Software developers in the US get paid

around $113,000 to $200,000 per year

with a median at around $150,000.

While the US government actually puts

the range closer to $132,000

compared to the Stack Overflow survey,

other countries like Australia, Canada,

UK, and Germany all fall behind the US

when it comes to wages for developers.

But when we look at job postings across

countries, Australia and Canada are

actually showing stronger numbers in

comparison to preandemic levels. And

meanwhile, the tech industry overall is

still recovering from the massive layoff

that started around the end of 2022 to

early 2023. And as the industry is still

recovering, the demand shifted away from

pure software developer roles to machine

learning and data engineering roles as

AI becomes more dominant. All of this

points back to the reality that while

there are 47 million developers in the

world, more than half of the developers

are in the age bracket of 18 to 34 years

old. And that is the exact demographic

for software developers that are

suffering the most when we look at the

headcount for current employment in the

industry. Welcome to kale writes code

where every second counts. Quick shout

out to Cotlin sponsoring this video.

More on them later. You probably have

seen this chart from Indeed hiring lab

before where it shows you an index of

software developer job postings across

time. But unfortunately there's a huge

misunderstanding when it comes to this

very chart. The chart here shows you 100

in the y-axis as a prepandemic level

baseline anchored in February 2020. And

relative to that point, we measure the

variance at each data point adjusted

with seasonality. In other words, around

February 2022 where the index showed a

value of 234, which means even in

consideration of trailing three-year

seasonality was about 134% higher than

the baseline. Meaning here we hire 24

people as opposed to hiring 10 people

for example. Now when we fast forward

from this peak to today, the most recent

data point shows that the US is at 69,

the UK is at 63, Germany at 58, Canada

at 78, and Australia at 122. So what we

are seeing here in the US is a slow

recovery towards the quote unquote

normal baseline at 100 in comparison to

February 2020. Now when we overlay this

with major flagship models starting with

GPT 3.5 in November 2022 that launched

ChachiPT to the world GBT4 in March 2023

and the first commercial reasoning model

01 announced in September 2024 followed

by the Deepseek moment that happened in

January 2025 along with many other

breakthroughs like Sonnet 3.5, Opus 4,

Gemini 2.5 and Gemini 3 all played a big

part in the industry. We also have many

coding agents mainly in 2025 that took

these models to disrupt the software

development industry and tools like clot

code, open code and more recently openi

codeex app really started to poke at the

industry whether we really needed all of

47 million developers that we have

currently in the world. So as you can

see there's a bit of a tugofwar between

where the baseline should be now that AI

is changing the industry. In other

words, is this an early sign that maybe

we will never reach back to the

prepandemic level and we now need a

different baseline than February 2020?

When we look at the Dora report, while

some early adopters of AI coding go all

the way back to 2022, a large majority

of developers actually started to use AI

coding at around 2024 and onwards, which

is pretty recent. And as the underlying

coding agents became more and more

complex, there's been this K-shaped

divergence between developers who

orchestrate their work entirely using AI

and developers who use AI merely as a

tool. And leading experts in the dev

industry like Uncle Bob started to

notice how when he goes from two agents

to three agents, he started to become

the bottleneck in coding. And similar

sentiment has been shared with Andre

Karpathy where he shared how he felt

behind as a programmer using AI coding

agents. As you can see, given this lag

in the AI adoption among developers, the

K-shaped adoption between coders who

embrace the tool and coders who say that

AI only moderately automates their task.

Now, all of this crescendos into the

split between entry-level developers who

are early in their career, who seem to

be suffering the most. And with that we

need to talk about cotlin. Intelligent

idea is an environment that lets you

build while working with whichever

coding agent that fits your workflow.

Whether it's codex app, clot code or jet

brains juni. I created a very simple

code that allows you to spin up a

backend system that could read customer

log documentation that I have here in

this file. And as easy as the code

looks, I can now use curl commands to

reach the endpoint to fetch information

contained in it. And now with their Juni

AI coding agent, you can easily use it

to continue coding by asking it to

further implement my project by adding a

new route that could access file by

file. Try out Cotlin's easytouse

back-end system and their Juni AI coding

agent that helps you ship products

faster. The common sentiment out there

is that AI will take away all

entry-level positions to start.

Enthropic CEO Dario Amade famously said

that in a few years 50% of entry-level

white collar positions will be impacted

by AI. While claims like this certainly

needs to be analyzed more thoroughly, it

does speak to at least some uncertainty

in the industry whether companies should

hire more developers, which is reflected

here where we're still behind the

prepandemic level. And looking at other

countries like Australia and Canada,

where they're nearing back to where they

used to be or even above the level in

Australia, the US still seems to be on

the fence about how hiring should look

like with AI in the mix. At the moment,

however, machine learning engineers saw

a nearly 40% jump in demand, data

engineers seeing 10% jump, while

front-end and mobile engineers actually

shrunk more than 5% going from 2024 to

2025. And meanwhile, the pace of how

fast AI is being used is not stopping.

According to the semi analysis report,

more than 4% of public repos in GitHub

use clot code, which goes to show how

fast AI is being adopted to those who

actually embrace it. And this kind of

acceleration will likely continue

causing the split between 52% of

developers who don't use AI agents while

the rest are trying to learn how to best

incorporate AI into workflow as fast as

possible. I think it's worth mentioning

why this K-shaped adoption is occurring

among developers. One of the biggest

components comes down to trust where a

large majority of developers simply

don't trust AI's response in their

coding needs. But this sentiment will

certainly improve as the underlying

model gets better in time. But the

biggest issue is when it comes down to

workflow. Just like how Uncle Bob

mentioned in his tweet, how when you go

from working with two agents to three

agents, he becomes a bottleneck. Coding

with AI, and I mean actually letting AI

do the majority of the work requires a

complete change in how you actually

work. And there's a big resistance when

it comes to this change. Using coding

agents like clot code and codex app is

vastly different than copying and

pasting code into a chat or even using

an extensionbased coding assistance like

client and rue. While client and rue

will assist in the actual implementation

task by task. Cloud code and codeex is

meant to abstract away the

implementation which changes the rule of

the game from coding to what's called

specdriven development or just

orchestrating between tasks. The codeex

app for example doesn't even show you

the code in the application and viewing

diff edits is done outside of the main

chat window. So given this underlying

change we're going through in terms of

AI adoption among developers lack of

trust in AI's response continual layoff

but hiring still being below the

baseline. Do you think this progression

will lead to AI replacing software

developers or do you think AI will

create even more demand for software

developers?