So, you're a senior engineer. I would

like to offer you something extremely

special. A re a real one-time deal here

there, fella. Okay, you instead of

coding, you can review junior and

mid-level AI generated PRs from here on

out. The trade I will give you this, the

single worst vesting schedule at any of

the fangle level companies. I know it's

a deal you just can't refuse. If this

doesn't sound familiar to you, then you

haven't heard the fantastical news

coming out of Amazon right now. It's a

it's a bit of a kind of a snaky tale.

So, we're going to kind of have to start

off when Adam met Eve. So, Adam, of

course, is the 14,000 job cut in October

and Eve would be the 16,000 job cut in

January. Amazon has been on a tear

recently doing a lot of laying off.

event. Even though the CEO of AWS has

largely been supportive of this idea of

keeping juniors around because it's very

very important to build up the talent

pipeline, the CEO Andy Jasse has been

very explicit. He wants Amazon to

operate like a startup. Yes, Amazon

wants to operate as a startup. The one

fang company that does give you the

worst investing schedule and is likely

to overwork you completely is

effectively a meat grinder would like to

be viewed as a startup. Right? So how

Amazon was able to release so many

people from their working has largely

been attributed to AI saying hey we're

getting a lot of you know we're really

making inroads here we're really

becoming more efficient we need to

become more lean we need to become more

efficient we need to be able to move

fast and the way to do that is to reduce

the amount of kind of organizational

complexity and then give the power of AI

to our employees and during all that

time Amazon actually set internally a

target to have 80% of developers using

AI coding tools at least once a week.

Now, my this is my favorite part is that

to use AI coding tools at Amazon, you

can't use Cloud Code or Codeex.

No, you have to use Amazon's in-house

tool, Kira, by January. Now, you're

probably asking yourself, "Okay, but

what's this business about seniors

having to just forever review code?"

Okay, I'm getting there. Okay, this is

called foreplay. We're we're getting to

the story of the high-profile events

that Amazon went down with. One of them

was caused by their very own internal AI

accidentally deleting an environment and

another one was 6 hours down for their

Amazon retail site. So, you kind of have

to put all this stuff together. They've

had large amount of organizational

change that's kind of led to maybe some

knowledge gaps that have been presented.

Second, they've been mandated to use

more AI to kind of fill in that

cognitive debt of the firings. And this

led to senior vice president Dave

Treadwell made the normally optional

weekly engineer meeting mandatory. Okay,

we're getting into some serious

business, boys. They are now Okay,

things are happening. 5 days ago, they

had one of the biggest kind of outages

they've had in a while. Lots of reports.

Down detector at the highest was

reporting 22,000 concurrent people

saying, "Hey, this thing is going down

for me." And during this meeting, he

said, "Gen AI tools supplementing or

accelerating production change

instructions leading to unsafe

practices." meaning that they don't have

a process anymore, right? Their process

was a certain way. They've let go of

30,000 people or 10% of their corporate

workforce and now they're filling in

cognitive debt. They're doing they're

having people run multiple jobs. Some of

the guard rails aren't quite as

wellestablished as they once were. And

their whole goal was to just fill it in

with AI. Now, you and I would would

probably make some changes after seeing

this, right? Because you and I would see

this and go, "Okay, maybe maybe I'm the

problem. Maybe we gota take the foot off

the gas and slow down there, bucko. Very

natural response to the situation. This

is not the direction Amazon decided to

take. No, no, no, no, no, no, no.

Instead, here was their plan. Okay, we

have been crushing down the levels of

organizational complexity. Okay, we've

been wanting to fill in the cognitive

debt with AI. For those that don't know,

by the way, cognitive debt simply means

there's parts of your organization in

which nobody knows how they run. you

have a debt in there, but it's purely

cognitive. So, if you just like say let

an AI write a half your program, you

don't know how half that program works.

You kind of know that it exists. It's

over there on the side, but you're just

like, I don't know how that thing works.

It just is there. It's the same things,

say if you fire 30,000 people. There's

just a lot of stuff you don't know

anymore. It it just exists. It runs. It

runs in production, but nobody really

knows why. Perhaps this is part of the

reason why that Dynamo DB kind of going

down and causing an entire issue at

Amazon that could have been attributed

to cognitive debt. So Amazon and their

infinite wisdom said, "Okay, we fired a

lot of people. We have a lot of

cognitive debt."

But we do have AI. All right. This is

how we're going to fix it. seniors,

senior devs, from here on out, if a

junior or mid-level engineer, that would

be like an L3, L4, make a change, an L5

needs to sign off on it. Seniors are now

responsible for the code changes made by

juniors. Now, some of you may think

that's not that can't be that big of a

problem. Well,

hold on there, bucko. Okay, let me show

you this right here. This is GT. That's

short for Gary Tan. In 45 days, he was

able to produce

150,000 lines of code and 150,000 lines

of test code. By the way, Rails was

originally designed to be a framework to

make a blog. I don't actually know how

one even I I'm not even sure how one has

that much code. It's a mystery. Now,

imagine you have an entire fleet of GTS.

And these GTs, well, they have mandates

from Amazon to use a lot of AI and

they're being measured and they're

required to use their specific Amazon

tool, Kira. You want to guess what's

going to happen? Well, like any metric

that you put down, if you say, "Hey, I

want to see people use X." Guess what's

going to happen? People are going to use

X. So, now a lot of people are going to

be using X. Especially with all these

layoffs and everything happening,

there's going to be a bit of panicking

going on. And so, we're going to want to

make upper management very, very happy.

So, I'm going to excel in the AI

category. I'm going to crush out some

bugs. Seniors are going to get just an

onslaught of insane AI dril coming

straight at them in the old merge

request and they have to sign off on it.

I'm not even sure if the seniors are

going to even have time to do anything

but review code. I think Amazon's going

to be the first company that's actually

realizing that changing really fast

actually has a lot of detriment. And I I

I do I'm going to throw something out

there. I know this is one of those

dangerous moments in life. never make a

prediction in a YouTube video, but I do

think we're going to see a lot more

people jump in and say, "Hey, we

actually have to slow down. You can't.

Nope, you're not allowed to move that

fast." And if you're kind of on Twitter

a lot, this would seem like an insane

take. People are going to say, "No,

you're completely wrong. You actually

have no idea what you're talking about."

But for those that are a little bit less

online, maybe this would be a good tweet

to read. We spoke to a company today

whose security team is so concerned by

AI code that they're considering banning

AI tools. Your first reaction might be

they're going to be left behind. But if

you are practical, their concerns aren't

invalid. If you are a huge multinational

or with tens of thousands of employees

and they just got a button that appears

to do their work, it's going to get

pushed a lot and the process around

knowing what is making it to production

is totally melting. Being honest, we're

all getting a bit lazier. See that Kira

related AWS outage as a real life

example. So, they're genuinely arguing

over how much this is going to be

allowed, especially since their net

productivity gains for the average dev

seems to be pretty low. And he's right,

right? Like, just think about what

Amazon's doing. If they have all this

speed happening, like, I'm going TO

PRODUCE SO MUCH CODE AND then they're

like, "Okay, now that you produce the

code, you need the senior to review it."

Like, what's going to happen? Dude, it's

going to be so much slower. Okay. Bam.

Okay. Now, you have to wait a couple

hours before the code even gets

reviewed. Once the code gets reviewed,

it gets sent back to you. You're going

to have to make changes. You're going to

have to go forward. Then the senior a

couple hours later is going to

re-review. It's like we've reintroduced

the slowest part back into the system

because the system itself is a little

bit silly. Like it it is when you think

about it. The system is a little bit

silly. This is what's being sold right

now. Hey, we have a tool that will

generate code. Cool. We have a tool that

will also do the GitHub PR. Cool. We

also have a tool that's going to be

generating all the tests. Okay, cool.

Like, so what are all those tools? Oh,

it's all the same thing underneath the

hood. Wait, wait a second. You're

telling me that you have a tool that

writes the code and then I have to have

a separate tool that costs $25 APR to

review the code that you wrote. Hey,

wait a second. Am I being bamboozled?

Hey, why don't you just write it better

the first time? Wait, how's that not an

option? Anyways, I just wanted to yap

about this cuz we're just in this

strange time, right? Because the tools

aren't quite there. They're not quite

there, but they're being sold as the

replacement currently. So, we are just

we're finding ourselves in probably the

most unstable time. And the most obvious

version of this is GitHub. GitHub has a

91.51%

update. 84 incidences in the last 90

days. There's almost not a single day

that a part of GitHub isn't down for 30

to 50 minutes every single day. It just

doesn't feel good. You know, we're just

in a funny time. Anyways, the obvious

solution, which if anybody's wondering,

is we just simply need more AI. I mean,

that's honestly what's the problem is is

that you're you're holding it wrong.

You're prompting it wrong. My shaman's

better than your shaman. And honestly,

in 6 months, all these problems are

going to go away anyways because Opus 5

is going to be dropping about September

time. And by that point, everything will

be fixed. Don't worry. Biggest coding,

it's largely been solved again. Hey, is

that HTTP? Get that out of here. That's

not how we order coffee. We order coffee

via SSH terminal.shop. Yeah. You want a

real experience. You want real coffee.

You want awesome subscriptions so you

never have to remember again. Oh, you

want exclusive blends with exclusive

coffee and exclusive content? Then check

out CRON. You don't know what SSH is?

>> Well, maybe the coffee is not for you.

Living the dream.