Any event, we have some mail. >> Mildly respectful. >> Yes.
Uh, we have mail from Ben. He says, "The fallac The subject line here is the fallacy of the disappearing software engineer. " And he writes, "I keep hearing on various podcasts about the disappearing software engineer and how the barbell effect will come for the software engineer.
This bleak prediction has a fatal flaw. I run a tech company in the valley and work to hire a lot of engineers. I'm here to say that the disappearing software engineer is nowhere close to being a reality and I don't think it will be a reality in the future.
The fundamental difference between software development and every other type of development is that the primary constraint is complexity, not physics. If AI makes engineers 10 times more efficient, a 100 times more efficient, or even a thousand times more efficient, you are still incentivized to hire as many engineers as you can because product road mapaps are infinite. A good technology company will never run out of things to build because that is their lever to make money in the world.
Will the activities of a software engineer change over time? Absolutely. But so long as your competitor is competing with you on software capability, you will be incentivized to build more capability, someone somewhere will have to translate those ideas and capabilities into a working product.
And that will take work, even if it's just typing into a prompt. Long live the software engineer. So Ben, what do you think?
This is obviously a response to our discussion of the barbell effect in engineering last week. my off-handed comment about downward pressure on engineering salaries over time. What comes to mind when you hear all of Ben's uh arguments there?
>> I mean, I I'm sort of intrinsically sympathetic. I like just in general when it comes to these AI debates. I think the number of problems to be solved and more importantly the capacity of humans to create problems to be solved is uh very large.
I mean, we somehow to our benefit, the problem of needing a podcast to listen to has been created and we're working hard to solve it. So, yeah. So, I I I I'm sympathetic to this.
>> Ben does this other Ben does also trigger some of my wanting to sort of be contrarian bits here where he's like, look, as long as your competitors compete with software capability, blah blah blah blah blah. That's like the fallacy of how companies actually win, right? Oh, Slack's going to win because it's better.
Oh, no, Microsoft's actually going to win because they have distribution. And they sort of glue all these things together. And, you know, we see large companies all the time, oh, we're going to increase our stock buyback because implicitly saying we don't have better things to spend the money on.
They could have been hiring more engineers all the time. Now, maybe you might say that actually it's because engineers are scarce, which I think would be might be Ben's counter here. And if they could hire more engineers with that money, they would find more stuff to do.
But I it's not I I think there's it's not not not not as simple as as sort of Ben puts it that there's an infinite number of things to do and to build in the fullness of time. Sure. But it is notable Microsoft did another set of layoffs.
We talked about, you know, Amazon talking about, you know, they're going to be reducing their workforce. Like there is at least some evidence. Again, maybe this is just big sclerotic companies that come up can't come up with better things to do.
>> Well, that's part of what I wonder about on this question because my main concern for engineers is not AI per se, but like at least in big tech, all the companies that have hired engineers over the last 15 years and driven salaries up are still, at least on the outside, it appears they're employing way more people than they need to. And if that's true, then a the efficiency gains from AI may be one reason that software engineers get squeezed in various ways. But also, all these companies have massive ongoing AI infrastructure costs that could lead to belt tightening down the road that sort of depresses the market.
>> That's a very important point, right? Like like the money is not infinite. It gets pulled from from pool A to pool B.
And so yeah, that's definitely a concern. I mean, I would say in the short term it seems like the the like >> definitely we still need engineers like a lot of this stuff is theoretical and demo projects and the actual complexity is huge. And the complexity of software in general means that if you sort of give too much to the AI to do it, you lose a handle of what's being built.
>> You're actually in a worse state than you were sort of at the beginning. It's it's very much a short-term gain for long-term loss. Uh, you know, tech debt is a real thing.
Tech debt as generated by AI sounds terrible, right? I mean, so a lot of this stuff is is sort of to be determined for sure. And, you know, OpenAI is hiring a bunch of engineers.
>> I was going to say anthropic throwing money all over the place. >> That's right. And so, so we'll see.
I I think like I said intrinsically I kind of tend to agree with Ben in this but when you zoom out and you know look at more sort of bluecollar work like how does automation affect say manufacturing >> people who are not concerned about say the US's trade deficit with China will point to the fact that the amount of revenue generated from manufacturing in the US is higher than ever but the number of people employed in manufacturing is lower than ever. And a big part of that is not just outsourcing, it's also automation. Look at agriculture.
We generate far more food than at any time in human history, but the number of people working in agriculture is much lower. It's like 2% of the population. It used to be like 90% of the population.
So will will engineering go in that direction? TBD uh will white collar work broadly go in that direction? It seems likely that that's going to happen and you have an ever more shift to services that are not necessary.
>> Everyone in the future will host a podcast and that's how we're going to solve the employment problems. >> I mean I'm feeling a little stiff sitting in this chair. May I'll go to massage later.
I mean talk about a like a created like job in the last like x number of years. So I I yeah I mean I there is a bit of fighting against like all of technological history. Technology sort of by definition >> increases efficiency that implies fewer people doing a greater amount of work.
>> And so um I'm >> ultimately in the end I've been debating with myself back and forth. I think I'm going to go opposite of Ben here. But I think Ben does make good points and it's very satisfying to have this conversation with Ben where no one knows who I'm talking about and I can explain to you right no matter what.
>> I know I saw his name was Ben as I opened the outline. I was like, "Oh boy, this could get confusing. " Um but yes, the point is well taken and it does always amuse me when Dario Amade for instance comes out and talks about how much of programming is going to be automated away within the next year and a half and then you see job postings at Anthropic, you see job postings at OpenAI.
So, I'm sure it's easier said than done and the timeline might be a bit longer than what some people forecast, but it does seem like structurally if for no other reason than actually spending on AI continues to be unbelievably expensive for all these businesses that have traditionally had much higher margins. And um I don't know where that nets out in the next couple of years. And when people start to say, well, wait a second.
This used to be a much more profitable business than it is now. Like you look at some of the evolution we've seen at Microsoft and what they're having to spend, what Meta is having to spend. Um it it seems reasonable that at some point people are going to try to constrict the spending elsewhere, but I guess time >> maybe it happens in engineering west, right?
like like Meta is hiring a lot of engineers. They're not hiring a lot of marketers. They're not hiring a lot of like all all these other job functions and where they've really trimmed they're really trying to trim sort of middle management in general.
And you know that that makes sense and I I think there's a bit where where Ben and I are looking at this this is still abusing me. We're looking at this in similar ways where I'm pushing back on Ben and saying look there's more to success in business than just product features. There's all the distribution bits that go there sale there like sales is >> Microsoft is such a good example the point you're making.
Yeah. >> Yes. But Ben's kind of saying the same thing within engineering itself.
There's a lot more that goes in engineering than typing out a bunch of boiler plate code. >> And I think anyone that's worked on a mildly complex software project. I just think about like Passport.
There's so much that goes into that.