February 13, 2023 • 53 mins
The US labor market looks rock solid. The unemployment rate is at its lowest level in 50 years, while layoffs continue to trend downward. But there's one glaring exception and that's the tech industry. Nearly every major tech company has announced layoffs in the last few months, which is exactly the opposite of how things played out over the last decade, when the sector was a bright spot in an otherwise sluggish job market. So what's going on? Why now? Who is getting cut? And will these tech workers quickly find new jobs? Can they apply their skills to the burgeoning AI space? On this episode of the podcast, we bring back Patrick McKenzie, the author of the Bits About Money newsletter, who previously worked at Stripe for six years. He talks about the current trends in tech employment and why it's still a good idea to become an engineer.
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:10):
Hello, and welcome to another episode of the Odd Lots Podcast.
I'm Joe Wisenthal and I'm Tracy Alloway. Tracy, have I
ever told you like my idea for like a two
part podcast or like a series of two part podcasts?
A series of two parts? Is it the debate one? No,
it's different. Okay, So I don't know about you, But
(00:31):
most of the time, after we do these interviews, I
usually have questions that I started kicked myself for not
having asked oh, yes, yes. So this often happens on
the podcast because we often touch on kind of wide
varying topics and we're not experts in a lot of
the things that we talked about, and often the first
(00:53):
episode is sort of you get to know your subject matter,
and then you leave it with even more questions. Yeah.
So I often kicked myself, like, oh, I should have
I should have obviously asked that obviously. And then the
other thing that happens is you post then the episode
comes out, and then like people on Twitter and elsewhere,
they'll talk about it. They talked about it, they're like,
what I'm curious about is actually like, oh, that's really
(01:13):
good question too. I should have thought of that. So,
like I thought, like a thing that we should do
maybe one day is schedule, like, have all episodes be
two parters where we do an interview with a guest,
take a week, sort of magrinate on it, think about it,
what are some questions we wish we would ask, and
then have the second episode schedule. I really like that idea.
(01:33):
It's sort of like it's almost the octopus model of
podcast episodes, where like one one episode just springs forth
a dozen you arms and legs that you can talk
about forever. So today we're kind of gonna we're kind
of gonna be doing that. We're this is I think
we might be This might be close. I'm not sure
what the record is, but this might be close to
(01:55):
the soonest we've ever had a guest on so soon
after they appeared the show. This is a pilot for
podcast two parters. The only other time I can think
of is we talked to Claudia Sam twice before the
pandemic and was because it was sort of like her
general views on how to like forecast recessions, and then
like two weeks later, like the pandemic was in view,
and it's like, oh, I shoot, this might be really bad,
(02:18):
so he had her on really fast. But this is
going to be close to that record. I think, yeah,
sounds good. All right, let's do it. So we recently
talked to Patrick McKenzie. He is a technology infrastructure financial
infrastructure specialist. He worked for Stripe for six years. He's
currently an advisor. He's the author of the Bits about
Money newsletter, and we talked to him about like corporate
(02:40):
I why it is the way it is? Why does
it seem to be like years and years behind what
we think of the cutting edge of software wise it
off and clunky wise it off and have these big
technical issues that can take a while to fix. That
was a great conversation. The public loved it. But there's
a lot going on in software these days. And the
other sort of big trend that we haven't really talked
(03:02):
about is that for the first time and I don't know,
maybe like fifteen years, we've been seeing all these tech layoffs, right,
And I think there was a little bit of tension
in that episode in that we were talking about why
corporate software is so bad. So it's almost like, well,
obviously there's a need for better software, and yet all
these big tech companies that ostensibly provide these services are
(03:23):
laying people off, but also in the broader macro picture.
Since we recorded that episode, we had a payrolls report
that came out much, much stronger than anyone expected. And
yet we've seen these big tech companies layoff people. And
so the question obviously becomes, is this something specifically about
(03:43):
tech or are these layoffs sort of the first sign
of something broader to come in the economy. And the
other thing that everyone points out to when when you
see these announcements from Meta and Alphabet, at various startups
and Microsoft and Amazon they all done it is like
they edited so many jobs over the last two years
that actually these layoffs are like fairly small in uh,
(04:07):
the grand scheme of things, even for these companies, based
on the amount of hiring that they've done. Yeah, but
I saw I saw a figure from Goldman Sex I
think it was Yannatsis, and he was talking about in
the tech sector, most of the companies that have been
laying people off grew their head count by over since
the pandemic basically because they thought all the pandemic trends
were going to keep going. Maybe there's a little bit
(04:28):
of labor hoarding. But it's a big figure. Absolutely. You know,
people in our industry, journalism, when journalists lose their jobs,
there's always trolls on Twitter saying, oh, learn to code.
You know, that's like a thing. I think at one
point even like Twitter start banning people for saying that.
But I guess the question is, right now, when you
see these layoffs, like should we learn to code or
is that not you know, is that not the career
(04:50):
safety that that it used to be? So all kinds
of questions about what is going on in the market
for tech talent. I'm sure someone's going to tell us
to learn to code. For the record, I can code
just in like very non useful languages. I have no well,
I coded in basic like yeah, yeah exactly, like C
plus plus and like some really basic HTML. But all right,
(05:11):
let's talk to someone who knows more about this question
than we do. We're bringing back Patrick. Thank you so
much for coming back on the podcast. Thanks very much
for having me. Absolutely so. I guess the question is,
before we even start talking about the recently off announcements,
why don't we start with like the hiring boom that
we really saw over the last two years, just massive
(05:32):
amount of headcount added and all these companies, we know
who they are, what drove that. So how about we
rule back history to two thousand nineteen and if you're
looking at recent history, as of two thousand nineteen, tech
has been on sort of an uninterrupted series of a
bunch of very good years, broad based expansion across the
(05:53):
entire industry, basically writing continuing to ride the wave that
had happened since the late pots. Is that how we
was it in English with the consolidation of mobile gains, etcetera, etcetera.
Then the pandemic happened and there was a brief pause
of okay, is this going to be an absolutely catastrophic
(06:15):
event for the entire world economy. Many bad things happened
during the pandemic, the way it played out for tech
was probably not how anyone would have expected one. There
was sort of a one two punch of a combined
fiscal response from governments built in the United States and
worldwide to stave off a huge economic disaster, which had
(06:36):
the effect of both putting money into consumers pockets and
also using the markets for assets, for example tech stocks,
which will come back to the importance so that in
a moment to a lot of the customers were, due
to various non pharmaceutical inventions, sitting at home with very
little to do other than use the internet. And so
(06:57):
a lot of commerce that had been possible on the
internet before the share of it that was soaked up
by the Internet in uh both sort of like semi
discretionary places like food delivery, but also much less discretionary
places like you know, core supermarkets suddenly shifted online in
a very very fast way. And so this combination of
(07:18):
there's more money slashing around and more of it is
falling into the online bucket led to absolutely blockbuster years
for tech companies, and it was a real like trying
to keep your fingers onto the rocket internally at the companies,
like the amount of new users that was on boarding,
at the rate of growth of the business, the raw
volumes of stuff that was going through the pipes made
(07:41):
it like difficult to keep everything up and running. And
a good news front, the business is largely successfully did
keep up and running during a time where society very
much needed them to. They also started to like readjust
their projections of what the future would look like, and
for a while it was looking like. Yeah. The phrase
(08:01):
that was going around was decades of growth were happening
every couple of weeks in terms of, you know, our
anticipated long term shift of the offline economy into the
online economy. And there was a big question of how
long does that continue for and is that pulling forth
growth growth that is happening in the future, is it
a one time spike, etcetera, etcetera. To to various structural
(08:24):
and competitive dynamics, a lot of firms spit simultaneously. This
is a pretty durable change. We find ourselves crushed by
the amount of demand we're seeing right now. We're going
to need to hire and hire aggressively to deal with
this and to position ourselves for what we see as
the you know, eventual coming out of the pandemic future.
And as a result of this, companies were mature companies
(08:47):
the Google's Amazons Facebooks of the world. We're hiring on
the order of like year over year growth across large
portions of their business. Somewhat earlier stage companies, companies that
might look like a stripe even though stripe as uh
somewhat larger these days, or early stage startups were on
boarding multiples of their pre pandemic had count as we
(09:11):
used over the course of the pandemic, so huge expansion
during the during the interval, and then as we came
out of the pandemic, companies assessed a number of things.
One the growth rates tended to go back to his
historical norms rather than this shot in the arm that
the pandemic was offering. Importantly, and you know, tech is
(09:31):
a wide sector. It touches every part of the economy
these days, so it's difficult to say with huge generalizations,
but as top line level things did not decline back
to two thousand nineteen. And again, two thousand nineteen was
not a bad year for tech. It was a, you know,
a pretty good year after a number of pretty good years.
So we haven't gone back to the pre pandemic baseline.
(09:52):
We haven't even stopped growing. In a number of cases,
the growth curve has just spent downwards, and so the
sustained like plus plus headcount growth over time didn't look
like it could be that could be sustained. And then
companies started to look at things that they had allowed
to happen for the course of the pandemic to characterize
(10:12):
these broadly. One of the things that happened during the
pandemic was due to the lockdowns and advisability of having
large numbers of people congregate in small pockets of air,
a bunch of companies went to both remote work and
remote hiring where they might not have had a huge
amount of institutional experience with that model of working before,
(10:34):
and after two to three years of working with these
newer cohorts of people, they've found that there are some
practices that they want to continue from this room at
work world into the future, and that there's some amount
of internal impetus to return to office and have sort
of a cultural reset around the office or headquarters, etcetera
(10:55):
as the sort of beating center of these firms. I've
worked remote for most of my career, myself of a
broad fan of the model. Let's say that there was
some cultural tension and companies on where the locus of
activity is going to be, whether it's going to be
in this online in zoom meetings and slack all the time,
or in the office high band with communication directly with
(11:16):
trusted peers, and a lot of companies wanted to have
a bit of a pullback towards the office, and then
they're looking more granularly at the classes of people to
the heart over the last couple of years, and found
that in comparison to prior classes, there was a bit
of cultural drift relative to where the companies want their
baselines to be, and also in some cases a bit
(11:38):
of a measured productivity difference versus where they wanted their
baselines to be. That's sort of expected because when you're
pulling out all the steps to hire, you like necessarily
you have to be a little less choosy than you
normally are. You have, you know, to the extent that
you describe any value at all to the in person
interview loop, which I describe relatively little value too, but
(11:59):
hopefully like slightly greater than zero. You you lose that
amount of signal and they're sort of hiring in a
slightly more challenge fashion than usual. And so I think
that companies will be pretty quiet about saying, but we'll
we'll say to themselves is we probably have a few
more regrets in like hiring classes than we did in
(12:21):
the hiring classes as a percentage, Patrick, this actually leads
to something that I want to ask you, but what
does blow actually look like in the tech sector, and
you know, is it something that only emerges as business
activity actually slows down, or even in one would you
(12:41):
have characterized tech as bloated. So it's difficult. This tech
sort of subsumes more and more of the economy into
its every increasing embrace to make like huge paint with
raw brush assertions across all of it. But let's see
where to start here. So one, the number of things
(13:04):
that are done in these large companies are extremely varied.
People might have a image that like most people who
could google our engineers, that's actually not the case. Depending
on the company we're talking about between twenty and the
people who work at the company are technologists broadly written,
they are software engineers, their system administrators, their designers at
(13:25):
some companies reporting to the same division. And then the
rest are every sort of worker that you would have
in any company in the economy. Lawyers, regulatory people, customer
sport agents, etcetera, etcetera, etcetera. Management layers upon layers of management.
So what does what does company growth look like? In
one case, it is staffing up more teams to work
(13:47):
on products that already exist. Sometimes staffing teams that sort
of like grow with the the rate of usage of
your products. So like customer service teams typically grow relatively
linearly with the usage of your service. Sometimes it's teams
that grow relatively nearly with the size of your organizations.
So as companies were having these sort of like unprecedented
(14:09):
amounts of employees getting on boarded every year, they needed
larger recruiting divisions to staff up there are other employees,
and it's just based on like the productivity math of
a recruiter. And you can, like finger to the wind
that if you hire a recruiter, that recruiter will be
able to hire twenty five people in the year. And
so if you need to hire four thousand people, then
(14:30):
you know, work math backwards, you require a hundred sixty
recruiters that you didn't have previously. That will tend to
cause your recruiting division to get larger as you are
doing a rapid expansion, and then it will contract faster
than the rest of your company will when you decide
to take your foot off the gas bottle. So those
are the the things that are sort of less inside
(14:51):
of your control. You you just need to keep doing
them to run the business, and then you're making some
more speculative investments on like what are what is our
new product line up going to look? Like? What features
are we going to add? And so the basic unit
of organization within an engineering organization these days is a
single engineering team will typically be like five to eight people,
(15:14):
and that team has a mental bandwidth to deal with
three relatively narrowly scoped problems. And so the more that
you want your software services suite, etcetera. To do, the
more like narrowly scoped problems that come into its domain,
more like five to eight people engineering teams you need,
and so you might find yourself in a position where
(15:34):
you've hired like five to eight people to work on
three relatively narrowly scoped problems somewhat opportunistically, and then when
you you know, come to three and are thinking very
rigorously around like, Okay, we think we're a little bigger
than we were when we were efficient back a couple
of years ago. We think the economic environment not be
(15:57):
might not be as strong and as we were model link,
Which of all the problems in our company are the
ones that we definitely need to keep focusing on, and
which can we refer into later or just our corridor
business right now, then perhaps like some of these nearroly
scoped problems are not at the top of our list.
And then if you consider, you know, like this product
that we thought we would bring to market in three
(16:19):
maybe it will not be brought to mark till then
there might be like ten teams implicated by that that
you do not have propletied for I have a lot
of questions. You know, when Ellen bought Twitter, and he,
like what you know, much more aggressive with the layoffs
than anything else that we've seen. There were all these
like vcs and stuff. A common to the Dirty Secret
(16:40):
and Silicon Valley is that all these companies could do
that they have fifty of their employees not really working
on anything and not really contributing anything. And like, thank
you Ellen for showing that this could be done. And
Twitter still is operating. Although I don't know how the
businesses or whether he cut too deep to the bone
or whatever, but like, would you hear that, like is
that the case that just like over the years, setting
(17:01):
aside the unrealistic expectations of one and maybe two, was
there just a wide scale over hiring relative to the
needs of the business. So tech has been in sort
of a land grab mode for essentially all of my
adult life. We certainly haven't hit the asom to oute
of how many things in the economy can be orchestrated
(17:23):
by software. We certainly haven't hit the assom tote of
how many human and human interactions will be intermediated by
a technical system happening over a smartphone, etcetera, etcetera. In
that sort of land grab mode, you aren't simply like
trying to answer what is the minimal set of things
we can do with the minimal number of people, but
are sort of opportunistically looking at what are the next
(17:44):
ten things that we can try such that one of
them becomes a company defining product feature, etcetera, etcetera. I
have a little bit of risk reflective contrarianism when people
say all tech companies are overstaffed by Could you cut
eighty percent of people who work at tech companies and
still have something functional at the end of the day.
(18:07):
Probably true, that would be extremely painful. But if you
went into a very different mode of operation and just
wanted them to continue the products and services they had
three years ago, possibly that could be done. Probably wouldn't
be optimal for any of them. That's one major reason
why nobody does it. There's also some not gone like cultural, etcetera.
(18:28):
Effects that make it virtually unthinkable. If you were an
executive at at a tech company and you were sufficiently
in your cups and had a had a heart to
heart with someone and said, what's the true number of
Like if I could wave a magic wand and no consequences,
where would our staffing be? Would probably be like eighty
five to ninetent of what it is currently. I think
(18:49):
I think most people would say, like, oh, there's a
bit of there's a bit of like I hate the
word fat in this context about you know, a little
bit of fluff around the edges, but we're not in
systemically a terrible place. And I think you know you
you would get different numbers from different people in different
parts of the organization, but that feels like plus or
(19:10):
minus right to me. Should be noted that I was
a beer worker b rather than the sort of executive
that would be tesked with making that kind of decision. Patrick,
(19:34):
you mentioned the sort of impetus towards creating company defining features,
and this is also something I've always wondered, is there
a bias in tech towards creating new products and our
employees and engineers you know, rewarded for doing new things
rather than maybe maintaining the old ones and perfecting those. Oh,
(19:57):
this is an extremely important thing to understand and the
behavior of the large tech companies from outside of them
that they all have what's called a PERF process in
the industry, it's called PURF outside is a performance review,
and the performance reviews are largely how a company takes
creative work that is done over this time scale of
like quarters and years, and it is often sort of
(20:21):
indevigable and very area and reduces it to a number
such that the company can dole out things of value
like promotions and bonuses and career paths, etcetera, etcetera. And
PERF happens on a semi annual or annual basis, and
the way it PERF works that most large tech companies
is heavily biases in the direction of getting your name
(20:42):
attached to new things that shipped in the world versus
you know, I was assigned to this legacy product, the
product did not go down for six months, you should
definitely give me a bonus on that basis. Oddly enough,
this is not straightforwardly the things that is in the
company's interests because all of the money is made by existing, well,
not all the money. The supermajority of money in the
(21:03):
tech companies is made by satisfying customers you already have
rather than getting new customers, and the supermajority of money
is made on your oldest and tourist products rather than
the new stuff. But institutionally, tech companies biased towards we
want our best people to be on the new things
all of the time. And if your individual best people
(21:25):
want to be, you know, doing the hard yards that
keeps the old stuff running, they will quickly be dissuaded
by their mentors and managers, etcetera, and say don't, no, no,
that is not the way to exceed expectations. You will like,
if you only do great maintenance work for the for
the next couple of years, you will be, you know,
(21:45):
severely career limited here. So figure out something new to
do and make sure your name is attached to it
in a way that is legible to your manager and
your manager's manager and this performance roview process. So let's
talk about the layoffs that we've seen. Because you said
something interesting in your first answer, which is that's sort
of like hiring discipline, hiring quality. During those crazy years
(22:07):
of one part of may have been loose, maybe not
as the standards were a little lower, or maybe people
just didn't fit or something like that. When companies these
days are now or recently making the decisions about who
they're going to let go, how skewed is it towards
that sort of recent cohort. Because the other thing I
(22:29):
could see is that look at many companies, you probably
have people who have been there forever who are getting
paid extremely high salaries or very good salaries just based
on the fact that they got some bump every single year.
Maybe they're not pulling their weight to some perceived degree
as much as they used to be. So how much
of the you know, when when are they when the
(22:50):
executive look and say, okay, we're gonna make cuts. How
much was it skewed towards the new cohort versus seeing
as like this is an opportunity to get rid of
some highly paid employees maybe don't add as much value
as they want. So a disclaimer off the top, layoffs
are like understandably traumatic for the people who go through them.
I don't want to minimize that. At the same time,
(23:11):
I think we often, particularly as as workers in this industry,
sort of like advocate responsibility for understanding the like structures
that cause these things to happen in ways that are
not in our interests. So broadly it's it's good to
have like open conversations about how these sort of decisions
are made. I think it is different on a firm
differing basis, but broadly speaking, you would not want your
(23:33):
simply to like roll back for the last six months
of your hiring. There's a couple of different reasons for that.
One is that when you're dealing with these complex ecosystems
that sufficiently large companies ecosystem to itself, there's all sorts
of levers that you are like managing a parallel and
one of those levers is that you are attempting to
(23:54):
balance the seniority ranges in various parts of your organization
such that you always have a mixed within some error
bars of how many people that you have they're acclimating
to the company versus how many who have acclimated and
could do productive work, versus how many are in that
senior mode where they can lately parachuting to consult on
things and do the architecture stuff that you're more intermediate
(24:17):
employees might not be able to do. Yet, if you
sort of create a bubble in the pipeline by concentrating
your cuts in the people that were only hired in
the last six months to two years, then you are
setting yourself up for a bubble a couple of years
from now where you have far too few people at
a portion of the experience curve to do work that
(24:39):
you urgently need to do work on a week by week,
quarter by quarter basis. And so if you come to
the conclusion that we've hired a few too many people
over the last couple of the last couple of months,
what are we going to do about that? You have
to distribute your cuts over a larger number of cohorts
than the most recent cohorts, or you will set yourself
(25:00):
fun for some pain. There's also some compliance and legal
issues that come up with is employees and you already
a protected class in particular jurisdictions, which also plays into
it into a little bit. But the biggest reason is
to avoid causing the operational issues for your company layoffs.
Is performance management that is a thing that exists in
the world. And so you know, if you were hearing
(25:22):
skeptical FECs on Twitter, that they would say about large
software companies is not merely that they were a little
bit flabby, but that they were a little bit uh
self assured of their position in the world, and it
had too many good years in a row. And if
you got attached to them, you could you get a
job in a corner office and not do all that
much and still be fine. I think that is a
(25:44):
little exaggerated, but let's say there's certainly cases of it,
and there's certainly some people like mature into a career
where they continue being impactful over years and decades, and
some people end up in sort of a tenured professor
mode where they've become critical to the organization because they
know a couple of things that the organization needs to know,
but they don't bring the same intensity that they used
(26:06):
to in their career. And then there are some people
who have like successfully created a niche for themselves inside
the company, but the company might not desire to exist,
and nobody wakes up in the morning and says today,
I want to do layoffs. But given a circumstance where
everyone in the industry is doing layoffs, some executives might say, okay,
(26:28):
it is a good time to reevaluate and like turn
up the heat a little bit on our performance management
and say, okay, is there anyone who has been coasting
a little too long? Is there anyone who has uh,
you know, created a secure little nest for themselves in
a way that that nest does not add a lot
of value to the company. Given that we we need
to usher some people on two new positions, let's start
(26:50):
with that first and then move to the cuts that
are going to take more mental energy to do. You know,
we're talking broadly about hiring discipline and the idea of bloat.
And this is a slightly loaded question, but to what degree,
if any, do you think the sort of maybe monopolistic
mode that some big tech companies have built around their
(27:13):
businesses has contributed to some complacency on the hiring front.
And a little adverse to the word monopolistic, but I
think I get what you're getting at, and that there
is certainly a lot of rent created in the technology
industry where these are some of the most effective businesses
(27:33):
ever created in any industry. Google AdWords will print a
ginormous amount of money next year, and almost no amount
of action taken by any set of first actors internal
or external ad to Google will cause Google AdWords to
not be worth many, many, many billions of dollars, and
so the margins on it are very high as well
(27:57):
in comparison to you know, we we're talking last time
about the airline industry, where the airline industry has struggled
mightily to maintain like single digit percentage positive margins over
a multi decade timeframe. Tech doesn't have that problem. The
nature of these very sticky products that shearsket size of them,
and the margins do tend to create a little more
(28:19):
room for that flabbiness than in exists in many industries
that have more of a cutthroat reputation. This is sort
of the polar opposite question. But nowadays we hear a
lot about the possibility of companies hoarding labor when it
comes to tech. How much of that do you think
(28:40):
has actually gone on in the sense that do you
see tech companies opportunistically hiring people just so their competitors
can't get their hands on them. I've heard this theory
advanced many times, and honestly, I don't think it is
very explanatory, and sometimes it's phrased Google would rather hire
a particular talented engineers so that they don't create a
(29:02):
startup and then eventually become competition to one of Google's products.
If hypothetically that were something that actually motivated executives at
tech companies, there would be a number of things that
would be easier to do than quote unquote labor hoarding
that we don't do institutionally. So in finance, there's this
institution of gardening leave. Tech doesn't institutionalize gardening leave at
(29:25):
any level almost anywhere in the industry. And if you
were thinking about let's prevent highly talented people from doing
interesting things for our competitors or for new startups that
they could create. The people in the industry that you
have the like tightest speed on their productivity level are
your existing employees, and so you would be you would think, oh, well,
like the natural place to start is like start with
(29:47):
people who are already work here and say, if you leave,
we would like to buy twelve months of your time
sight unseen, and no one does that. And there's other
things that you can do Broadly tech is there's always
a bit of push and pull between the needs of
a company and the needs of employees, but broadly tech
who is strikingly pro worker relative to many industries in
the United States. These things that are done that would
(30:08):
be consistent with the labor hoarding hypothesis just are not done.
You know, you can talk to the people that are
involved in the decisions that that are read on the
outside as labor hoarding, and they never advanced that as
a reason to you know, buy up. A new company
that has four engineers attached to it is typically phrased
something more similar to, well, this is a team that
(30:29):
seems already jelled. They're clearly highly highly productive individual contributors,
and we could have a bunch of engineering recruiters work
for months to find for similarly talented individuals or the
m and a team can like tick one box off
in Q one and get them all in the door
for the price of one low check. Let's do it.
The notion of like and let's take this team off
the table, so they don't, you know, have a market
(30:51):
success in three years and create something competitive with us
never comes up. Okay, we started talking about why the
(31:14):
hiring boom happened. In the first place, we've talked about
maybe some of the decisions on who is getting cut.
Let's talk about the sort of prospects for the people
that have lost their jobs and or the people that
are thinking about going into a career in tech. So
how quickly do you perceive that the people losing their
jobs over the last several months are finding new offers? Like,
(31:35):
let's start really simple. Can I tack something onto that,
which is how how fungible are these types of jobs
in reality? Yeah, A long time ago, in a place
far far away, during the dot com crash, I was
graduating from university and the Wall Street Journal was which
read the Wall Street Journal every day with my father
(31:55):
growing up. It was how I learned to read the
Wall Street Journal. Could do no no wrong in my
as as a an undergraded engineer, and the Wall Street
Journal was pretty decided that yep, engineering as a field
is done in the United States of America. Henceforth, all
engineering will happen in Asia. And I said, oh, chucks,
I really wanted to get an engineering job. I guess
(32:16):
I have to move to Asia, and so I did.
Oh back, now we know the origin story. This is
the backstory of how I ended up spending my entire
adult life in Japan. Now that ended up being a
good read like a good life decision for me for
entirely unrelated reasons. But it turns out there were, in
fact engineers hired between two thousand and four and two
(32:37):
thousand and twenty three in the United States, and so
reports of the field's demise were heavily exaggerated. If you
are considering a career in engineering, every reason you had
to consider a career in engineering in is like still
a reason to do it. So this like minor wobble
that will be forgotten in a matter of months. Please
(32:59):
don't allow it to like cause you to make major
drastic life decisions. Although life is what happens when when
you're busy dealing with these little wabbles. Okay, so that
out of the way. How fungible are people? Broadly speaking?
In the early levels of career, tech tends to cast
a very wide net and hire people for what's often
(33:20):
called horsepower, with the expectation that they will be able
to specialize over time. There is some degree of worry
that if you spend ten years or fifteen years in
a particular industry doing. The quote often used is have
the same year ten years in a row, then you
will end up over specialized and only be available for
doing that sort of thing in the future. Depending on
(33:42):
the thing you are doing, there might be a sharply
limited set firms for which that is relevant. But broadly speaking,
the engineers that were hired to do anything in the
first five seven is years of their career are broadly
expected to be able to do not quite anything, but
like a large subset of all the things that a
(34:03):
tech employer could want an engineer to do. And so
the liquidity in the tech market within like a broad
class like recruiters or engineers, etcetera liquidity between job titles,
exact roles, exact companies business model of the company is
very high. And I'm forgetting what Joe's original question was, Well,
so are they finding just a short term like you
(34:26):
must hear from people, you must talk to people like, uh,
people just got cut off? Are they recruiters already reaching
out to them from different companies? Structurally tech company this
would be a bad pool quote. Structurally tech companies are
like sharks. Okay, we're gonna we're gonna pull their quote.
Just just like sharks, like the way that their gills work.
(34:48):
They have to keep swimming or they stop getting oxygen.
And that's an unfortunate thing for most creatures. The tech
companies because of their staffing models, they and that thing
we talked about earlier, where they are constantly mixing the
number of people at each level of seniority within the company.
They have to keep hiring. And so even if an
individual company decides like, Okay, we're going to like push
(35:09):
pause for six weeks and do quote unquote hiring freeze
one the amount of time and they can actually do
that and not severely damage the business is limited. So
it pauses always temporary unless the company is going down
the tubes. And like the large tech companies certainly are
not going down the tubes. Some startups might get shaken
out at the march and tow to UH funding constraints,
(35:30):
et cetera, But the like the overall business of the Internet,
continues to grow apace. So pauses are temporary nature. And
there exists, you know, like many different companies inside that
the broader ambit of tech. Some of them might be
positive any given moment, Some of them are you know,
still attempting to make new investments for three and some
(35:53):
of them while they're not in uh sort of rapid
growth mode, growth mode for doing things like you know,
we have to back fill for people who are leaving
the company, and in a typical year at a typical
tech company, that may be like ten percent of our
engineering staff. So if we've got two thousand engineers, we
have two hundred engineers that we are slate to hire
in Interestingly, one of the things that cost a bit
(36:15):
of the over hiring was companies have this model for
what percentage of people will leave in a year and
therefore how many you need to hire just to stay
at the current level of employment that you have. And
when the economy started wobbling in, what happened was the
rates of voluntary attrition that companies, meaning that people who
resigned out of their own volution, went lower than the
(36:36):
model predicted. And because you need to like set in
place a process that takes months to hire people, but
the process of deciding not to quit is not visible
for those months, that resulted in sort of like a
hiring overhang, and so companies overshot their targets for how
many people would be in the company, which doesn't sound
(36:57):
like an easy thing to do, but it is a
very easy thing to do if there is a sort
of like sharp change and employee behavior with regards to
things that they have total control over and don't have
to announce to you, like deciding to leave or not
leave in a statistical fashion. Are there signs that tech
workers should look out for that they're about to be
laid off? Like do you stop being a signed new projects?
(37:19):
Do your access codes get cut off? Does someone come
take your stapler off your desk? Like? What exactly are
the warning signs that you might be in the danger zone? Many,
many tech people have a large degree of stress with
regards to whether I'm doing well, am I on the list, etcetera, etcetera,
And I don't want to add to that stress. Broadly.
(37:41):
You should. You should have an understanding of how performance
is calculated at your company, and consider that official view
of your performance to be perhaps more important than you
would naively believe it too, because the official view where
you're the entirety of your performance is reduced down to
like one number, I'm a full or for this six months.
(38:01):
That is the only view that is going to be
available to someone who might be two, three, four steps
above you on the ladder when they're going to make
hard decisions in a hypothetical future where they're making hard decisions.
So the things that cause formal visibility to accompany are
anomalously important, and the career oriented people around you, who
(38:22):
are very good at work in those systems their advantages
will find advantages based on that. But I wouldn't, you know,
over rotate on perfect The only thing we're thinking about
seems simple, But just do great work and then make
sure people are aware of the fact that you did
the great work, and then things will tend to work
out in a career fashion over only a long period
(38:43):
of time, not gun. So I have a question that
bridges this conversation with the conversation we had last week
about I T. And I realized I should have asked
it last time. This whole episode has been because it's
actually I actually only just had one question from last time.
I had to come up with a whole excuse for
why we needed to have you back out just so
(39:03):
I could ask this. But occurred to me, you know,
like in the in the business press, we're always reporting
on c E O s getting fired or let go
and hired. Sometimes CFOs I don't see much coverage of,
like c t O S or c I O is
like the people who run the internal tech systems being
let go for poor performance. I actually think the only
time I can ever remember hearing any sort of CTO
(39:26):
or something losing their job for poor performance is probably
like fifteen years ago when Twitter was always having the
fail wills and like they weren't scaling very well during
the boom years. And other than that, I can't actually
recall a time in which I like read a story
about it, you know, a CTO being laid off for
bad performance. How often does that happen? And you know,
(39:47):
in the context of whether we're talking about tech companies
or all. You know, I think we were talking about
Southwest and others last time, Like how often do the
head of do those positions lose their job because they say, like,
our I is not good. So it's a complicated subject
for a number of reasons. One is that the degree
of saliency of CTO most companies to the media is
(40:08):
relatively low. The degree of saliency of many things that
are very important in the tech industry to the media
is lower than many people in the tech industry, but
like and that is one cause of the frequent conflict
between the media and tech. But be that doesn't two
people get laid off for for performance. Yes, one relatively
frequent thing that happens relative to the incidents of senior
(40:29):
senior executives departing is the uh sort of like fall
on your sword motion if there is a significant outage.
Is a thing that frequently happens, are frequently relative to
all causes for a departure and hesitant to give you
the example because Tokyo is a small town, but there
are a number of banks, both in Japan and outside
(40:50):
of Japan that have had disabling computer outages for like
days to weeks at a time, where that is an
extremely extremely thing to be avoided for a bank and
rules up fairly directly to the head of I T
or the CEO. And there are cases where either the
head of I T or the CEO of have left
us a result of doing that. There is one thing
that I do like about the culture that is Japanese management,
(41:13):
where in the sort of like ritualized speech that an
executive gives it that they will often say police don't
blame the people that had their hands on the keyboards
during this. The fact that this was allowed to happen
was a result of managements miss decisions or taken over
the course of years. I presided over them, and as
a result this uh, this allage. Even if you know
(41:34):
it was one person individually fat fingering something that took
us down for a week, this belongs at my door,
and I'm resigning to take responsibility for it. There are
many things I don't love about Japanese management culture, but
that bit I do like. Another thing is there are
reasons for companies to be other than other than maximally
(41:55):
public about the fact that we are removing a senior
executive for cost. If you remember the over the course
of the last couple of years, the I T sector
has been in sort of like massive boom mode. Companies
are extremely protective of their brand with respect to engineering candidates.
Nobody wants to join a organization that exists under a
cloud who's CTO just got fired for being an idiot.
(42:17):
So the thing that might happen is like, oh, well,
the previous VP of engineering wasn't quite up to enough.
Maybe they can be shuffled onto a different project, and
we're going to hire a CTO above them. If you've
already hired a CTO, that's a bit of a bit
of a more difficult thing. But like shuffles with regards
to who are the most important people in the engineering
(42:39):
organization and is there a separate product organization? Do they
report to the same people, etcetera, etcetera, are sometimes caused
by like X isn't getting it done. We want to
like shuffle in why, But we don't want that to
be seen as a repudiation of X. Not because we
care about X's opinions so much, but we care about
how this will be read by internal engineers who we
want to keep attaptioned to the company and external candidas. Alright,
(43:03):
one last small question they'll probably eject, a question that
we can talk about for a long time, but just
real quickly. So the one one area within tech that
seems like almost certainly going to be hiring like crazy
for years at this point is anything to do with AI.
And you know we all know what's going on there.
How much are the skills that some of these like
sort of cutting edge AI companies in need of. How
(43:28):
much are these skills that sort of legacy or existing
tech workers might have, or how much are the skills
that they need something that like you really need years
of like focus training in the specific area to satisfy
what these companies need. Can I can I add another
thing onto the back of that place? How many coding
jobs will something like chat GPT destroy? Ye? Should people
(43:50):
stop learning to code? Yeah? Yeah, talk about talk about
So I have a glib but true answer with respect
to our advanced AI techniques going to destroy programming jobs.
The first program or class of programs that we had
where an advanced computer was obviating the need for human
programmers was called the compiler, where instead of doing you know,
(44:12):
complex low level and instructions directly and assembly and speaking
sort of natively the language of the computer, you use
what we're called high level programming language is like C
back in the day. And then the compiler would you know,
use its magic AI powers to turn that C into
assembly language so that you didn't have to laboriously do
the assembly language itself. So every technology that gives programmers
(44:34):
more power, more capability to do things that are valuable
for human society probably increases the aggregate demand for programmers
is sort of like my high level view on the
world and if yet to see a contrary example to that.
So an interesting question with regards to AI is what
are the like, what series of steps is going to
(44:56):
be necessary to take it to market in a way
that it actually creates the value for individual people land
for society, and that it seems to have Latin within it.
And if you look at like chat GPT, if you
like I view it as an iceberg, there's the above
the waterline part and below the waterline part, and below
the waterline part has some let's say deep deep magic
(45:20):
there bracketting out that magic for the moment, it seems
like the above the waterline part was very important in
why everyone has heard chat GPT and probably used it
if you're listening to this podcast, but it hasn't heard
of like similar efforts at Google, etcetera. The reason is
that there was a you know, a product focused team
that made a relatively pedestrian piece of software like a
(45:43):
chat interface, but made it really, really good and like
work on that to the point where people's interactions with
the underlying large language model would be like sufficiently effervescent
that it would screenshot that interaction has shared it over
to Twitter, and so everything that above the water line
part is amenable to the to the technologies and tactics
(46:07):
that existing engineers have with no modification whatsoever. There you're
talking to the back end. The back end is implemented
in a different kind of magic than your back ends
usually are. But the back end has always been magic too.
That is like part of the answer. There's an interesting question,
like how much of the work is going to be
that above the waterline part. The productization of these you know,
(46:29):
creating like new forms of user interfaces, new models for
interactions with users, new metaphors that we have to teach
to people, like new you know there there might be
an entire field and like education and how to do
I don't know, prompt engineering, well, prompt engineering being how
do you type in the right series of incantations to
the machine so that the the spirited some and stuff
(46:51):
out of the ether does the right thing for you? So, like,
what percentage of the work will happen there? First? What
percentage of the work will happen on these like core
under the hood model things. A sub sub question to
that too is like okay, so for the work happening
at that model layer. Is that work going to happen
at every company that consumes models, or is it going
(47:13):
to happen primarily at open AI and Google and Microsoft.
And we can count the number of firms that like
need this these engineers on a single hand. In a
world where we count the number of firms that need
like dedicated hard AI researchers on a single hand, that
probably implies like lower total employment of them than in
the world where every firm that touches AI has its
(47:36):
own AI practice on staff. But it's at least as
of like the current state play deeply uncertain where that
will shake out. And so these are some of the
questions that get debated upon people at both the AI
firms and also like you know, if if you are
a VC that's adventure that's investing in the space, you
are probably having like a number of interesting dinner conversations
(47:56):
on okay, where does the value accrue in this chain?
Where does most of the work get done? What do
these products like expose themselves to in the life of
the user. Is it's something deeply under the hood or
is it integrated into their daily operations? Do they know
they're using an AI do they know they're using software?
Is it something that they're like directly typing in or
is it something that they're interfacing with someone who's doing
(48:19):
the typing on their path, etcetera, etcetera. Well, Patrick, this
was absolutely great talking to you. We could talk for
a long time, but instead we'll just talk to you
in a few weeks again when we have a million
more questions. Now, I'm big patigias, but I learned a
lot and really appreciate you coming back on the show.
Thanks very much for having me, and I always happy
to be come back. Thanks Patrick. That was fun, so Tracy,
(48:52):
there were so many interesting elements of that conversation. I'm
really glad we had Patrick back. I'm not even sure
where to begin, but to start, you know, his point
about the hiring boom during the pandemic, I thought was interesting,
not just that maybe these companies had a sort of
unrealistic expectations about how long does growth boo would last,
but that when you're hiring that fast and under sort
(49:12):
of extremely unusual situations, like you have that drift where
maybe you're like there's a little bit of a we're
not that happy with the class. And then also that point,
but you also can't just hire everyone who came in
recently for reasons of like seasoning and like skill level growth. Well,
to me, it kind of I guess hammers home the
point that three years on from the start of the
(49:36):
COVID pandemic, we are still experiencing these normal developments. And
it kind of gets to the macro versus micro point
about some of the recent payrolls figures. You know, all
the tech layoffs that have been announced. Are they saying
something about the wider economy or is this really a
tech specific problem? And I think, I mean, I can
(49:58):
kind of argue it either way. I think I come
away from that conversation thinking, well, you know, one were
really unusual periods in terms of hiring for the big
tech companies, and to some extent it seems reasonable that
that starts to get rolled back a little bit. But uh,
you know, it's also I take his point, is he
and I suspect it's probably true, which is that if
(50:21):
a year ago you were thinking you wanted to go
into engineering or coding or something like that, very little
about what we've seen so far in three should make
you change your mind. I thought that was really interesting
too about like sort of the questions about AI and
how so it's like, as he pointed out, like there
are other you know, places working on very similar, if
(50:42):
not equal technology. What sort of made things breakthrough recently
was the consumerization of some of the chat interface or
some of these AI images imaging things. So like how
much of like to go to market for this stuff?
Ultimately isn't sort of like familiar experiences that people already have.
It reminds me a lot and I don't mean this
(51:03):
necessarily in a negative way, but it reminds me a
lot of crypto in the sense that, like, yes, there
is a lot of hype around AI, but also in
the sense that this is a new technology that people
can actually participate in. And so the use of the
AI image generators, chat GPT, it kind of brings it
to people in the same way that they are able
(51:24):
to experiment with, you know, blockchain and different types of
money using crypto, and so it suddenly becomes a lot
more salient for people in that way. Yeah, you know,
like there's a good example because like with crypto, like
if you're like interacting with like core protocols are like
developing on ethereum like that's going to be a limited
a limited number of people know how to do that.
(51:44):
But if you're building like an exchange, there are a
lot of I mean, everyone can have a wallet, right
you're marketing or stuff like that. There are still all
of these roles within crypto that have like sort of
like consumer facing analogs to any other industry. Yeah, I
need to look up the compile tiler. That sounds interesting. Yeah,
I'm gonna go off in Google um deep learning compiler.
(52:05):
I guess for so far a job security, we still
need to learn to code. Huh. I think we need
to learn AI. I don't know. Probably I don't know,
but you know C plus plus, which is what I
learned in a little bit of no because it's obsolete.
Like no, I don't think anyone uses c post plus
and they certainly don't use it for for AI and
(52:27):
machine learning stuff. I should have asked Patrick what language?
What coding language Python? When we have them back in
the Yeah, okay, yeah, our next episode with Patrick will
be about which coding language we should all be learning
in the future. Shall we leave it there? Let's leave
it there. This has been another episode of the All
Thoughts podcast. I'm Tracy Alloway. You can follow me on
Twitter at Tracy Alloway. And I'm Joe wisn't Thal. You
(52:49):
can follow me on Twitter at the Stalwart. Follow our
guest Patrick McKenzie. He's at Patio eleven and check out
his Bits about Money newsletter. Follow our producers Carmen Rodriguez
at Carmen Armand and Dash Bennett at Dashbot. And check
out all of our podcasts Bloomberg under the handle at podcasts,
and for more odd Lots content, go to Bloomberg dot
(53:12):
com slash odd Lots when we push the transcripts Tracy
and I blog. If we have a weekly newsletter that
comes out every Friday, go there and sign up. Thanks
for listening
Hello, and welcome to another episode of the Odd Lots Podcast.
I'm Joe Wisenthal and I'm Tracy Alloway. Tracy, have I
ever told you like my idea for like a two
part podcast or like a series of two part podcasts?
A series of two parts? Is it the debate one? No,
it's different. Okay, So I don't know about you, But
(00:31):
most of the time, after we do these interviews, I
usually have questions that I started kicked myself for not
having asked oh, yes, yes. So this often happens on
the podcast because we often touch on kind of wide
varying topics and we're not experts in a lot of
the things that we talked about, and often the first
(00:53):
episode is sort of you get to know your subject matter,
and then you leave it with even more questions. Yeah.
So I often kicked myself, like, oh, I should have
I should have obviously asked that obviously. And then the
other thing that happens is you post then the episode
comes out, and then like people on Twitter and elsewhere,
they'll talk about it. They talked about it, they're like,
what I'm curious about is actually like, oh, that's really
(01:13):
good question too. I should have thought of that. So,
like I thought, like a thing that we should do
maybe one day is schedule, like, have all episodes be
two parters where we do an interview with a guest,
take a week, sort of magrinate on it, think about it,
what are some questions we wish we would ask, and
then have the second episode schedule. I really like that idea.
(01:33):
It's sort of like it's almost the octopus model of
podcast episodes, where like one one episode just springs forth
a dozen you arms and legs that you can talk
about forever. So today we're kind of gonna we're kind
of gonna be doing that. We're this is I think
we might be This might be close. I'm not sure
what the record is, but this might be close to
(01:55):
the soonest we've ever had a guest on so soon
after they appeared the show. This is a pilot for
podcast two parters. The only other time I can think
of is we talked to Claudia Sam twice before the
pandemic and was because it was sort of like her
general views on how to like forecast recessions, and then
like two weeks later, like the pandemic was in view,
and it's like, oh, I shoot, this might be really bad,
(02:18):
so he had her on really fast. But this is
going to be close to that record. I think, yeah,
sounds good. All right, let's do it. So we recently
talked to Patrick McKenzie. He is a technology infrastructure financial
infrastructure specialist. He worked for Stripe for six years. He's
currently an advisor. He's the author of the Bits about
Money newsletter, and we talked to him about like corporate
(02:40):
I why it is the way it is? Why does
it seem to be like years and years behind what
we think of the cutting edge of software wise it
off and clunky wise it off and have these big
technical issues that can take a while to fix. That
was a great conversation. The public loved it. But there's
a lot going on in software these days. And the
other sort of big trend that we haven't really talked
(03:02):
about is that for the first time and I don't know,
maybe like fifteen years, we've been seeing all these tech layoffs, right,
And I think there was a little bit of tension
in that episode in that we were talking about why
corporate software is so bad. So it's almost like, well,
obviously there's a need for better software, and yet all
these big tech companies that ostensibly provide these services are
(03:23):
laying people off, but also in the broader macro picture.
Since we recorded that episode, we had a payrolls report
that came out much, much stronger than anyone expected. And
yet we've seen these big tech companies layoff people. And
so the question obviously becomes, is this something specifically about
(03:43):
tech or are these layoffs sort of the first sign
of something broader to come in the economy. And the
other thing that everyone points out to when when you
see these announcements from Meta and Alphabet, at various startups
and Microsoft and Amazon they all done it is like
they edited so many jobs over the last two years
that actually these layoffs are like fairly small in uh,
(04:07):
the grand scheme of things, even for these companies, based
on the amount of hiring that they've done. Yeah, but
I saw I saw a figure from Goldman Sex I
think it was Yannatsis, and he was talking about in
the tech sector, most of the companies that have been
laying people off grew their head count by over since
the pandemic basically because they thought all the pandemic trends
were going to keep going. Maybe there's a little bit
(04:28):
of labor hoarding. But it's a big figure. Absolutely. You know,
people in our industry, journalism, when journalists lose their jobs,
there's always trolls on Twitter saying, oh, learn to code.
You know, that's like a thing. I think at one
point even like Twitter start banning people for saying that.
But I guess the question is, right now, when you
see these layoffs, like should we learn to code or
is that not you know, is that not the career
(04:50):
safety that that it used to be? So all kinds
of questions about what is going on in the market
for tech talent. I'm sure someone's going to tell us
to learn to code. For the record, I can code
just in like very non useful languages. I have no well,
I coded in basic like yeah, yeah exactly, like C
plus plus and like some really basic HTML. But all right,
(05:11):
let's talk to someone who knows more about this question
than we do. We're bringing back Patrick. Thank you so
much for coming back on the podcast. Thanks very much
for having me. Absolutely so. I guess the question is,
before we even start talking about the recently off announcements,
why don't we start with like the hiring boom that
we really saw over the last two years, just massive
(05:32):
amount of headcount added and all these companies, we know
who they are, what drove that. So how about we
rule back history to two thousand nineteen and if you're
looking at recent history, as of two thousand nineteen, tech
has been on sort of an uninterrupted series of a
bunch of very good years, broad based expansion across the
(05:53):
entire industry, basically writing continuing to ride the wave that
had happened since the late pots. Is that how we
was it in English with the consolidation of mobile gains, etcetera, etcetera.
Then the pandemic happened and there was a brief pause
of okay, is this going to be an absolutely catastrophic
(06:15):
event for the entire world economy. Many bad things happened
during the pandemic, the way it played out for tech
was probably not how anyone would have expected one. There
was sort of a one two punch of a combined
fiscal response from governments built in the United States and
worldwide to stave off a huge economic disaster, which had
(06:36):
the effect of both putting money into consumers pockets and
also using the markets for assets, for example tech stocks,
which will come back to the importance so that in
a moment to a lot of the customers were, due
to various non pharmaceutical inventions, sitting at home with very
little to do other than use the internet. And so
(06:57):
a lot of commerce that had been possible on the
internet before the share of it that was soaked up
by the Internet in uh both sort of like semi
discretionary places like food delivery, but also much less discretionary
places like you know, core supermarkets suddenly shifted online in
a very very fast way. And so this combination of
(07:18):
there's more money slashing around and more of it is
falling into the online bucket led to absolutely blockbuster years
for tech companies, and it was a real like trying
to keep your fingers onto the rocket internally at the companies,
like the amount of new users that was on boarding,
at the rate of growth of the business, the raw
volumes of stuff that was going through the pipes made
(07:41):
it like difficult to keep everything up and running. And
a good news front, the business is largely successfully did
keep up and running during a time where society very
much needed them to. They also started to like readjust
their projections of what the future would look like, and
for a while it was looking like. Yeah. The phrase
(08:01):
that was going around was decades of growth were happening
every couple of weeks in terms of, you know, our
anticipated long term shift of the offline economy into the
online economy. And there was a big question of how
long does that continue for and is that pulling forth
growth growth that is happening in the future, is it
a one time spike, etcetera, etcetera. To to various structural
(08:24):
and competitive dynamics, a lot of firms spit simultaneously. This
is a pretty durable change. We find ourselves crushed by
the amount of demand we're seeing right now. We're going
to need to hire and hire aggressively to deal with
this and to position ourselves for what we see as
the you know, eventual coming out of the pandemic future.
And as a result of this, companies were mature companies
(08:47):
the Google's Amazons Facebooks of the world. We're hiring on
the order of like year over year growth across large
portions of their business. Somewhat earlier stage companies, companies that
might look like a stripe even though stripe as uh
somewhat larger these days, or early stage startups were on
boarding multiples of their pre pandemic had count as we
(09:11):
used over the course of the pandemic, so huge expansion
during the during the interval, and then as we came
out of the pandemic, companies assessed a number of things.
One the growth rates tended to go back to his
historical norms rather than this shot in the arm that
the pandemic was offering. Importantly, and you know, tech is
(09:31):
a wide sector. It touches every part of the economy
these days, so it's difficult to say with huge generalizations,
but as top line level things did not decline back
to two thousand nineteen. And again, two thousand nineteen was
not a bad year for tech. It was a, you know,
a pretty good year after a number of pretty good years.
So we haven't gone back to the pre pandemic baseline.
(09:52):
We haven't even stopped growing. In a number of cases,
the growth curve has just spent downwards, and so the
sustained like plus plus headcount growth over time didn't look
like it could be that could be sustained. And then
companies started to look at things that they had allowed
to happen for the course of the pandemic to characterize
(10:12):
these broadly. One of the things that happened during the
pandemic was due to the lockdowns and advisability of having
large numbers of people congregate in small pockets of air,
a bunch of companies went to both remote work and
remote hiring where they might not have had a huge
amount of institutional experience with that model of working before,
(10:34):
and after two to three years of working with these
newer cohorts of people, they've found that there are some
practices that they want to continue from this room at
work world into the future, and that there's some amount
of internal impetus to return to office and have sort
of a cultural reset around the office or headquarters, etcetera
(10:55):
as the sort of beating center of these firms. I've
worked remote for most of my career, myself of a
broad fan of the model. Let's say that there was
some cultural tension and companies on where the locus of
activity is going to be, whether it's going to be
in this online in zoom meetings and slack all the time,
or in the office high band with communication directly with
(11:16):
trusted peers, and a lot of companies wanted to have
a bit of a pullback towards the office, and then
they're looking more granularly at the classes of people to
the heart over the last couple of years, and found
that in comparison to prior classes, there was a bit
of cultural drift relative to where the companies want their
baselines to be, and also in some cases a bit
(11:38):
of a measured productivity difference versus where they wanted their
baselines to be. That's sort of expected because when you're
pulling out all the steps to hire, you like necessarily
you have to be a little less choosy than you
normally are. You have, you know, to the extent that
you describe any value at all to the in person
interview loop, which I describe relatively little value too, but
(11:59):
hopefully like slightly greater than zero. You you lose that
amount of signal and they're sort of hiring in a
slightly more challenge fashion than usual. And so I think
that companies will be pretty quiet about saying, but we'll
we'll say to themselves is we probably have a few
more regrets in like hiring classes than we did in
(12:21):
the hiring classes as a percentage, Patrick, this actually leads
to something that I want to ask you, but what
does blow actually look like in the tech sector, and
you know, is it something that only emerges as business
activity actually slows down, or even in one would you
(12:41):
have characterized tech as bloated. So it's difficult. This tech
sort of subsumes more and more of the economy into
its every increasing embrace to make like huge paint with
raw brush assertions across all of it. But let's see
where to start here. So one, the number of things
(13:04):
that are done in these large companies are extremely varied.
People might have a image that like most people who
could google our engineers, that's actually not the case. Depending
on the company we're talking about between twenty and the
people who work at the company are technologists broadly written,
they are software engineers, their system administrators, their designers at
(13:25):
some companies reporting to the same division. And then the
rest are every sort of worker that you would have
in any company in the economy. Lawyers, regulatory people, customer
sport agents, etcetera, etcetera, etcetera. Management layers upon layers of management.
So what does what does company growth look like? In
one case, it is staffing up more teams to work
(13:47):
on products that already exist. Sometimes staffing teams that sort
of like grow with the the rate of usage of
your products. So like customer service teams typically grow relatively
linearly with the usage of your service. Sometimes it's teams
that grow relatively nearly with the size of your organizations.
So as companies were having these sort of like unprecedented
(14:09):
amounts of employees getting on boarded every year, they needed
larger recruiting divisions to staff up there are other employees,
and it's just based on like the productivity math of
a recruiter. And you can, like finger to the wind
that if you hire a recruiter, that recruiter will be
able to hire twenty five people in the year. And
so if you need to hire four thousand people, then
(14:30):
you know, work math backwards, you require a hundred sixty
recruiters that you didn't have previously. That will tend to
cause your recruiting division to get larger as you are
doing a rapid expansion, and then it will contract faster
than the rest of your company will when you decide
to take your foot off the gas bottle. So those
are the the things that are sort of less inside
(14:51):
of your control. You you just need to keep doing
them to run the business, and then you're making some
more speculative investments on like what are what is our
new product line up going to look? Like? What features
are we going to add? And so the basic unit
of organization within an engineering organization these days is a
single engineering team will typically be like five to eight people,
(15:14):
and that team has a mental bandwidth to deal with
three relatively narrowly scoped problems. And so the more that
you want your software services suite, etcetera. To do, the
more like narrowly scoped problems that come into its domain,
more like five to eight people engineering teams you need,
and so you might find yourself in a position where
(15:34):
you've hired like five to eight people to work on
three relatively narrowly scoped problems somewhat opportunistically, and then when
you you know, come to three and are thinking very
rigorously around like, Okay, we think we're a little bigger
than we were when we were efficient back a couple
of years ago. We think the economic environment not be
(15:57):
might not be as strong and as we were model link,
Which of all the problems in our company are the
ones that we definitely need to keep focusing on, and
which can we refer into later or just our corridor
business right now, then perhaps like some of these nearroly
scoped problems are not at the top of our list.
And then if you consider, you know, like this product
that we thought we would bring to market in three
(16:19):
maybe it will not be brought to mark till then
there might be like ten teams implicated by that that
you do not have propletied for I have a lot
of questions. You know, when Ellen bought Twitter, and he,
like what you know, much more aggressive with the layoffs
than anything else that we've seen. There were all these
like vcs and stuff. A common to the Dirty Secret
(16:40):
and Silicon Valley is that all these companies could do
that they have fifty of their employees not really working
on anything and not really contributing anything. And like, thank
you Ellen for showing that this could be done. And
Twitter still is operating. Although I don't know how the
businesses or whether he cut too deep to the bone
or whatever, but like, would you hear that, like is
that the case that just like over the years, setting
(17:01):
aside the unrealistic expectations of one and maybe two, was
there just a wide scale over hiring relative to the
needs of the business. So tech has been in sort
of a land grab mode for essentially all of my
adult life. We certainly haven't hit the asom to oute
of how many things in the economy can be orchestrated
(17:23):
by software. We certainly haven't hit the assom tote of
how many human and human interactions will be intermediated by
a technical system happening over a smartphone, etcetera, etcetera. In
that sort of land grab mode, you aren't simply like
trying to answer what is the minimal set of things
we can do with the minimal number of people, but
are sort of opportunistically looking at what are the next
(17:44):
ten things that we can try such that one of
them becomes a company defining product feature, etcetera, etcetera. I
have a little bit of risk reflective contrarianism when people
say all tech companies are overstaffed by Could you cut
eighty percent of people who work at tech companies and
still have something functional at the end of the day.
(18:07):
Probably true, that would be extremely painful. But if you
went into a very different mode of operation and just
wanted them to continue the products and services they had
three years ago, possibly that could be done. Probably wouldn't
be optimal for any of them. That's one major reason
why nobody does it. There's also some not gone like cultural, etcetera.
(18:28):
Effects that make it virtually unthinkable. If you were an
executive at at a tech company and you were sufficiently
in your cups and had a had a heart to
heart with someone and said, what's the true number of
Like if I could wave a magic wand and no consequences,
where would our staffing be? Would probably be like eighty
five to ninetent of what it is currently. I think
(18:49):
I think most people would say, like, oh, there's a
bit of there's a bit of like I hate the
word fat in this context about you know, a little
bit of fluff around the edges, but we're not in
systemically a terrible place. And I think you know you
you would get different numbers from different people in different
parts of the organization, but that feels like plus or
(19:10):
minus right to me. Should be noted that I was
a beer worker b rather than the sort of executive
that would be tesked with making that kind of decision. Patrick,
(19:34):
you mentioned the sort of impetus towards creating company defining features,
and this is also something I've always wondered, is there
a bias in tech towards creating new products and our
employees and engineers you know, rewarded for doing new things
rather than maybe maintaining the old ones and perfecting those. Oh,
(19:57):
this is an extremely important thing to understand and the
behavior of the large tech companies from outside of them
that they all have what's called a PERF process in
the industry, it's called PURF outside is a performance review,
and the performance reviews are largely how a company takes
creative work that is done over this time scale of
like quarters and years, and it is often sort of
(20:21):
indevigable and very area and reduces it to a number
such that the company can dole out things of value
like promotions and bonuses and career paths, etcetera, etcetera. And
PERF happens on a semi annual or annual basis, and
the way it PERF works that most large tech companies
is heavily biases in the direction of getting your name
(20:42):
attached to new things that shipped in the world versus
you know, I was assigned to this legacy product, the
product did not go down for six months, you should
definitely give me a bonus on that basis. Oddly enough,
this is not straightforwardly the things that is in the
company's interests because all of the money is made by existing, well,
not all the money. The supermajority of money in the
(21:03):
tech companies is made by satisfying customers you already have
rather than getting new customers, and the supermajority of money
is made on your oldest and tourist products rather than
the new stuff. But institutionally, tech companies biased towards we
want our best people to be on the new things
all of the time. And if your individual best people
(21:25):
want to be, you know, doing the hard yards that
keeps the old stuff running, they will quickly be dissuaded
by their mentors and managers, etcetera, and say don't, no, no,
that is not the way to exceed expectations. You will like,
if you only do great maintenance work for the for
the next couple of years, you will be, you know,
(21:45):
severely career limited here. So figure out something new to
do and make sure your name is attached to it
in a way that is legible to your manager and
your manager's manager and this performance roview process. So let's
talk about the layoffs that we've seen. Because you said
something interesting in your first answer, which is that's sort
of like hiring discipline, hiring quality. During those crazy years
(22:07):
of one part of may have been loose, maybe not
as the standards were a little lower, or maybe people
just didn't fit or something like that. When companies these
days are now or recently making the decisions about who
they're going to let go, how skewed is it towards
that sort of recent cohort. Because the other thing I
(22:29):
could see is that look at many companies, you probably
have people who have been there forever who are getting
paid extremely high salaries or very good salaries just based
on the fact that they got some bump every single year.
Maybe they're not pulling their weight to some perceived degree
as much as they used to be. So how much
of the you know, when when are they when the
(22:50):
executive look and say, okay, we're gonna make cuts. How
much was it skewed towards the new cohort versus seeing
as like this is an opportunity to get rid of
some highly paid employees maybe don't add as much value
as they want. So a disclaimer off the top, layoffs
are like understandably traumatic for the people who go through them.
I don't want to minimize that. At the same time,
(23:11):
I think we often, particularly as as workers in this industry,
sort of like advocate responsibility for understanding the like structures
that cause these things to happen in ways that are
not in our interests. So broadly it's it's good to
have like open conversations about how these sort of decisions
are made. I think it is different on a firm
differing basis, but broadly speaking, you would not want your
(23:33):
simply to like roll back for the last six months
of your hiring. There's a couple of different reasons for that.
One is that when you're dealing with these complex ecosystems
that sufficiently large companies ecosystem to itself, there's all sorts
of levers that you are like managing a parallel and
one of those levers is that you are attempting to
(23:54):
balance the seniority ranges in various parts of your organization
such that you always have a mixed within some error
bars of how many people that you have they're acclimating
to the company versus how many who have acclimated and
could do productive work, versus how many are in that
senior mode where they can lately parachuting to consult on
things and do the architecture stuff that you're more intermediate
(24:17):
employees might not be able to do. Yet, if you
sort of create a bubble in the pipeline by concentrating
your cuts in the people that were only hired in
the last six months to two years, then you are
setting yourself up for a bubble a couple of years
from now where you have far too few people at
a portion of the experience curve to do work that
(24:39):
you urgently need to do work on a week by week,
quarter by quarter basis. And so if you come to
the conclusion that we've hired a few too many people
over the last couple of the last couple of months,
what are we going to do about that? You have
to distribute your cuts over a larger number of cohorts
than the most recent cohorts, or you will set yourself
(25:00):
fun for some pain. There's also some compliance and legal
issues that come up with is employees and you already
a protected class in particular jurisdictions, which also plays into
it into a little bit. But the biggest reason is
to avoid causing the operational issues for your company layoffs.
Is performance management that is a thing that exists in
the world. And so you know, if you were hearing
(25:22):
skeptical FECs on Twitter, that they would say about large
software companies is not merely that they were a little
bit flabby, but that they were a little bit uh
self assured of their position in the world, and it
had too many good years in a row. And if
you got attached to them, you could you get a
job in a corner office and not do all that
much and still be fine. I think that is a
(25:44):
little exaggerated, but let's say there's certainly cases of it,
and there's certainly some people like mature into a career
where they continue being impactful over years and decades, and
some people end up in sort of a tenured professor
mode where they've become critical to the organization because they
know a couple of things that the organization needs to know,
but they don't bring the same intensity that they used
(26:06):
to in their career. And then there are some people
who have like successfully created a niche for themselves inside
the company, but the company might not desire to exist,
and nobody wakes up in the morning and says today,
I want to do layoffs. But given a circumstance where
everyone in the industry is doing layoffs, some executives might say, okay,
(26:28):
it is a good time to reevaluate and like turn
up the heat a little bit on our performance management
and say, okay, is there anyone who has been coasting
a little too long? Is there anyone who has uh,
you know, created a secure little nest for themselves in
a way that that nest does not add a lot
of value to the company. Given that we we need
to usher some people on two new positions, let's start
(26:50):
with that first and then move to the cuts that
are going to take more mental energy to do. You know,
we're talking broadly about hiring discipline and the idea of bloat.
And this is a slightly loaded question, but to what degree,
if any, do you think the sort of maybe monopolistic
mode that some big tech companies have built around their
(27:13):
businesses has contributed to some complacency on the hiring front.
And a little adverse to the word monopolistic, but I
think I get what you're getting at, and that there
is certainly a lot of rent created in the technology
industry where these are some of the most effective businesses
(27:33):
ever created in any industry. Google AdWords will print a
ginormous amount of money next year, and almost no amount
of action taken by any set of first actors internal
or external ad to Google will cause Google AdWords to
not be worth many, many, many billions of dollars, and
so the margins on it are very high as well
(27:57):
in comparison to you know, we we're talking last time
about the airline industry, where the airline industry has struggled
mightily to maintain like single digit percentage positive margins over
a multi decade timeframe. Tech doesn't have that problem. The
nature of these very sticky products that shearsket size of them,
and the margins do tend to create a little more
(28:19):
room for that flabbiness than in exists in many industries
that have more of a cutthroat reputation. This is sort
of the polar opposite question. But nowadays we hear a
lot about the possibility of companies hoarding labor when it
comes to tech. How much of that do you think
(28:40):
has actually gone on in the sense that do you
see tech companies opportunistically hiring people just so their competitors
can't get their hands on them. I've heard this theory
advanced many times, and honestly, I don't think it is
very explanatory, and sometimes it's phrased Google would rather hire
a particular talented engineers so that they don't create a
(29:02):
startup and then eventually become competition to one of Google's products.
If hypothetically that were something that actually motivated executives at
tech companies, there would be a number of things that
would be easier to do than quote unquote labor hoarding
that we don't do institutionally. So in finance, there's this
institution of gardening leave. Tech doesn't institutionalize gardening leave at
(29:25):
any level almost anywhere in the industry. And if you
were thinking about let's prevent highly talented people from doing
interesting things for our competitors or for new startups that
they could create. The people in the industry that you
have the like tightest speed on their productivity level are
your existing employees, and so you would be you would think, oh, well,
like the natural place to start is like start with
(29:47):
people who are already work here and say, if you leave,
we would like to buy twelve months of your time
sight unseen, and no one does that. And there's other
things that you can do Broadly tech is there's always
a bit of push and pull between the needs of
a company and the needs of employees, but broadly tech
who is strikingly pro worker relative to many industries in
the United States. These things that are done that would
(30:08):
be consistent with the labor hoarding hypothesis just are not done.
You know, you can talk to the people that are
involved in the decisions that that are read on the
outside as labor hoarding, and they never advanced that as
a reason to you know, buy up. A new company
that has four engineers attached to it is typically phrased
something more similar to, well, this is a team that
(30:29):
seems already jelled. They're clearly highly highly productive individual contributors,
and we could have a bunch of engineering recruiters work
for months to find for similarly talented individuals or the
m and a team can like tick one box off
in Q one and get them all in the door
for the price of one low check. Let's do it.
The notion of like and let's take this team off
the table, so they don't, you know, have a market
(30:51):
success in three years and create something competitive with us
never comes up. Okay, we started talking about why the
(31:14):
hiring boom happened. In the first place, we've talked about
maybe some of the decisions on who is getting cut.
Let's talk about the sort of prospects for the people
that have lost their jobs and or the people that
are thinking about going into a career in tech. So
how quickly do you perceive that the people losing their
jobs over the last several months are finding new offers? Like,
(31:35):
let's start really simple. Can I tack something onto that,
which is how how fungible are these types of jobs
in reality? Yeah, A long time ago, in a place
far far away, during the dot com crash, I was
graduating from university and the Wall Street Journal was which
read the Wall Street Journal every day with my father
(31:55):
growing up. It was how I learned to read the
Wall Street Journal. Could do no no wrong in my
as as a an undergraded engineer, and the Wall Street
Journal was pretty decided that yep, engineering as a field
is done in the United States of America. Henceforth, all
engineering will happen in Asia. And I said, oh, chucks,
I really wanted to get an engineering job. I guess
(32:16):
I have to move to Asia, and so I did.
Oh back, now we know the origin story. This is
the backstory of how I ended up spending my entire
adult life in Japan. Now that ended up being a
good read like a good life decision for me for
entirely unrelated reasons. But it turns out there were, in
fact engineers hired between two thousand and four and two
(32:37):
thousand and twenty three in the United States, and so
reports of the field's demise were heavily exaggerated. If you
are considering a career in engineering, every reason you had
to consider a career in engineering in is like still
a reason to do it. So this like minor wobble
that will be forgotten in a matter of months. Please
(32:59):
don't allow it to like cause you to make major
drastic life decisions. Although life is what happens when when
you're busy dealing with these little wabbles. Okay, so that
out of the way. How fungible are people? Broadly speaking?
In the early levels of career, tech tends to cast
a very wide net and hire people for what's often
(33:20):
called horsepower, with the expectation that they will be able
to specialize over time. There is some degree of worry
that if you spend ten years or fifteen years in
a particular industry doing. The quote often used is have
the same year ten years in a row, then you
will end up over specialized and only be available for
doing that sort of thing in the future. Depending on
(33:42):
the thing you are doing, there might be a sharply
limited set firms for which that is relevant. But broadly speaking,
the engineers that were hired to do anything in the
first five seven is years of their career are broadly
expected to be able to do not quite anything, but
like a large subset of all the things that a
(34:03):
tech employer could want an engineer to do. And so
the liquidity in the tech market within like a broad
class like recruiters or engineers, etcetera liquidity between job titles,
exact roles, exact companies business model of the company is
very high. And I'm forgetting what Joe's original question was, Well,
so are they finding just a short term like you
(34:26):
must hear from people, you must talk to people like, uh,
people just got cut off? Are they recruiters already reaching
out to them from different companies? Structurally tech company this
would be a bad pool quote. Structurally tech companies are
like sharks. Okay, we're gonna we're gonna pull their quote.
Just just like sharks, like the way that their gills work.
(34:48):
They have to keep swimming or they stop getting oxygen.
And that's an unfortunate thing for most creatures. The tech
companies because of their staffing models, they and that thing
we talked about earlier, where they are constantly mixing the
number of people at each level of seniority within the company.
They have to keep hiring. And so even if an
individual company decides like, Okay, we're going to like push
(35:09):
pause for six weeks and do quote unquote hiring freeze
one the amount of time and they can actually do
that and not severely damage the business is limited. So
it pauses always temporary unless the company is going down
the tubes. And like the large tech companies certainly are
not going down the tubes. Some startups might get shaken
out at the march and tow to UH funding constraints,
(35:30):
et cetera, But the like the overall business of the Internet,
continues to grow apace. So pauses are temporary nature. And
there exists, you know, like many different companies inside that
the broader ambit of tech. Some of them might be
positive any given moment, Some of them are you know,
still attempting to make new investments for three and some
(35:53):
of them while they're not in uh sort of rapid
growth mode, growth mode for doing things like you know,
we have to back fill for people who are leaving
the company, and in a typical year at a typical
tech company, that may be like ten percent of our
engineering staff. So if we've got two thousand engineers, we
have two hundred engineers that we are slate to hire
in Interestingly, one of the things that cost a bit
(36:15):
of the over hiring was companies have this model for
what percentage of people will leave in a year and
therefore how many you need to hire just to stay
at the current level of employment that you have. And
when the economy started wobbling in, what happened was the
rates of voluntary attrition that companies, meaning that people who
resigned out of their own volution, went lower than the
(36:36):
model predicted. And because you need to like set in
place a process that takes months to hire people, but
the process of deciding not to quit is not visible
for those months, that resulted in sort of like a
hiring overhang, and so companies overshot their targets for how
many people would be in the company, which doesn't sound
(36:57):
like an easy thing to do, but it is a
very easy thing to do if there is a sort
of like sharp change and employee behavior with regards to
things that they have total control over and don't have
to announce to you, like deciding to leave or not
leave in a statistical fashion. Are there signs that tech
workers should look out for that they're about to be
laid off? Like do you stop being a signed new projects?
(37:19):
Do your access codes get cut off? Does someone come
take your stapler off your desk? Like? What exactly are
the warning signs that you might be in the danger zone? Many,
many tech people have a large degree of stress with
regards to whether I'm doing well, am I on the list, etcetera, etcetera,
And I don't want to add to that stress. Broadly.
(37:41):
You should. You should have an understanding of how performance
is calculated at your company, and consider that official view
of your performance to be perhaps more important than you
would naively believe it too, because the official view where
you're the entirety of your performance is reduced down to
like one number, I'm a full or for this six months.
(38:01):
That is the only view that is going to be
available to someone who might be two, three, four steps
above you on the ladder when they're going to make
hard decisions in a hypothetical future where they're making hard decisions.
So the things that cause formal visibility to accompany are
anomalously important, and the career oriented people around you, who
(38:22):
are very good at work in those systems their advantages
will find advantages based on that. But I wouldn't, you know,
over rotate on perfect The only thing we're thinking about
seems simple, But just do great work and then make
sure people are aware of the fact that you did
the great work, and then things will tend to work
out in a career fashion over only a long period
(38:43):
of time, not gun. So I have a question that
bridges this conversation with the conversation we had last week
about I T. And I realized I should have asked
it last time. This whole episode has been because it's
actually I actually only just had one question from last time.
I had to come up with a whole excuse for
why we needed to have you back out just so
(39:03):
I could ask this. But occurred to me, you know,
like in the in the business press, we're always reporting
on c E O s getting fired or let go
and hired. Sometimes CFOs I don't see much coverage of,
like c t O S or c I O is
like the people who run the internal tech systems being
let go for poor performance. I actually think the only
time I can ever remember hearing any sort of CTO
(39:26):
or something losing their job for poor performance is probably
like fifteen years ago when Twitter was always having the
fail wills and like they weren't scaling very well during
the boom years. And other than that, I can't actually
recall a time in which I like read a story
about it, you know, a CTO being laid off for
bad performance. How often does that happen? And you know,
(39:47):
in the context of whether we're talking about tech companies
or all. You know, I think we were talking about
Southwest and others last time, Like how often do the
head of do those positions lose their job because they say, like,
our I is not good. So it's a complicated subject
for a number of reasons. One is that the degree
of saliency of CTO most companies to the media is
(40:08):
relatively low. The degree of saliency of many things that
are very important in the tech industry to the media
is lower than many people in the tech industry, but
like and that is one cause of the frequent conflict
between the media and tech. But be that doesn't two
people get laid off for for performance. Yes, one relatively
frequent thing that happens relative to the incidents of senior
(40:29):
senior executives departing is the uh sort of like fall
on your sword motion if there is a significant outage.
Is a thing that frequently happens, are frequently relative to
all causes for a departure and hesitant to give you
the example because Tokyo is a small town, but there
are a number of banks, both in Japan and outside
(40:50):
of Japan that have had disabling computer outages for like
days to weeks at a time, where that is an
extremely extremely thing to be avoided for a bank and
rules up fairly directly to the head of I T
or the CEO. And there are cases where either the
head of I T or the CEO of have left
us a result of doing that. There is one thing
that I do like about the culture that is Japanese management,
(41:13):
where in the sort of like ritualized speech that an
executive gives it that they will often say police don't
blame the people that had their hands on the keyboards
during this. The fact that this was allowed to happen
was a result of managements miss decisions or taken over
the course of years. I presided over them, and as
a result this uh, this allage. Even if you know
(41:34):
it was one person individually fat fingering something that took
us down for a week, this belongs at my door,
and I'm resigning to take responsibility for it. There are
many things I don't love about Japanese management culture, but
that bit I do like. Another thing is there are
reasons for companies to be other than other than maximally
(41:55):
public about the fact that we are removing a senior
executive for cost. If you remember the over the course
of the last couple of years, the I T sector
has been in sort of like massive boom mode. Companies
are extremely protective of their brand with respect to engineering candidates.
Nobody wants to join a organization that exists under a
cloud who's CTO just got fired for being an idiot.
(42:17):
So the thing that might happen is like, oh, well,
the previous VP of engineering wasn't quite up to enough.
Maybe they can be shuffled onto a different project, and
we're going to hire a CTO above them. If you've
already hired a CTO, that's a bit of a bit
of a more difficult thing. But like shuffles with regards
to who are the most important people in the engineering
(42:39):
organization and is there a separate product organization? Do they
report to the same people, etcetera, etcetera, are sometimes caused
by like X isn't getting it done. We want to
like shuffle in why, But we don't want that to
be seen as a repudiation of X. Not because we
care about X's opinions so much, but we care about
how this will be read by internal engineers who we
want to keep attaptioned to the company and external candidas. Alright,
(43:03):
one last small question they'll probably eject, a question that
we can talk about for a long time, but just
real quickly. So the one one area within tech that
seems like almost certainly going to be hiring like crazy
for years at this point is anything to do with AI.
And you know we all know what's going on there.
How much are the skills that some of these like
sort of cutting edge AI companies in need of. How
(43:28):
much are these skills that sort of legacy or existing
tech workers might have, or how much are the skills
that they need something that like you really need years
of like focus training in the specific area to satisfy
what these companies need. Can I can I add another
thing onto the back of that place? How many coding
jobs will something like chat GPT destroy? Ye? Should people
(43:50):
stop learning to code? Yeah? Yeah, talk about talk about
So I have a glib but true answer with respect
to our advanced AI techniques going to destroy programming jobs.
The first program or class of programs that we had
where an advanced computer was obviating the need for human
programmers was called the compiler, where instead of doing you know,
(44:12):
complex low level and instructions directly and assembly and speaking
sort of natively the language of the computer, you use
what we're called high level programming language is like C
back in the day. And then the compiler would you know,
use its magic AI powers to turn that C into
assembly language so that you didn't have to laboriously do
the assembly language itself. So every technology that gives programmers
(44:34):
more power, more capability to do things that are valuable
for human society probably increases the aggregate demand for programmers
is sort of like my high level view on the
world and if yet to see a contrary example to that.
So an interesting question with regards to AI is what
are the like, what series of steps is going to
(44:56):
be necessary to take it to market in a way
that it actually creates the value for individual people land
for society, and that it seems to have Latin within it.
And if you look at like chat GPT, if you
like I view it as an iceberg, there's the above
the waterline part and below the waterline part, and below
the waterline part has some let's say deep deep magic
(45:20):
there bracketting out that magic for the moment, it seems
like the above the waterline part was very important in
why everyone has heard chat GPT and probably used it
if you're listening to this podcast, but it hasn't heard
of like similar efforts at Google, etcetera. The reason is
that there was a you know, a product focused team
that made a relatively pedestrian piece of software like a
(45:43):
chat interface, but made it really, really good and like
work on that to the point where people's interactions with
the underlying large language model would be like sufficiently effervescent
that it would screenshot that interaction has shared it over
to Twitter, and so everything that above the water line
part is amenable to the to the technologies and tactics
(46:07):
that existing engineers have with no modification whatsoever. There you're
talking to the back end. The back end is implemented
in a different kind of magic than your back ends
usually are. But the back end has always been magic too.
That is like part of the answer. There's an interesting question,
like how much of the work is going to be
that above the waterline part. The productization of these you know,
(46:29):
creating like new forms of user interfaces, new models for
interactions with users, new metaphors that we have to teach
to people, like new you know there there might be
an entire field and like education and how to do
I don't know, prompt engineering, well, prompt engineering being how
do you type in the right series of incantations to
the machine so that the the spirited some and stuff
(46:51):
out of the ether does the right thing for you? So, like,
what percentage of the work will happen there? First? What
percentage of the work will happen on these like core
under the hood model things. A sub sub question to
that too is like okay, so for the work happening
at that model layer. Is that work going to happen
at every company that consumes models, or is it going
(47:13):
to happen primarily at open AI and Google and Microsoft.
And we can count the number of firms that like
need this these engineers on a single hand. In a
world where we count the number of firms that need
like dedicated hard AI researchers on a single hand, that
probably implies like lower total employment of them than in
the world where every firm that touches AI has its
(47:36):
own AI practice on staff. But it's at least as
of like the current state play deeply uncertain where that
will shake out. And so these are some of the
questions that get debated upon people at both the AI
firms and also like you know, if if you are
a VC that's adventure that's investing in the space, you
are probably having like a number of interesting dinner conversations
(47:56):
on okay, where does the value accrue in this chain?
Where does most of the work get done? What do
these products like expose themselves to in the life of
the user. Is it's something deeply under the hood or
is it integrated into their daily operations? Do they know
they're using an AI do they know they're using software?
Is it something that they're like directly typing in or
is it something that they're interfacing with someone who's doing
(48:19):
the typing on their path, etcetera, etcetera. Well, Patrick, this
was absolutely great talking to you. We could talk for
a long time, but instead we'll just talk to you
in a few weeks again when we have a million
more questions. Now, I'm big patigias, but I learned a
lot and really appreciate you coming back on the show.
Thanks very much for having me, and I always happy
to be come back. Thanks Patrick. That was fun, so Tracy,
(48:52):
there were so many interesting elements of that conversation. I'm
really glad we had Patrick back. I'm not even sure
where to begin, but to start, you know, his point
about the hiring boom during the pandemic, I thought was interesting,
not just that maybe these companies had a sort of
unrealistic expectations about how long does growth boo would last,
but that when you're hiring that fast and under sort
(49:12):
of extremely unusual situations, like you have that drift where
maybe you're like there's a little bit of a we're
not that happy with the class. And then also that point,
but you also can't just hire everyone who came in
recently for reasons of like seasoning and like skill level growth. Well,
to me, it kind of I guess hammers home the
point that three years on from the start of the
(49:36):
COVID pandemic, we are still experiencing these normal developments. And
it kind of gets to the macro versus micro point
about some of the recent payrolls figures. You know, all
the tech layoffs that have been announced. Are they saying
something about the wider economy or is this really a
tech specific problem? And I think, I mean, I can
(49:58):
kind of argue it either way. I think I come
away from that conversation thinking, well, you know, one were
really unusual periods in terms of hiring for the big
tech companies, and to some extent it seems reasonable that
that starts to get rolled back a little bit. But uh,
you know, it's also I take his point, is he
and I suspect it's probably true, which is that if
(50:21):
a year ago you were thinking you wanted to go
into engineering or coding or something like that, very little
about what we've seen so far in three should make
you change your mind. I thought that was really interesting
too about like sort of the questions about AI and
how so it's like, as he pointed out, like there
are other you know, places working on very similar, if
(50:42):
not equal technology. What sort of made things breakthrough recently
was the consumerization of some of the chat interface or
some of these AI images imaging things. So like how
much of like to go to market for this stuff?
Ultimately isn't sort of like familiar experiences that people already have.
It reminds me a lot and I don't mean this
(51:03):
necessarily in a negative way, but it reminds me a
lot of crypto in the sense that, like, yes, there
is a lot of hype around AI, but also in
the sense that this is a new technology that people
can actually participate in. And so the use of the
AI image generators, chat GPT, it kind of brings it
to people in the same way that they are able
(51:24):
to experiment with, you know, blockchain and different types of
money using crypto, and so it suddenly becomes a lot
more salient for people in that way. Yeah, you know,
like there's a good example because like with crypto, like
if you're like interacting with like core protocols are like
developing on ethereum like that's going to be a limited
a limited number of people know how to do that.
(51:44):
But if you're building like an exchange, there are a
lot of I mean, everyone can have a wallet, right
you're marketing or stuff like that. There are still all
of these roles within crypto that have like sort of
like consumer facing analogs to any other industry. Yeah, I
need to look up the compile tiler. That sounds interesting. Yeah,
I'm gonna go off in Google um deep learning compiler.
(52:05):
I guess for so far a job security, we still
need to learn to code. Huh. I think we need
to learn AI. I don't know. Probably I don't know,
but you know C plus plus, which is what I
learned in a little bit of no because it's obsolete.
Like no, I don't think anyone uses c post plus
and they certainly don't use it for for AI and
(52:27):
machine learning stuff. I should have asked Patrick what language?
What coding language Python? When we have them back in
the Yeah, okay, yeah, our next episode with Patrick will
be about which coding language we should all be learning
in the future. Shall we leave it there? Let's leave
it there. This has been another episode of the All
Thoughts podcast. I'm Tracy Alloway. You can follow me on
Twitter at Tracy Alloway. And I'm Joe wisn't Thal. You
(52:49):
can follow me on Twitter at the Stalwart. Follow our
guest Patrick McKenzie. He's at Patio eleven and check out
his Bits about Money newsletter. Follow our producers Carmen Rodriguez
at Carmen Armand and Dash Bennett at Dashbot. And check
out all of our podcasts Bloomberg under the handle at podcasts,
and for more odd Lots content, go to Bloomberg dot
(53:12):
com slash odd Lots when we push the transcripts Tracy
and I blog. If we have a weekly newsletter that
comes out every Friday, go there and sign up. Thanks
for listening
Hosts And Creators
Joe Weisenthal
Tracy Alloway
Popular Podcasts
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Therapy Gecko
An unlicensed lizard psychologist travels the universe talking to strangers about absolutely nothing. TO CALL THE GECKO: follow me on https://www.twitch.tv/lyleforever to get a notification for when I am taking calls. I am usually live Mondays, Wednesdays, and Fridays but lately a lot of other times too. I am a gecko.
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.