Simon says, "let AI do it all"

They have a new junior front-end engineer join the team, and Simon is like all excited and is like, oh, schedules a call. Simon got on his soapbox and was like, you especially are not going to have a job in 2 years.

No, Simon, you're not gonna have a job.

You're listening to Off the VPN, the show where 2 work besties get real about working in tech. Okay, so Jasmine. When you hear someone is obsessed with AI, what do you think that means? Obsessed? Yes.

Do they have a robot girlfriend? They use it for everything, including, um, romantic connection. That's the first thing I think of.

For love. Yeah, that was a great answer. Oh, okay. We'll talk about a person today who is, quote, obsessed with AI. So we'll see, we'll see more about what that means in this particular person's case. But yeah, so we're back on the AI train today for this story. So let's lock in. This story is told from the perspective of an engineering manager who we will call Taylor.

Okay, Taylor. Tay-tay.

Tay-tay works for a startup that's been in the process of scaling up. Taylor describes the company as having a great product. It solves a real customer problem. It's not like groundbreaking or anything, and there's been a lot of recent competition, so I think sales and that type of thing have, has just become much more challenging in this new landscape with a lot of other companies coming to market.

Not the only company on the block.

Nope, not anymore. Taylor has a colleague who is very important to this story, who we will call Simon. And Simon is a very senior engineer with over 30 years of experience. He's only been at this particular company for about 2 years, but overall is just like considered to be super senior.

30 years. You've seen a lot. You've seen it all.

You've seen it all. Really? Yeah. Yeah. No, actually. Like actually. Yeah. According to our EM Taylor, Simon is the type of person who is super senior, has a lot of experience, but also gets distracted by shiny new things. Like you already know what this means, right? You can envision this perfectly.

Well, this is Simon. Simon is the one getting distracted by new things.

That's good.

They're 30 years in and, um, they're still like exciting new stuff, which is great. I feel like you could be in the bucket of folks who are really well seasoned. They kind of like, like the things that they like and they want to do the things and they have a routine and a thing that works. They don't really experiment.

So that's a really good point.

But also, yeah, it can also be very distracting.

That's, that's our perspective. It's a nice thing. Taytay does not think so. But like, aside from all of that, I think Simon is regarded as like a very well-respected engineer, is clearly very talented, has a lot of experience, as we've said, is a reliable coworker, like all really good things until somewhat recently.

Oh yeah.

Simon's like whole vibe has changed.

What's Simon saying?

How long have you been waiting to say that?

I was waiting for the right moment. Yeah.

Yeah. That was a great one. That was a 10 out of 10.

It's all about timing.

It is. All right. So like in Taylor's words, Simon has, quote, gone off the deep end. So at the top of this, at the top of the episode, I said that this was going to be about AI. And people obsessed with AI. So what are you imagining for Simon?

He asked for the company holiday party if he could bring his robot girlfriend. I'm just kidding.

Okay, okay, that was the only right answer.

Yeah, that was, that was it. Uh, I mean, deep end, like, and you're saying this is with AI specifically, like, maybe like they're trying to, like, they're trying to, like, train something, and it's, like, doing, like, this, like, extra stuff where, you know, you're trying to train your own model or build your own thing. And, like, actually, that is, like, very much the wrong thing to do 90% of the time. Uh, it's expensive, and, like, the actual thing that you get is not much. You have to be a really special use case for that, or I don't know, maybe he's trying to go off onto this whole like automation using LLMs, like more of like coding systems. There's some companies who are doing like only prompting you can use to touch this code base and let's see. So maybe they're developing a system like that. Basically premise being like some over-engineered thing that is a little bit not practical. Uh, would be my guess. Wow.

Or the robot girlfriend. Um, or robot girlfriend, wife, husband.

I don't know.

They just want them at the holiday party. Minus that, your answers were so spot on. Really? Kind of. Yeah. Like actually kind of wild. So Simon basically like goes around this company and is like, all of us are going to be completely like out of jobs in 2 years. And like, I don't care because I'm retiring soon. And you know what, why don't I just expedite this retirement? And like, I'm actually going to try actively to like replace all of us at this company by like trying to train models and like do this like super science project that has like nothing to do with anything the company is trying to do as a business. Um, so that's like one side of it. I think eventually Simon gives up on this or realizes it's not gonna happen. Like you need to show up and do your job.

Right.

But Simon ends up doing the second thing that you said, which is essentially just like prompt-only engineering.

Oh, okay.

Let me back up for a second too. Like this company has adopted AI coding tools, like everyone has, and they love them. They've found a lot of success and it's been working for them. But Simon is taking things a little bit too far. So Simon exclusively uses agents to code, will not commit anything to his pull requests himself. Has like a hard rule.

We're paying 30-year experience money. I'm not paying— I think that's what I'm paying for.

We sure are.

There's a big difference between AI-generated code and the one Simon can generate. We're not there at like even remotely.

Yeah. And yet Simon refuses to write a single line of code himself.

Simon says no.

Simon says no. Sometimes he'll take a Jira ticket and just feed it to an agent and have that agent generate a PR, which—

That's cool.

That's cool. But if you're just like jumping from take the ticket, agent writes the code, and then putting up the PR, I feel like we've missed a bunch of steps in between there. Simon's approach to this, or workflow really, is again, review the code that the agent put up, but like refuses to make edits himself. So he will comment on the pull request and tell the agent what to do to go back and make changes. I see. So it's like all very like just prompt-driven, iterative.

So productivity has absolutely plummeted for Simon.

Oh my gosh. Yes, ma'am. Yes, it has.

The amount just to do the smallest change, I'm sure, has like quadrupled. Like, I— you're just spinning wheels all day, especially the comments too, and being like, let me just wait for it to come back. That's crazy.

Yeah, yeah. And like, to have this rule just applied to all different types of changes, even for like the— like, some changes I can imagine it would just take you longer to type it out to tell the agents than it would for you to just go and fix it. Yes. So it's a situation, and you are absolutely spot on, that this is creating problems for the rest of the team and for like just the company in general, because now Simon's work is taking a lot longer. And as we've already stated, Simon is a senior engineer. So Simon's working on like impactful, important stuff. And so suddenly all the important impactful stuff is taking like 3 times longer. People are like, what's going on? Simon though insists it's worth it for the long-term gains.

You gotta talk, Simon. I like the experiment of setting up the Jira ticket system and figuring out like, is this something we can do? I feel like there's maybe workflows that, that actually would be excellent for kind of like getting it to start something, but probably just being your only thread of work is, we're not there yet. So I think there's like use cases where developing these systems and companies are doing this, right? They're more like isolated code bases. They're maybe not production code bases, prototyping, stuff like that. Like something that's like a little sectioned off, like that would be great use. But it feels like he's losing the plot a little bit, like of what this company's doing. Also early, like context of like, they need to be thinking about creating like competitive moats and differentiation. So I imagine like things are heating up, so having any sort of drag like this is just not in the best interest of the company. I think having exploratory stuff timeboxed is a great idea to figure out how you could leverage that, but ultimately it comes down to like, how are we making more money? How are we making more sales? Like it always comes down to that. And so, uh, if you can do that in service of that for the company, then great. But if you're not, then it doesn't make sense.

Yeah. At the end of the day, like someone needs to sign your check and like, I, I also don't know what long-term gains Simon is actually referring to at this point, because I feel like these tools are rapidly evolving all of the time. So it's like you could, I don't know, get to the most efficient workflow today. And then like in 3 days from now, there's going to be a completely new thing that just takes over. And so I don't really buy that, but okay, Simon.

I think Simon's in the wrong place. If he's really excited about this, he should go get a gig on developing these tools and processes. That's like his whole job.

Um, yeah, yeah. I have a feeling though, because Simon wants to retire soon, he's not looking for a new gig. But I agree with you, he's trying to get the system to work for him so that he can collect without doing work.

Yeah.

He has nothing to lose.

Simon got nothing to lose.

Yeah, he doesn't. Yeah. But he, yeah, he's messing up the team.

I've never heard such full AI adoption. I actually think it's really interesting because like really super, super senior engineers are often said to not be adopters of AI and there's like a lot of friction to, so to see this circumstance with the senior engineer like going full all in because they're retiring is pretty funny where I feel like there's usually, it's met with so much skepticism.

Simon is definitely a bit of an anomaly, I would say, in this way. But yeah, the rest of the team is pissed at Simon. One thing in particular too is like, because Simon's not really writing any of this code, there's a difference between when you write something yourself versus if you're just reviewing someone else's code. Like I can review one of your PRs and then like, yeah, it looks good. The next day if you ask me details about it, I probably couldn't tell you, but if I wrote it, I would hopefully be able to tell you all of the details about it. And so what's happening is that Simon is basically merging code that ends up having a lot of bugs in it and then has to go and fix those bugs, but is not familiar with the code that was really written because it was written by AI. And so he's spending a lot of time, just extra time debugging and figuring out, okay, like what the heck is going on? I imagine also using an agent to do it. So it's just like this cycle of like, pushing forward things that you don't really know what they're doing or what they're intending to do.

Layden's like, I don't know what's going on. You know what's going on? I don't know what's going on. You know what's going on? I don't know what's going on. Let me ask Claude.

I know.

Let me ask the other Claude. Let me ask the other version.

I'm literally picturing Simon. He has like Claude. He has like a bunch of GPT models. He has like Gemini. He's like asking all of them like, whoever can debug this. Please help. But yeah, it's like, it's not looking good. So back to Taylor, our EM who is observing all of this. What do you think an appropriate response is for Taylor to like try and address Simon and what he's been doing?

I think the best way to provide the feedback is to describe impact. Of things, it's really important to say like, here's the behavior that I've observed. Here is the impact that I'm seeing. How can we turn this around? Here's what I think. Um, I think there's a lot of value in him using these tools and figuring out how to leverage them for the company. I think like he needs a reminder of this or that energy to be redirected in a way. Probably the best thing to say is like, you could talk about it or even just having actual metrics provided of what the difference is as far as like output and how that has downstream effects. We're seeing more quality issues, like we're seeing that we hired you as this super senior engineer, this is kind of the expectation we have of you, and now that you've completely changed your workflow, and we love that you're testing these tools, there is a significant drop, and so we need to change something about the way that you're working here because this is not sustainable. This is a line also of what is your relationship to Simon as the manager on the team and not their manager. Maybe the appropriate conversation should come from their manager. So as a manager, maybe talk to their manager, but that's always tricky because sometimes it's better to come from yourself because you're the one of the observed behavior, having the observed behavior. But I would start there and have that conversation, but I think you need to have like really clear story and what the impact is. Of these changes. And I think I would say that and say, hey, why don't we, why don't you like chew on that of like, how can we make this like the outcome of this be the best for the team, for the company? And then let's come back and we'll start spitballing about how to like use some of the things you've learned and mix it with existing practices and try to like actually improve productivity because you can't just like fully adopt AI, everything is AI. Like it just doesn't work like that. It just doesn't. Like, you, you just can't let the coding tools and robots run on their own. They don't have the context that humans can hold in their brain. They don't have business context. There's so many problems, and like, we're trying to improve them every day, and we're trying to do all these things, but ultimately we're still very much augmented, and you have to figure out the balance. And he's on the complete opposite end of the balance.

I think you bring up a really good point. About downstream impact and obviously looking at numbers before the way things used to work and after is a really good way of just using objective data to understand what the impact has been. So yeah, I think you mentioned quality being a big one. I think throughput is also something that kind of matters here to some extent. We already talked about how everything's been slowed down, so you can look at things like cycle times, and obviously I don't think we should just be looking at these things like on an individual level or like in a vacuum. It's like, let's talk about the work that matters and sort of what we committed to doing and why we think it's important and sort of how AI coding tools might be affecting that and sort of the goals, like the business level objectives or goals that we've had. Also, I think this is a really good opportunity to talk to the team about some more qualitative things as well. Right. So I mean, they said that everybody is pretty much using AI coding tools and they have seen a lot of success with them, but I don't know, like if we dig into those details, what about the perception of how you think your teammates are using them?

Oh, interesting. Not just like how I'm using it, but mm-hmm. Yeah. Like I'm thinking of like even like a few episodes ago we had the, really terrible AI slop PRs, right? Exactly. There's, there is an impact just as you as a developer, but surrounding you, right? If you're putting up PRs that are not your native language or something that you're, you have expertise in, maybe your skill level is low and you're asking AI to develop these things and you're like, I don't know, probably works. Maybe. I think I, I don't really totally get it. You're putting that effort now on the reviewer of the expert, right? And that has impact. So that's actually really interesting, like asking around. You know what this makes me think too, and I'm actually really glad that you brought this up, is like, I, I'm gonna stand 10 tones down when I say this. I think that there's still, no matter what we do, just because of the market and there's so much money flowing for AI-centric companies that are developing with AI, companies that are developing AI products, the shovels for the AI gold mine rush, whatever, any, anything revolving AI is gonna get you funding, it's gonna get you money. It's gonna get you attention. Because of that, there is just like this, this bubble, there's this like mirage about how these tools actually, like what they actually do. And there's like this consistent tension and we've seen it with our customers. Like for context, Allison and I work on our AI impact platform. So this is something we think about like on a day-to-day. We talk to people, senior engineering leaders all the time where basically like There are boards, there are like C-suite people that are saying like, hey, like we are expecting X percent of code is now AI generated. And like, who really knows how like realistic those numbers are? They're kind of bloated because the market is doing that. It's like survival at this point to be like, AI number up, up, up, up, up, like number up. But the reality is very different, which I think is why it's so important to know that like you adding AI and you're expecting all of these results. That you can't just add AI. It's like not a strategy. I think this is a good, excellent example of like, there is a level of finesse and understanding of how your organization works and its needs and its tech stack, and you need to be able to use these tools well. And these things are evolving every day, as you said earlier, like the actual technology itself is improving, but also people are adapting and finding new ways to introduce this into workflows. There are new, like, glue type products that are coming out. So maybe not the LLMs getting better, but the actual, like, these operational ways of using it, right? So that it makes it a little bit easier for your workflow. So everything's changing, but you expect to, like, add these things and productivity goes up. But actually, in this very beautiful case, AI made everyone's productivity just plummet. Who even knows how much this company, like, lost in terms of, like, deliverables in adopting in this way. There's a balance to strike in augmenting these teams. I know it's a bit of a rant, but like, I think I feel so strongly because I get so annoyed. Everybody in tech industry needs to be more aware and needs to be able to catch these misnomers. They need to be able to actually like lead organizations more effectively by knowing it. If you have like your, you know, people saying like you need to have X more percent faster and your engineers are like, LOL, what? Like you're gonna contort in weird ways. Your, uh, your org is gonna just be less productive. It's gonna do, uh, we've seen it where these giant PRs are getting shipped, but actually it's getting stuck in review and like the overall cycle time is just slower. I think we need to have more realistic expectations and realize like there's a lot more that needs to be done to integrate these and adopt these tools. And that work is not going to end. We're gonna be doing this for a really long time. Yep.

Yep. Case is not closed. Yeah.

End of rant.

End of rant. It's no longer Simon Says, Jasmine Says. But yeah, no, that's all right. And it's funny, like, we are clearly so opinionated about this, and we work on this literally every single day, so we're thinking about it all the time. And just knowing where this story goes It just shows that there's a lot of catching up I think leaders have to do to try and adapt with engineers adopting these tools that I think is probably pretty hard to do when you're not actively using these tools yourself.

They should all use it. I'm gonna say one more, one more sec. Hold on.

Yeah, let's go. Let's go. Let's go.

If you are talking about AI in any capacity, if you are making decisions that involve this, You have to have used them yourself. Yes. Period.

Period.

Point blank. And if you have opinions but haven't actually used it, I can't ride with that because there's too much stuff. You need to have lived experience. I just, I just think there's too many talking heads talking about things they haven't actually tried themselves. I don't mean that to be exclusionary in any way, but I think now is more important than ever because there is so much like garbage out there. Everything's inflated. Truth can only come from experience, I think, at this present moment. So unless you're showing me a thing you did, a thing you did, or you experienced, or you seen with your own eyes, like, I just— I'm going to take it with a grain of salt. Yeah, soapbox away.

That's a really good point. And I'd also argue that it feels easier than ever to carve out time for these things because I mean, think about all of the educational content that is out there for how to use Claude to do this, how to use ChatGPT to do that. Like there's just a lot of actually helpful educational tools to get started. I mean, a lot of them aren't even for coding. Like some of them are just like interacting with AI for everyday people. And so I just think there's no excuse really to not even try and engage. So yeah, I completely agree with you. Taylor, I don't know. Yes.

What's Taylor do?

I don't know if Taylor as an EM is using AI. I have no idea. I can't tell you that. But what I do know is that Taylor does not go to Simon and give direct feedback.

Ah.

Taylor doesn't even go to Simon's manager.

Ah.

Taylor goes to the head of engineering.

Ah.

Which then causes the head of engineering to start to come to the team standups. Yeah, you, this is so, you did not see this coming.

Imagine, imagine like coming into standup and your head of engineering just show, haven't seen them like in a while. Like they're just in your standup. I'd be like, is this a setup?

Yeah. And not only is the head of engineering showing up to team standups, head of engineering is now like, ugh, why is it taking so long?

Oh my God, stop. These people do not know. I don't know. Like, listen, more than half of your job as a leader is to be able to deal with these things. And that is just like a very— that is— wow.

No, this was not it. This was not—

it could have been handled very simply in a conversation.

And guess what? It doesn't work because here's, here's where we go from here.

Well, of course not. No, I know it wasn't because what you're doing, like, someone in here should maybe know the answer and hope that Simon, like, is like, oh, it's me, I'm the problem, I see the error of my ways. Like, no, that's not how it works. You say, hey, you don't do that, period.

Yep.

It's like, it's a very simple— it's simple math. I don't know how you have a job like your, uh, job like that. And do crazy— like, in what universe did that sound like a good idea? I'm very judgy today, clearly.

No, I— this, this story is— I feel like we're allowed.

Because like, also, if you're a leader, a senior leader like that, you should have awareness of your impact in a room. Anything you say is going to be received in a way— like, there is no 'let me kick it with the ICs' like that. That has been removed from you. 2 promotions ago, 3 promotions ago.

Yeah.

You know, you don't, you don't get that experience anymore. So this just lacking a sense of awareness.

Um, yes.

And also you're just tipping the scales on a degree. Also, you are paying these people. This is a manager's job to be able to handle. You as a head of engineering shouldn't be dealing with this. I would be looking to my EMs and being like, why can't you handle this, right? That feels like an expensive waste of money. We have competitors to go deal with. Why are we doing this?

We have other things on the plate. Yeah, I know, I know. That was such a left field. I was like, what? But it doesn't work. And this is what happens to Simon, okay? So recent, like somewhat recently after that happens, they have a new junior front-end engineer join the team. And Simon is like all excited and is like, oh, schedules a call and let me show you how to use AI for coding. The junior dev calls Taylor after meeting with Simon in tears because apparently Simon, Simon got on his soapbox and was like, you especially are not going to have a job in 2 years.

You're not gonna have a job.

But you especially, because you are, not only are you junior, You are a frontend engineer and AI is so good at writing frontend code. You especially will not have a job.

Wow.

Yeah, that is wild.

Listen, if there's anything you could say about Simon is, uh, he says what's on his mind.

Simon's direct.

Have I told you this story where I, um, I was making my first resume before entering like full-time workforce. And I had given it to this person at like my previous internship. Actually, I'm wearing it like my Harvard internship. And, uh, I had them review it and they were a software engineer in their previous life and now they worked, uh, in the school and did things. I don't remember, but I was like, hey, can you look at this? And they're like, are you sure you wanna do this? They'll take you in and they'll work you to the bone and they'll chew you out. Uh, was what they, that was their advice to me becoming a software engineer.

They'll take you in and treat you well. It actually holds true.

No, I'm just kidding. I'm kidding.

Barely a human being standing here. Yeah.

But I remember like having like an episode. Yeah. Adding fear is just not a— in what use was that for them?

This did not help anybody or anything. This was just like pure, like Simon off the rails. And I feel very bad for that junior engineer, but hopefully she's started to actually play with these coding tools and is, oh yeah, no, I'm good. I'm fine. Like, it's okay. But now finally Taylor decides, oh, okay, this is terrible. I'm going to escalate to Simon's manager.

Oh my God. That's not, not uncommon to have a senior engineer make your junior engineer cry. That's—

no, definitely.

This is definitely a new one. I haven't heard this one before. Yeah.

Um, this is new. This is new for sure in the AI era. This is new. So gets escalated, whatever. And then the next thing we know, which is like a whole month later, Simon sends a message to the team and it's just like, I will be taking my talents elsewhere where they will be appreciated. It was a pleasure.

See you all.

And leaves the company.

Wow, that was kind of— that was kind of baller.

I guess he's not retiring, but I'm guessing that feedback was not received. Probably not. I'm taking my talents elsewhere where they will be appreciated.

What a diva.

Yeah, it's a little much. I respect it, but that's what I said at the beginning.

I'm like, it feels like they have an interest in something else. Maybe someone would wanna be like, please build this super autonomous AI machine, compose this for us. And that might be a job that exists completely. That might be like a new DevOps type team where it's just, or DevOps will evolve to like DevOps AI where it's just like developing AI systems in your company. It's just like composing those things like CI/CD, but AI, ID, I don't know. I don't know what it stands for, but we can move on.

Yeah. Don't ask too many questions. No, just keep going. Yeah. Simon.

Yeah, probably.

That's what I imagine Simon going off and doing because yeah, that was not it.

Whenever I leave a job in the future, I'm gonna be like, I'm gonna make sure it's very dramatic like that.

My talents were not appreciated.

I will go somewhere where my talents are appreciated. Thanks for listening to Off the VPN.