Why are some software engineers

embracing AI coding while others seem to

be getting irritated just by the

concept? I'm getting completely opposite

reactions from engineers with the same

level of training and experience. I run

a 20 person dev team and I'm seeing this

split happening inside my own team. And

it looks like it's happening inside

teams everywhere right now. 10 months

ago, one of my senior engineers, Vitily,

left these comments during a code review

of one of my AI generated PRs. He called

the code awful. He said it shouldn't

exist in our codebase. This week, the

same Vitili sent me an unprompted 8-page

proposal to build our first AI agentic

coding system. And in his message, he

named the exact moment he was mocking

with a winking emoji. What flipped in

his head in 10 months is the answer to

the question I just asked. And it's

about something far deeper than the

code. If you're an engineer or you've

got engineers around you, you're inside

this transition whether you noticed or

not. The only question now is are you

leaning towards being an AI native

software engineer or a niche handcoder?

When I see people talking about AI

coding online, it gets framed as two

camps. Hand coders versus AI native

coders or old school versus new school

or real engineers versus vibe coders.

That framing is wrong. It's a scale, not

a divide. At one end, you've got the

fully AI native software engineers. They

write the PS and spec sheets in plain

English and they get AI to write the

code, review the code and test it. Their

craft has moved from typing to

specifying, architecting and managing

agents. And then at the other end, you

got the 100% handcoders. Every single

line typed, every character earned,

their craft is in their fingers along

with the discipline and pride that came

with learning how to code by hand. To

them, AI written code isn't real work.

It's unearned. It's cheating. And who's

in between those two? Most of the

industry. From developers using AI for

autocomplete to letting AI write most of

the code but still keeping an eye on

things. Some leaning towards more AI,

some leading towards less. This isn't

just my opinion, by the way. The data

backs it up. Google says 75% of their

new code is AI generated. Microsoft says

20 to 30%. And Anthropic's clawed code

teams say they're at 100%. All major

teams are now mostly AI assisted. But

then look at the developer sentiment

side. The Stack Overflow 2025 survey

showed that 84% of software engineers

were using AI or thinking about it, but

only 29% trusted. Trust actually dropped

11 points from 2024. That gap is where

the resistance lives. And I've been

collecting that resistance in the

comments of my videos. Do you understand

how insane this is? What sort of

product are you making?

Everything you said was just absolute

A tech bro attempts to grasp

software development. Typing code was

never a bottleneck. If you think so, you

know nothing about development. I am so

happy to read that you guys are having

such a hard time. That last one, read it

again. Happy you're having a hard time.

That's not a technical disagreement.

That's something else.

When you spent many years writing code,

you've built a self-image around the act

of typing. Your reflexes, your habits,

the way you think while you type. All of

it tuned to the keyboard. If I walk in

and say, "Stop writing code, write a

spec." Instead, I'm not asking you to

learn a new tool. I'm asking you to

change who you are. And that's hard for

anyone. One line that captures this

perfectly came out in MIT Technology

Review at the end of last year. It's

from an engineer called Luciano Noigen

who'd been leaning hard on AI coding

tools at work and then tried to build a

side project without them and realized

how dependent he'd become. I was feeling

so stupid because things that used to be

instant became manual, sometimes even

cumbersome.

Feeling stupid. That's what this

transition does to the muscle memory and

to the self-image built on it. The

technical term for what happens next is

cognitive dissonance. That happens when

your identity says, "I'm a craftsman

that writes every single line of code by

hand." And the world says that that

craft is being marginalized. When you

experience cognitive dissonance, your

brain has two options. Update your

self-image, which is painful, slow, and

requires humility, or reject the new

evidence, which is easier, immediate,

but isolating over time. Most people

pick option two first, and it's deeper

than psychology, actually. Your nervous

system biologically can't tell the

difference between transformation and

threat. Both fire the same alarm. When

something familiar starts ending, the

muscle memory, the discipline, the

identity that came with writing every

line by hand, your body reacts in the

same way as if something physically

dangerous was happening. Fight or flight

goes up. That's the resistance

underneath the resistance. Here's the

cleanest example of that that I've seen

from my own comments.

He is completely mistaken.

Specificationdriven development is not

the future. But I won't explain why.

Read that last bit again. But I won't

explain why. That's cognitive dissonance

in real time. He's not making an

argument. He's defending an identity.

The I won't explain why is the tell

because there isn't an argument

underneath. There's just discomfort. See

the traits that make someone brilliant

as an engineer. Persistence, focus, the

ability to hold a complex model in your

head are the same traits that make it

hard to change the model when the model

is what makes me valuable. The anger

isn't malice, it's grief dressed up as

opinion. And one of the other comments

deserves engaging with seriously because

it's a strongest handcoder argument that

I've heard. Typing the code was never a

bottleneck.

He's right. Typing the code was never

the only bottleneck, but the bottleneck

has moved. One of the hard parts used to

be typing the code. Now the hard part is

specifying what to build and reviewing

what the AI built. Those are different

skills. And if you don't develop them,

you're stuck. Not because typing is the

bottleneck, but because the bottleneck

moved and you didn't move with it. These

angry comments come from engineers who

are still standing where it used to be.

So, here's the main question I'm

grappling with. Why do some engineers

adapt and others dig in? I've watched

this inside my own team for three years.

The honest answer is in three parts.

One, AI tools have gotten dramatically

better over a short space of time. A

year ago, AI coding for production work

was quite poor most of the time.

Vitali's comments on my code 10 months

ago were technically justified. The code

was often bad. Opus 4.5 was decent. 4.6

was a step. 4.7 is something else

entirely. The engineers who tested AI in

2024 and concluded this isn't ready

weren't wrong then. But if you haven't

retested in the last 3 months, your

opinion is out of date. Two, identity.

The engineers who adapted the earliest

on my team weren't the youngest or the

smartest. They're the ones whose

self-image wasn't attached to the

typing. The ones who already saw

themselves as problem solvers first,

code writers second. Same skill set,

different identity, different outcome.

Three, humility. Not the soft version,

the technical version. The willingness

to be a beginner again at something

hard. There's a line I came across

recently that captures it well. Getting

better feels like getting worse at

first. Most senior engineers got senior

by being the smartest person in the

room. AI coding makes them wrong on

repeat until they're not. The ones who

could sit with that made the turn. The

ones who couldn't didn't. Rational

identity humility in that order.

Here's the thing. None of this is new.

Every digital transformation in living

memory has followed the same shape. Look

at photography. Until the early 2000s,

all photography was based on film. dark

rooms, chemicals, printing paper. I

watched this transition up close. My

father run a photography business. His

craft was developing the film, working

with the chemicals and exposing prints

in a dark room. Then digital came. He

stuck with what you knew for a while. He

was a master of the old craft. But his

competitors switched. They got faster.

They started winning more work. So one

day he made the call. He renewed the

whole lab. Within a month, he was

converted to digital. He set his

identity aside and adapted. But the

decision itself took him years. The

execution took weeks. He didn't switch

when digital arrived. He switched when

he saw it was the only way forward.

Today, most professional photography is

digital. Same with music. Vinyl passed a

billion dollars in revenue in the US

last year, the first time since 1983.

Real niche, real money, but not how the

mainstream listens. The pattern is

consistent. The old craft doesn't die,

it becomes a respected niche. The

mainstream production moves on. And the

legends of hand coding agree with me.

John Carmarmac, who founded ID Software

in the '90s and created games like Quake

and Doom, posted this earlier this year.

Coding was never the source of value,

and people shouldn't get overly attached

to it. Problem solving is the core

skill. If Carmarmac says it, the whole

you're not a real dev so you wouldn't

know claim doesn't quite land because

he's a real dev and he's saying the same

thing. Something has definitely shifted

over the last 10 months or so. And we're

not the only ones noticing it. DHH,

creator of Ruby on Rails, he was an AI

skeptic for years. Summer last year, he

said, "I can literally feel competence

draining out of my fingers." 6 months

later in January this year, he said,

"Just last summer, I spent more time

rewriting what I wrote than if I'd done

it from scratch. That has now flipped."

Those comments were only 6 months apart.

That's the same pattern as with Vitil,

just on a bigger stage. Let me bring

this back home to what actually happened

inside my team. 10 months ago, Vitily

wrote these three comments on an AI code

commit I made. This file is just awful.

Terrible turnary operator. This should

not be here. Real technical feedback. He

was right at the time. Then he retested.

This week, Vitili sent me an eight-page

proposal to build our internal agentic

coding system, a Slackbot for codebase

Q&A, linear ticket to merge request

automation, risk mitigations, branch

protection, mandatory human plan

approval. And this came entirely from

him. I didn't prompt him to do this. And

in his message, he wrote, "Those MRs he

used to send me that gave me plenty to

comment on. He named the moment he was

mocking me 10 months ago." He

acknowledged his own past position. He

updated his view when the evidence

changed. That's the story I'm seeing

play out everywhere. Most teams are

still where Vitaly was 10 months ago.

Quick word from today's video sponsor

before I get to the two paths. I'm

originally from Barcelona. I'm half

Spanish, half English. And I used to

work as a translator as a teenager. So,

this product hits close to home for me.

Palabra AI is realtime speech

translation that keeps your actual voice

60 plus languages translated in under a

second. Here, watch this.

Hello. How is everything?

Well, here I am speaking Spanish. My

voice is being translated into English

in real time. This is going to change

the world of translation forever. This

is incredible, isn't it? You can use

Palava for online meetings, broadcasts,

in-person events, or consume it directly

via their API. The majority of my dev

team is in Ukraine. Their English is

good, but sometimes they need to switch

to Ukrainian to explain something

properly. Balava solves that. And

normally with live events, if you need

translation, you'd have to pay like a

grand a day for each human interpreter.

This plugs in at a fraction of the cost

and does the job as well or better and

doesn't get tired. This is the future of

translation and this is why they're my

first sponsor. They've given me a 30%

discount for anyone watching and the

code is Axelmist 30. That's Axelmist all

one word then 30. Enter Axelmist 30 when

you sign up and the 30% comes straight

off. So scan the QR code or click the

link in the description and use the code

Axel 30 and start a 3-day free trial to

test it out yourself. Okay, back to the

video. The two paths.

So, here's where I think we're heading.

And to confirm, these are my estimates.

I think that within a year, 90% of all

production code will be written by AI.

Humans will be writing specs, reviewing

plans, managing agents, and approving

merge requests. Within 3 years, hand

coding will be considered a niche craft,

beloved, respected, and real, but not

how mainstream production happens, like

film photography or like vinyl. I could

be wrong about the timelines. Anthropic

CEO Dario said in March last year that

we'd hit 90% of all production code

written by AI within 6 months, and we're

still at around 42%. So, predictions

miss, but the direction we're going in

is unambiguous. I think there are two

likely paths forward from here. Path

one, adapt. Learn to write specs in

plain English. Get comfortable letting

AI write the code while you architect,

review, and manage. Find the new craft

inside the new workflow. Most mainstream

production work is going this way. Path

two, choose to be part of the hand

coding niche. If your love is the

keyboard, the typing, the craft of

writing every line yourself, that's

valid. There's a place for you. Just

understand that the place is getting

smaller. Both paths are honorable. Both

involve real work. Both let you keep

building things you're proud of. What

doesn't work is the third path. Denial.

Insisting the change isn't happening.

Attacking the people reporting it,

pretending the world will rewind. That

position has never won in any

technological transition. Not in

photography, not in music, not in

telecoms, and not in coding either.

Vitili chose path one 10 months ago. He

was the loudest skeptic on the team.

This week, he's the one leading the

build. He had the humility to update his

view when the evidence shifted. If

you're a founder or operator working

through the same transition in your dev

team or anywhere else in your business,

I write about it every week at

axelmolist.com.

Join my newsletter and subscribe for

more content like this. And pick your

path. Adapt or pick the niche. Both are

honorable. Just don't stand in the

middle defensive and angry at the people

who picked. See you in the next one.