0:02 Everyone is an architect now. Everyone
0:04 is a staff engineer. You think you're a
0:06 staff engineer or not, doesn't matter.
0:08 You're an architect, you're a staff
0:10 engineer. So the idea is very simple.
0:11 Here's what companies are expecting.
0:14 What was traditionally considered coding
0:15 work is now being done by AI coding
0:16 agents, right? They're going to handle
0:18 the implementation work. Used to be the
0:21 developer's main job. Now the real work,
0:24 the stuff that actually matters is the
0:26 thinking, right? The architecture, the
0:28 context, knowing what to build and where
0:31 to put it. So the conclusion is everyone
0:34 is now working at an architect level or
0:36 a staff engineer level. And the
0:38 conclusion seems kind of optimistic. It
0:42 says that AI is pushing everyone up the
0:44 ladder, right? Junior engineers are
0:46 thinking like seniors and seniors are
0:47 thinking like staff engineers and
0:49 everybody levels up. Sounds great,
0:53 right? But here's the problem. I've been
0:55 in this industry long enough to
0:58 recognize that there is a pattern here,
1:00 right? Every time it shows up, this
1:02 pattern is dressed a little differently,
1:04 but underneath it's always the same
1:07 thing. It's about more expectations and
1:09 the same paycheck. Let me give you some
1:11 history, right? Because this is new.
1:14 About 15 years ago, the industry
1:17 invented a term called the full stack
1:19 engineer. And if if you were around for
1:21 it, you'd remember the pitch, right?
1:22 Instead of just being a front-end
1:25 developer or just a back-end developer,
1:27 you could be both, right? You'd be more
1:29 versatile. you would be more valuable.
1:31 You would understand the whole system.
1:35 And that sounds empowering. That sounds
1:37 like, well, who would not want to be
1:39 more capable? Of course, I want to do
1:42 that. But here's what actually happened.
1:44 Companies used to take what used to be
1:45 two roles, right? A front-end person and
1:47 a backend person. And they basically
1:49 collapsed it into what? One person now
1:52 handles both. So, the scope doubled, but
1:55 the salary more or less stayed the same.
1:57 It kind of went up a little bit and then
1:59 it went back to normal. All right,
2:01 that's that's one story. And it happened
2:04 again with DevOps, right? There was this
2:06 whole you build it, you run it kind of a
2:08 thing, right? That was the philosophy.
2:10 Developers shouldn't just write code and
2:12 throw that over the wall to an
2:15 operations team. You should own your
2:16 code in production. You should
2:18 understand deployment and you should
2:19 understand monitoring and
2:22 infrastructure. Again, at a high level,
2:23 this sounds empowering, right? who
2:25 wouldn't want more ownership and more
2:29 control. But in practice, it basically
2:31 meant that developers were now on call
2:33 for production issues. There was a point
2:36 of time when developers wouldn't do on
2:38 call, right? There were ops people who
2:40 would do on call. Now it's become so
2:42 normalized that everybody kind of
2:44 assumes that yeah, you're a developer,
2:45 you're on call, right? you're going to
2:48 be managing CI/CD pipelines, you're
2:50 going to be writing Terraform config,
2:51 and if you're running Kubernetes, you're
2:52 going to be debugging the Kubernetes
2:56 clusters and all of that on top of your
2:58 regular feature work. So basically, it's
3:00 more scope again, but the same
3:04 compensation. And now we have the AI
3:06 version of this pattern. Everyone is a
3:08 staff engineer. Everyone is an
3:09 architect. You're not just writing code
3:12 anymore. You're expected to think
3:14 architecturally. You're expected to
3:17 manage context across different systems.
3:19 You're expected to kind of steer AI
3:23 agents to produce the right output. You
3:24 have to review everything that it
3:27 generates and you have to make strategic
3:30 decisions about what to build, is it the
3:33 right thing to build and so on. So the
3:37 scope has expanded again massively and I
3:38 think you can guess what's happening to
3:40 the paycheck. So let me tell you what I
3:42 think is happening here. So this is not
3:46 about AI elevating engineers to a higher
3:48 level of work. I mean that's what it
3:50 looks like on the surface and maybe for
3:53 some individuals it's genuinely that.
3:56 And yes AI is going to make people
3:57 productive. I'm not going to debate
4:01 that. But as an industry trend what's
4:03 really happening is something that I'd
4:06 call the expectations trap. Right?
4:10 Here's how it works. AI tools take a
4:13 medium level engineer and it allows him
4:15 or her to produce output that looks like
4:17 a staff level engineer or an architect
4:19 level engineer, right? They can they can
4:22 scaffold systems very quickly. They can
4:23 use AI to maybe even generate
4:26 architectural diagrams. They can produce
4:29 code which looks nice and it can work
4:32 across multiple systems. The output
4:34 looks impressive. So companies look at
4:37 that and go great now we can get staff
4:39 level output from mid-level engineers
4:41 which means they don't have to hire as
4:44 many staff engineers or architects right
4:46 and they don't need to promote mid-level
4:49 people to staff because well they're
4:50 already producing at that level say
4:53 right so why even promote them so the
4:55 point here is if everybody is a staff
4:58 engineer and everybody's an architect
5:00 well then nobody is a staff engineer and
5:02 nobody's an architect so the title
5:06 expands s to cover more people but the
5:08 compensation does not expand with it.
5:10 That's inflation, title inflation,
5:13 right? So this is exactly what happened
5:17 with the senior engineer role, right? So
5:20 101 15 years ago having to say like
5:22 you're a senior engineer meant
5:23 something, right? It meant that you had
5:26 expertise. You were like one step above
5:29 a junior and you had something to denote
5:31 that step, right? you've been through
5:33 maybe a bunch of production incidents
5:35 and you kind of know the system well.
5:37 You could make some architectural
5:40 decisions, right? You don't have to have
5:41 as much and you didn't have as much
5:43 authority as a staff level or an
5:45 architect, but still you did make
5:47 architectural decisions within a certain
5:48 scope and you would live with the
5:50 consequence. You would have ownership of
5:53 a certain thing. But now I think
5:55 companies just hand out senior titles
5:57 after like what two years of experience,
5:59 three years of experience. I have
6:01 noticed people come out of boot camp
6:03 boot camp graduates who had no
6:06 experience before they show up as senior
6:07 developers on LinkedIn within like one
6:11 or two years. Um there is this uh
6:12 interesting comment that I read some
6:16 time back about um how each of these
6:18 titles are going through the same
6:20 inflation cycle right it's title
6:23 inflation just like currency inflation
6:26 is when your money is valued less and
6:28 less and you have to pay more money to
6:30 get the same value similarly with title
6:33 inflation a title has less and less
6:36 value and you have to do more work to
6:38 justify the impact or the expectations
6:40 of that title, right? It's title
6:42 inflation. And that's exactly what this
6:45 is. We are watching in real time the
6:46 architect role or the staff engineer
6:49 role go through that same devaluation
6:53 that the senior engineer went through a
6:55 while ago and I think it is also being
6:58 contributed and accelerated by AI. I'm
7:00 going to link a post in the description
7:02 which is what kind of got me thinking
7:04 about this how everybody is a staff
7:07 engineer and that post makes one point
7:08 which I think is very insightful. It
7:10 says that the engineers who succeed and
7:13 who thrive won't be the ones who are
7:15 best at just prompting AI and getting
7:18 the AI to do things that they want it to
7:21 do. They will be the ones who can best
7:23 manage context. All right, knowing the
7:25 code base or understanding the business
7:28 domain or kind of carrying the full
7:30 picture of what the system does and why,
7:31 right? And I completely agree with that.
7:34 I've told about that in a bunch of live
7:35 streams and I think a couple of videos
7:37 as well. I think context is everything.
7:38 Even before AI, context is everything.
7:40 It's the it's the thing that separates
7:42 like someone who just walked into the
7:44 team to someone who has kind of been
7:46 there and had made decisions and kind of
7:49 knows how things are laid out, right?
7:51 But the problem that nobody's addressing
7:53 is well how does this context come
7:56 about? Where does it come from? It comes
7:58 from actually doing the work. It comes
8:00 from implementing, writing code for many
8:02 years and debugging a bunch of stuff,
8:04 dealing with production issues, right?
8:07 That work is what builds the context
8:09 over months and years, right? You build
8:12 judgment by actually seeing doing
8:13 something and seeing the consequence for
8:15 it and not by just handing it off to an
8:17 AI and getting it to do the work. So in
8:19 that sense there is a bit of a paradox
8:21 here. So on the one hand you know you
8:23 can say writing code was never the hard
8:26 part. Yes AI can write code but that's
8:28 not the hard part. It is the strategic
8:30 thinking and the architectural vision
8:33 that's harder to develop. Right? But by
8:36 dismissing that code is the easy part
8:40 that makes it easy for companies and
8:44 like the CTOs to pay less for it. Right?
8:45 Implementation is where you build the
8:47 context. Implementation is also what
8:50 makes you valuable but implementation is
8:53 also what's devalued because hey AI can
8:56 do the job right you have to implement
8:59 in order to earn the right to make
9:02 architectural decisions because you have
9:04 lived with a consequence of making bad
9:06 architectural decisions. So if we
9:08 automate that away for junior
9:10 developers, if they skip straight to
9:13 staff level thinking without years of
9:16 building things, how do they develop
9:18 that judgment that makes staff level
9:20 thinking valuable in the first place?
9:23 There was this recent uh news item about
9:26 the AWS CEO putting it very bluntly that
9:28 eliminating junior developers is one of
9:30 the dumbest things I've ever heard. And
9:32 I completely agree with that. Dad even
9:34 made a video that eliminating junior
9:36 developers is basically companies
9:38 shooting themselves in the foot. His
9:40 reasoning was that if you don't have
9:42 people learning fundamentals today, you
9:43 don't have anyone who actually
9:46 understands anything 10 years from now,
9:47 right? You've optimized for short-term
9:50 velocity and you've created a longerterm
9:53 knowledge crisis. I actually go beyond
9:55 that. I say even forget longer term.
9:57 Even in the short term, you're having
9:58 problems by not having junior
10:01 developers. But let's set that aside. We
10:04 we know that this is a problem. The
10:06 other thing about everyone being a staff
10:09 engineer is the velocity expectation.
10:12 I'm seeing this in all the companies of
10:14 you know my friends who I've talked to.
10:16 This is changing every day. All right?
10:19 There is a velocity expectation. AI
10:21 doesn't change what you're expected to
10:23 do. It just changes how fast you're
10:26 expected to do it. I've had people go
10:28 like managers say, "Hey, what is the
10:30 estimate for this work?" and then you
10:31 give the estimate and then they go well
10:33 really well just give it to chat GPT or
10:35 just give it to claude it's going to get
10:37 it done in an hour right think about it
10:40 if an AI can generate a working
10:42 implementation in minutes something that
10:44 a developer would take one day or two
10:46 what does it do to your sprint planning
10:49 what does it do to your deadlines the
10:51 answer is they compress if the tool can
10:53 write your code in seconds the
10:55 expectation becomes that you ship in
10:58 hours what you used to ship in days
11:00 right it's not you have more free time.
11:01 It's just that you're expected to do
11:04 more. And the people that are setting
11:06 this expectations like product managers
11:08 or engineering heads and like even
11:11 executives what they're doing is they're
11:12 seeing AI generated demos, right?
11:14 They're seeing the before and after for
11:16 the productivity charts and they're
11:18 saying, "Okay, if the demo could be done
11:21 so much in so quick of a time, why are
11:23 we not seeing productivity improvement
11:26 in our staff?" Right? But here's what
11:28 they're not accounting for. The code
11:30 generation is faster, sure, but the
11:31 thinking hasn't gotten faster, right?
11:33 Understanding requirements and making
11:35 design trade-offs, reviewing the AI
11:37 output that takes so much time, right?
11:39 Looking for a subtle bugs and
11:42 integrating the generated code into
11:45 existing systems. If the AI generates
11:46 one-off thing, you have to think like
11:48 how does it fit into an existing system?
11:51 How do how do you handle edge cases? All
11:54 that still takes human time. And in many
11:56 cases, it actually takes more time now
11:58 because you're reviewing code that you
12:00 didn't write and you might not fully
12:03 understand. So yes, the productivity
12:05 gain is real, but it is smaller than
12:09 perceived because there are a few more
12:10 things that you need to do. So there is
12:12 a gap there and that gap is causing
12:15 burnout already in developers and I
12:16 think it's going to happen more and more
12:18 until the expectations actually get
12:22 aligned. Okay, so now I've told you a
12:23 bunch of stuff. This is actually what's
12:25 happening in the industry now. So what
12:26 do you do with all of this? I don't want
12:28 to just point to a problem and just
12:29 leave it there, right? I want to be
12:32 clear. This is not an anti- AI take.
12:34 Okay? AI coding tools are genuinely
12:36 useful. I use them. I have to say this
12:38 in every video because anytime I say
12:40 something critical about AI, people are
12:41 going to go, "Oh, you don't know. AI is
12:44 actually cool." Dude, I use AI, right? I
12:46 literally use AI every day and I build
12:49 systems with AI. I have a 9 toive job in
12:50 a software company and I build with AI,
12:53 right? The problem isn't the tools. The
12:55 problem is the narrative around the
12:57 tools. And the narratives can be pushed
12:59 back on. Right? If you are in that
13:00 position and you're being asked, like
13:03 you're a mid-level developer and you're
13:05 being asked to do staff level work
13:07 because the AI makes it possible.
13:09 Recognize what's happening. It's not
13:10 that you're being empowered. You're
13:13 being given more scope because there is
13:15 this implicit assumption that you should
13:18 be doing more. And as long as you don't
13:20 meet that assumption that's in someone
13:23 else's head, you will be considered as
13:25 someone who's not doing up to what you
13:27 could be doing. You're not doing up to
13:30 the potential, right? So I think it
13:31 comes down to communication and setting
13:34 expectations. There are a lot of people
13:36 who are at jobs where we don't have the
13:38 liberty to walk out of because the
13:41 market is brutal right now. But I think
13:43 it just comes down to how you
13:45 communicate what is happening. And if
13:47 there are gaps there that your manager
13:49 isn't seeing, it's up to you to show
13:52 them, right? You cannot just absorb this
13:54 extra pressure because it's not
13:55 sustainable. The pressure is going to
13:57 keep going and you're going to have to
14:00 keep delivering. So as much as possible,
14:04 the idea is to communicate. Whether it
14:05 works or not, it's a different issue.
14:07 But you have the opportunity and your
14:09 responsibility to yourself to
14:11 communicate. You can say, "Well, yes, it
14:13 you think it takes an hour, but here's
14:15 what it actually took. It took me an
14:18 hour to do this, and here were the
14:19 problems that I had, and I had to fix
14:22 it, right?" Point it out. Imagine if
14:23 you're working with a teammate and you
14:25 say, "Well, my, you know, this teammate
14:27 delivered code in an hour, and then you
14:29 pick it up, and then it's a mess and you
14:31 had to clean it up." Don't you tell your
14:33 manager that? You say, "Well, yes, you
14:34 thought the teammate fixed it in an
14:37 hour, but it took this long." and you
14:38 make basically make the other person
14:41 accountable and you say that this is
14:42 what you had to do. You're going to have
14:44 to do the same thing with AI, right?
14:45 Consider AI as your teammate and say if
14:47 your manager says, well, AI delivered it
14:49 in an hour, you go, well, no, this is
14:50 what it delivered and these were the
14:53 things that I had to do, right? And also
14:55 be skeptical about industry narratives
14:58 that that conveniently benefits the
15:00 employers, right? Everyone is a staff
15:02 engineer. Everyone's got to think
15:04 architecturally is a catchy headline,
15:08 but it does imply something, right? If
15:09 someone tells you that your role has
15:11 been elevated, you got to think, okay,
15:13 what is happening? Is it just more work
15:16 or is it actually an empowering move? If
15:17 it's an actually empowering move, sure
15:19 you it's it's good for your career, you
15:22 should take it. And finally, you got to
15:23 think about the long game, right? people
15:26 who will be most valued in five or 10
15:28 years aren't the ones who are fastest
15:30 with generating code with AI because AI
15:32 is being commoditized. Everybody's using
15:35 AI. So you're going to be competing with
15:36 bunch of others who are doing the exact
15:39 same thing. The people who shine and
15:42 people who have potential for jobs in
15:44 the future are people who genuinely
15:46 understand systems and who have deep
15:47 context because context is what
15:50 differentiates and you build context
15:52 through years of actually doing the work
15:55 and not just directing AI to do it. Sure
15:57 you can use AI tools to do it but then
15:59 see where it fits. See what is the
16:01 understanding behind what was asked of
16:03 you. Right? Someone gives you a task
16:05 don't just give it to AI. Find out why
16:07 it's being given to you and what is the
16:09 output of the AI and how it fits into
16:12 the picture. Did your AI contribution
16:14 assist AI assisted contribution actually
16:16 address the need? What was the need and
16:18 kind of learn from it? I think that's
16:20 the key and there are no shortcuts to
16:21 that. I'm going to post the link to this
16:23 article in the description. It's it's
16:25 worth a read. Um and and I've gone
16:28 through bunch of these cycles and uh
16:32 it's happened over and over again. full
16:34 stack engineering was reasonable, DevOps
16:38 was reasonable and they both got to like
16:40 we as developers got used to justifying
16:43 the expanding scope without expanding
16:46 compensation. So the real question is
16:49 well sure AI can make you work like a
16:51 staff engineer but whether the industry
16:53 will pay you for it. If the industry
16:55 says well it's you're working like a
16:56 staff engineer not because of yourself
16:58 but because of the tools we give you
17:01 which is AI then you're going to have to
17:02 call out well it's not just that I'm
17:04 doing staff level thinking which is
17:06 different from a junior level thinking
17:08 which is a delta which is expected of
17:10 you which you're not going to be
17:12 compensated for and that's the key right
17:14 so let me know in the comments what you
17:16 think I'm genuinely curious how uh this
17:19 is playing out for you in your company
17:20 are you being asked to do more are you
17:23 being asked to deliver faster and
17:24 saying, "Well, yeah, I can take care of
17:28 it." Or are your u managers and senior
17:31 leadership a little more practical and
17:33 actually testing what's really
