover.

00:36:35: okay lots of topics lots of possibilities.

00:36:38: you know given One-to-two days That We Could Work Through.

00:36:42: And then What I Do is I have slides, but i don't use them.

00:36:48: Unless really feel they augment something.

00:36:50: that is going to be a way for us move along the journey.

00:36:54: Usually first question will ask in any of boot camps what do you think you want take away from this day or two days?

00:37:05: Then frame everything around it unless its off-piste say you know I'm in the right room but most of time it's on track and its things we could explore.

00:37:19: And so what i do is, look for a journey through everything that I've got in store to give me more value in those two days.

00:37:35: It's not easy.

00:37:36: Nice,

00:37:36: but I've learned this gets the best conversations.

00:37:40: it also really helps to not go in and pretend that i have all the answers To their specific situations Because you know as she had this at The last workshop Iran over in Berlin Where we were exploring one of the first questions was how do You convince people?

00:38:03: And really the question is a wrong question.

00:38:06: We're not trying to convince anyone, we are trying help them approach it and figure out for themselves... ...and make it visible to them and attractive.. ..and there's ways that you can bring people on an adoption journey.

00:38:19: A collective of people has very little to do with forcing or controlling any one Or even just telling what their feature does.

00:38:29: It's much more nuanced than that.

00:38:31: But as you have one person in the room go, no.

00:38:33: In my context we tell them to use it and when I dug into it with this incredible human being in front of me was like great!

00:38:41: This is a new perspective.

00:38:43: And he was saying well We work with external developers?

00:38:47: We have external development teams.

00:38:48: We don't have any developers.

00:38:50: We're trying to create platform that very specifically tells them how we want them To Work.

00:38:55: The SDLC Is Very Clear.

00:38:57: It governed and he worked in a government regulated industry.

00:39:03: And so that made perfect sense.

00:39:05: No, you don't have to think about how humans are attracted to your platform.

00:39:10: You actually tell them this is it?

00:39:12: This is the way.

00:39:13: if you don' do it this way contractually We've got a problem.

00:39:17: Yeah, I thought yeah That's another perspective.

00:39:19: but If i'd gone In with a module Like a load of slides that say this is how human behavior could be modified.

00:39:27: He would have sat there for two hours going none.

00:39:30: Of us is applicable to my world.

00:39:32: Instead because I run it more fluidly we can explore his world and go right?

00:39:37: This is gonna Be different view, but he's gonna be something you know.

00:39:41: There still might be things in what all those topics they could cover but they're going to be adapted very differently in your context, which is why I love doing the workshops.

00:39:51: So do and while there's very few slides

00:39:56: it makes sense to do it like that And from experience It Is way harder To Be That fluid and dynamic.

00:40:04: you have to actually Do a lot more prep?

00:40:06: And Have A Lot More Things Ready To Go.

00:40:08: But

00:40:09: From My Perspective And Get This Feedback From People At Events When you have customized your delivery to the place that they are, it takes so much more out of it.

00:40:22: And like I said conversations is better?

00:40:25: That's quite selfishly why I do fair enough because I get enormously challenged by and uh i used to deliver courses back in the day for spring source.

00:40:37: um In London when I was delivering those courses at four-day courses and you've got a lot of modules, exercises.

00:40:45: All-of-it great content really first class training.

00:40:50: however I realized that when i delivered these four day courses all my excitement came from the conversations between the modules an over lunch.

00:41:03: so what is maximized?

00:41:05: That habitat for learning?

00:41:08: it's all habitat engineering in my head so I just have a habitat where i have a bunch of people with completely diverse contexts coming into a room wanting to learn how to build platforms that support possibly constrain but um usually support and help their people thrive, then there's obviously an eye towards the agents as well now.

00:41:31: Okay?

00:41:31: But There are moves and shakes, decisions we can make but you will make them for your world.

00:41:39: And all of the fun is in those conversations around these things.

00:41:44: It maximizes that fact that you're in person sharing space with other human beings.

00:41:50: otherwise it could just be a video!

00:41:55: I've written another book out there called How to Speak which i'm still writing actually because And it relates to this, right?

00:42:05: So one of the things I realized early on is that when i do a keynote even.

00:42:10: I sometimes won't have slides because its called A Talk and I really talk about something!

00:42:17: And I studied...I used to study comedians who rarely have slides how they can capture an audience take them often quite long-winded stories which, you know have a payoff in the end and they hear I laugh.

00:42:31: And there's all that sort of thing that comes with it.

00:42:34: but what i studied was That connection that you're looking for With?

00:42:39: The people In the room.

00:42:41: so i strive For that when i do keynotes and now When i do my regular talks and when i Do My workshops i want to walk away from that feeling like we've built A bit Of a connection around those challenges You had Around the things i'd Like To Share.

00:42:55: And to do that, I find more often than not some of these more grandiose mechanisms like slide design don't really actually serve me.

00:43:04: Um i can achieve those things a lot better without those things in the way

00:43:08: right?

00:43:10: Nobody says we all have to use PowerPoint.

00:43:14: No!

00:43:15: That's it.

00:43:16: no one mandates.

00:43:17: but my goodness are there whole books written on how to lay out slides And in fact, I mean again on name drop a little.

00:43:24: but i was trained by Damian Conway as an Incredible public speaker.

00:43:29: if you get chance to see him.

00:43:31: See him He's far better than I am.

00:43:33: he probably far better night ever will be.

00:43:35: here is a master of the craft?

00:43:37: He has gone deep into this and I had this one day with him where he was training me on how to do these things.

00:43:44: and One other thing she pointed out what two things they took away from it As well as a whole host of other things.

00:43:50: But two things that relate to this, is that when he's designing slides.

00:43:55: He starts with one word in the middle That just frames The thing he wants To talk about and it's partially for him Partially For the room but It doesn't give you everything.

00:44:06: so always people are gonna read it And they're not... You're Not listening to you anymore.

00:44:10: So thats a simple Thing there.

00:44:12: I also smiled at the time because there are so many books on slide design, which primarily i think about using slides to communicate people who were not in the room that want to read those slides.

00:44:24: But if you're In The Room You Want To Engineer A Desire To Listen To You Which Is Not Gonna Come Through The Slides.

00:44:33: And It's The Same On The Courses.

00:44:35: If I Put a Slide Up I'm Usually Wary Because Its Distracting Everybody From What Im Saying.

00:44:42: But yeah, you're right.

00:44:43: No one told us we had to use slides to do talks and I weren't my favorite speakers of all time although he had an interesting form for his speaking was Abraham Lincoln.

00:44:54: You know i've even studied his Speaking approach And the beautiful storytelling that he could Do not only as a leader but as a speaker?

00:45:03: The two things are kind Of entwined.

00:45:06: so Yeah, I recommend anybody who's thinking of doing this sort of thing at All Look at those people, learn little bits.

00:45:13: Start practicing do it by doing and look for connection.

00:45:18: Yeah absolutely yeah.

00:45:20: I actually got my early introduction to public speaking from my uncle who was doing professional illusion magic.

00:45:29: Oh yes attention!

00:45:32: Look over this hand over here don't look at the

00:45:34: side.

00:45:35: You obviously have a lot of things going on.

00:45:37: you've a book you've just finished, other books in the work.

00:45:40: A lot of workshops and as you admitted when he started this book, The Severed Engineer two-and-a-half years ago... You started writing a different book than you ended up with.

00:45:52: so talk to me a minute.

00:45:55: what are the industry developments?

00:45:57: What is like the pivotal moments both during the writing even before that leading into inspiration that have been kind of pivotal or fundamental to giving you, to this point in the thought process?

00:46:11: It's a very good question because when I started writing the book one could easily argue that genetic engineering was very much his infancy.

00:46:23: You talked earlier about trust and i don't think there was even much cool, per trust in output or contribution at that point.

00:46:32: It was more aspirational.

00:46:33: it was I think this can because of the way... Because what i did is..I went deep on how LLMs work and if I could figure that out And then I went deep On How Humans Work In Our Brands And Then Going Okay these two things are different.

00:46:48: So There's got advantages in This Collaborative Relationship.

00:46:54: Okay, so at that point we were very much in a directive relationship.

00:46:58: right?

00:46:58: do this thing.

00:47:00: tell me what you've done.

00:47:01: show me.

00:47:01: What?

00:47:01: You're done now it's wrong.

00:47:03: I'll try and figure out how to prompt you better next time.

00:47:06: That was the relationship.

00:47:07: And well they kind of movements and changes that happened over writing the book Were more about How that collaborative relationship would change.

00:47:19: So there would be more potential for agents to do more work for you.

00:47:24: There would be the dangers of cognitive surrender to their decision-making process, there will be the danger that Margaret Ann Story and her research group came up with these three debts – technical debt we know off but when it comes a cognitive intent debt what they're doing is losing touch on how this works.

00:47:49: factors became more possible as the agents and models become more powerful, more able in their contribution.

00:47:59: So what I found was that I was plotting a course which was very much tracking with what the LLMs were capable of bringing And what I realized late on in the process, and it's always late on is you go.

00:48:19: Oh no!

00:48:21: No...I was trying to maintain decision making which is relevant regardless of what the collaboration ends up being.

00:48:30: so i was very fortunate as a bit lucky.

00:48:32: um..i must have been a bit lucky too find myself looking at What I feel is one of the problems at heart, challenges that are hard to be arrangement.

00:48:41: If i could have written a book on nuances of LLMs and think it would be very rapidly out-of-date.

00:48:46: Um...I could've written a deep dive in how it works but what's not rapidly our date need make good decisions And don't see changing either at some time.

00:49:03: So hopefully I had a lot of lovely comments back on the book.

00:49:07: that it's one.

00:49:07: those books they feel is going to stand the test of time, you never know!

00:49:12: You never know.

00:49:13: and...you know i'm sure people thought they were writing an incredible book about how to work in an IDE two years ago?

00:49:22: And now they're saying whoops..I've not written a book for all times at this point but we can't tell I think anybody who, and actually I like this about our current moments in time.

00:49:37: Anybody that claims we know exactly how the world's going to feel as a software engineer?

00:49:43: In five years' time they're lying!

00:49:47: And i think it is awesomely interesting.

00:49:50: It

00:49:50: has brought me back into industry with some respect.

00:49:55: so when you get very ill start re-evaluating What are you doing with your life?

00:50:01: Was what's, was your reason for being here.

00:50:04: And this renewed focus in habitats and places where we can be creative out the role of platforms a roll of agents in that I think.

00:50:15: well it certainly brought me back kicking and screaming into our industry and loving It For It.

00:50:22: i know That We Get I Know We Feel Attacked At Times many facets of what LLMs and artificial intelligence will bring.

00:50:34: And I don't think all those concerns or worries are overblown, it's fair to be concerned and worried.

00:50:43: It is also fair that you feel your identity has changed because the craft has been changing.

00:50:49: Your relationship with tools have changed.

00:50:52: That means our relationship for you too!

00:50:58: But what's exciting for me is that we are redefining, well it means to build software.

00:51:06: We're seeing a greater democratic surge of people being able to be builders and creators.

00:51:14: It's on the part with you know suddenly everyone realizing they can do most of the work They wanted to do in Excel then now realising they can Do it with agents?

00:51:22: And thats great!

00:51:22: You Know its easy as a software engineer of thirty years to go, yeah but they're not going to get TDD and then I'm gonna get this.

00:51:29: And you know that's okay.

00:51:31: for a lot things we don't need that discipline or rigor.

00:51:36: there will always be things We do need that disciplinary.

00:51:39: for that You need different mechanisms available too To support that working relationship.

00:51:46: is it changes?

00:51:47: But i think It's the most exciting time in software engineering ever known um and its most creative.

00:51:54: It's fun.

00:51:55: I can work with Claude primarily at the moment, sometimes co-pilot and create things very quickly... Things that are disciplined!

00:52:05: Things they have TDD And all of these other thing i would normally create.

00:52:09: But now im working in new ways That are speedier.

00:52:15: As long as ive that responsibility baked into my workflows The harnesses around those workflow then I get taken with those decisions as well, which is where the fun continues.

00:52:28: Sure

00:52:29: so what is something in the industry that you're really keeping a close eye on to see how it changes over the next six or twelve months?

00:52:40: Everybody i think is looking carefully at LLMs mostly from the perspective of Particularly recently where there's been interesting, I'm going to call these sovereignty decisions around access to frontier models.

00:53:05: That is something that everyone should be closely paying attention too – these LLMs become critical infrastructure for our workflows and so we don't want them may be made in some ignorance.

00:53:22: And so we have to be sensitive.

00:53:24: that and watching closely, I think make a good sense.

00:53:28: The other thing i'm looking very carefully at is again the work of the team i am aware over Canada and the sort of exploration they are doing around cognitive debt an intent debt how when agents writing code They're creating systems quickly.

00:53:50: How can we retain our sovereignty in that relationship?

00:53:55: The journey's not over, right.

00:53:57: In some respects We've recognized the challenges or some of the challenges.

00:54:01: I think there are more and What i try to do is leave you with a launchpad To create your own handbook as a sovereign within this new world because you will navigate this along with me.

00:54:14: As these debts are managed, as our decision-making is augmented and remains important to us that's something I'm paying a lot of attention

00:54:25: too.".

00:54:26: And it speaks my interest as well if i'm being honest.

00:54:30: like I love exploring how the human brain works and how we relate to our tools now our tools change us.

00:54:36: so

00:54:41: Certainly, I've said for a long time that I feel psychology is just as relevant to our work of software engineers and engineering background.

00:54:50: That's probably more true today even than it was a decade ago

00:54:54: but... It IS!

00:54:55: There's this lovely book out there which can push beyond my own which is Rewilding by Tudor Gerber & Son Wardley a fantastic book that again positions the same message in some respects as Alistair's work and my book does.

00:55:13: And then Rewilding Does As Well, which is I think it's a free book as well.

00:55:17: It's available to.

00:55:18: anyone can explore it so i think isn't kind of wiki set up um its fabulous stuff.

00:55:23: what they're doing?

00:55:26: Can you elicit decisions?

00:55:27: Can you explore decisions, can we explore legacy code bases and Decisions being the raw stuff of software engineering.

00:55:35: I think is a key message across The board.

00:55:38: And if you think about it what is more essential to decision-making then how do our brains work?

00:55:45: Sure sure

00:55:46: Would you say.

00:55:46: there are a couple qualitative or quantitative metrics.

00:55:51: You recommend people pay attention too as their kind of doing some this introspection and learning to gauge their work or progress?

00:56:01: I think the Dora Four are still there as sort-of flag bearers.

00:56:10: What is your lead time for evolving towards alongside change failure rate, those sorts things that you've got balance rather than just looking at one?

00:56:23: I think we're still, in the early days as well of figuring out how best to place the ROI on these things.

00:56:33: And i think the ROI is going to be interesting to discern because I think the ROY Is In The Habitat and it's had any effect that habitat has On the capabilities Of the inhabitants.

00:56:45: so at the moment We're Still Looking At Things We Can Measure Rather Than Things That Maybe We'd Like To Be Able To Try And Explore a Measure.

00:56:54: there's that continuing love affair.

00:56:57: That we have in source lines of code, I find that remarkable like...I think when i joined the industry everyone knew that was an insane metric.

00:57:06: but we now do that one again.

00:57:10: look at how much percentage of our code base is contributed to with AI?

00:57:14: Like that's a good thing or a bad thing?

00:57:17: it's a meaningless thing.

00:57:19: perhaps would be another way looking at it!

00:57:21: But its measurable.

00:57:23: And the other one, of course is the measurement of token usage.

00:57:27: So I think there's also the token maxes and you know general insanity as well.

00:57:34: but i also thing that theres a certain degree infatuation in metrics.

00:57:39: we can easily measure where all interesting things are in habitat which is relationships relationships that we have with our tooling.

00:57:53: So when people feel, say they feel like their faster what does that mean?

00:57:59: It's not coming out in raw blow through put this on times.

00:58:04: so where is it happening?

00:58:06: What he was actually carrying there and his dad valuable as well Right.

00:58:09: Because sometimes it can be in someone going, actually I now can operate my environment and feel like i'm contributing but you're not gonna see that immediately or even for some time.

00:58:19: the raw metrics your recording quantitatively.

00:58:23: so we are at a place where most of value is being discerned from qualitative exploration whereas going to be a confused picture for a little while.

00:58:39: I don't think it's gonna have the right fidelity for awhile, but if you ask quantitative questions when you can explore the experience people are having... The lived experience of software engineering is what i like to call that then in that i think he's going to mean much of the value that you see.

00:58:57: and then you need to listen.

00:58:58: when people say look this-this is changing the way i relate and work But it's frustrating because I can't listen to what they say next.

00:59:07: Because in that is a platform engineering move, you know?

00:59:11: It there's something that they find hard-to-do and i've found this a lot in regulated environments as well as non-regulated where people are saying... ...I want to be able do this and I can'.

00:59:21: This would make my life fundamentally easier.

00:59:23: And AI is just an amplification of many of these factors.

00:59:31: And so listen to the noise produced by the amplification and things that you think matter, flow have changed.

00:59:41: those doorometrics will improve.

00:59:45: they'll be lagging but improved because actually worried about something more leading than lagging which is what's a qualitative experience using this thing?

00:59:56: Yeah absolutely.

00:59:58: So what is the number one message or takeaway you want to leave the audience with today?

01:00:05: Well, that's a brilliant question.

01:00:06: The number-one message I'm gonna have to borrow from an author I absolutely adore who was no longer sadly with us but Douglas Adams did so much for us in this industry and beyond.

01:00:24: And uh...I would say don't panic.

01:00:28: one of the things I'd like to borrow.

01:00:30: You're still that decision making party, you should be.

01:00:34: there's still.

01:00:35: this is a decision-making person.

01:00:37: when i mentor people thinking about our industry and everything from schools through universities...I come in with..this is still a great industry to find yourself within.

01:00:52: yes we are relearning our craft, we are adjusting so much at once.

01:00:58: So to tell you exactly what will be doing in five years time is quite difficult but I can't tell you what we might be thinking and i can tell you how value as humans would be in that particular mixture...so don't panic!

01:01:13: Our decision making is still crucial..I don't see anything yet on the horizon.

01:01:18: it's going And that makes it exciting.

01:01:22: We can still be creative, we can still contribute... ...we can still have a lot of fun and enjoyment and sheer joy in this industry!

01:01:32: I encourage people to come into it for those reasons.

01:01:35: So yeah, anybody who's listening and thinking about reading the book or has read the book.

01:01:41: even It is all about joys from remaining sovereign.

01:01:46: Don't panic you're still the sovereign.

01:01:49: I love it Russ, thank you so much for taking the time and sharing your insights.

01:01:55: Greatly appreciate having here.

01:01:58: Thank you very much for inviting me on.

01:01:59: it's been absolute pleasure

01:02:01: Awesome!

01:02:01: Thank You So Much.