Why I Don’t Have Enjoyable With Claude Code • Stephen Brennan

Stephen Brennan • 23 January 2026

LLM-based coding brokers like Claude Code & Codex are all the craze proper now.
Rightfully so, as a result of these instruments are literally getting good. They’re really
on the level the place individuals, each programmers and fewer technical customers, can use
them to create options and even complete initiatives with first rate outcomes. I’ve a
lot of emotions that I can’t cowl in a single weblog put up, however one factor seems like
it’s changing into clear to me: I’ll seemingly by no means love a instrument like Claude Code,
even when I do use it, as a result of I worth the duty it automates.

Whenever I take advantage of Claude Code, I discover that I cease having enjoyable making software program.
That’s attention-grabbing, as a result of many individuals report the other: that coding brokers
make computer systems enjoyable once more! I imagine that this is because of a distinction in values.

We Automate Tasks We Don’t Value

Like different applied sciences, AI coding instruments assist us automate duties: particularly,
those we don’t worth. I take advantage of my dishwasher as a result of I don’t worth the method
of hand-washing dishes. I solely worth the top consequence: clear dishes. Fabric is
created with mechanical looms & knitting machines, as a result of the financial worth is
the ensuing material, not the method of making it. Yet I nonetheless crochet & knit
some gadgets by hand, as a result of I do benefit from the course of.

People who love utilizing AI to create software program are loving it as a result of they don’t
worth the act of making & understanding the software program. At least, they worth it
far lower than the top consequence. I believe that’s fairly a traditional view of software program &
computing: it’s a means to an finish. But to me, creating & understanding
software program is a worthwhile pursuit by itself! I benefit from the means of representing
an issue in code, and I take pleasure in studying & constructing a psychological mannequin of programs so
that I can higher perceive and debug them. The ensuing software program product might
have worth, nevertheless it’s not the one worth, and even the first worth to me. Put
merely, I’m not a “product-focused” developer.

Now this isn’t to say that I worth writing all code. Plenty of code is boring!
I gained’t profit a lot from writing a whole bunch of strains of boilerplate code. And
it’s to not say that I don’t worth the product, or care deeply about designing
software program that solves issues! But the rationale I bought into software program, and the
cause I proceed to do it, is that I simply actually like computer systems and need to
be taught extra about them. The present zeitgeist of AI coding is to make use of the AI to
do as a lot as potential as rapidly as potential, and for me that simply throws out
the infant with the bathwater.

For the product-focused individuals on the market, I’m positive a instrument like Claude Code is a
godsend. Finally, you’ll be able to inform a system what you need and get a consequence, give
suggestions, and iterate. You can nonetheless make some technical selections, however in a
sense you develop into a “manager” of AI programs. You communicate by way of outcomes and
necessities, and also you let your underlings deal with the main points. If the one worth
being created is the top consequence, then that completely is smart. But for me, it’s
simply not the case.

Becoming More Honest About Goals

As a results of this, I’m beginning to be extra aware about my targets when it
involves software program. When I crochet or knit one thing, my purpose will not be simply to
get a blanket, stuffed animal, scarf, and so on. I might rapidly purchase these issues
cheaply. The purpose is to create one thing with my palms and my time. The worth
of the article, for myself or particularly as a present, is within the time and care I put
in whereas making it. I’m beginning to apply that mentality to my targets about
software program.

I used to say that I wished to “make X” for some worth of X. But often I would like
to construct that X in order that I can be taught one thing. Maybe I need to perceive the
drawback higher. Or possibly I need to begin a venture in a brand new programming language
so I can be taught it. I’m discovering it extra useful to acknowledge these issues as
a part of the purpose, and discern the extent to which I care in regards to the course of
versus the consequence.

This is useful as a result of once I’m trustworthy about my purpose, I can select the suitable
course of to truly get what I would like. There are undoubtedly duties the place I simply
worth the consequence, not the method. I can be taught & use these AI instruments in these
areas, leaving myself extra time for the elements of the method I do take pleasure in. I’ve
all the time possessed the laziness mentioned to be a
virtue of programmers. It’s a simple promote if I
can automate one thing that I genuinely don’t need to be doing! But if I
really need to be taught one thing, I’ve to do it the arduous approach.

Will I Get Replaced By AI?

That’s nice for me, however what about my profession? While I could worth the method of
coding and the training I get from it, the actual world tends to be chilly and
uncaring. The economic system might not worth my studying (not less than within the short-term),
and employers are turning an increasing number of to AI to “increase productivity.” It
can be silly for me to imagine that AI gained’t change our business. That mentioned,
I believe there’s plenty of nuance as to whether skilled software program engineers are
really at risk, and I believe that this query of values & targets immediately ties
into that nuance. The worth of software program improvement, even at large corporations, is
not the output alone.

My job will not be really to put in writing code. I’m employed to repair buyer bugs in
Linux. I learn plenty of code. I write code to assist diagnose, and in the end repair
these bugs. I write code to breed and get extra accustomed to some bugs. I
work on debuggers and associated tooling, in an effort to create extra highly effective instruments to
assist me in debugging. I give you methods to debug points whereas respecting the
(often onerous) constraints of consumers.

So once I write code, the trick isn’t often understanding write the code, however
understanding what the characteristic ought to be. Even when the coding job itself is
easy, I nonetheless discover I get worth out of it. I’m all the time constructing my
information and expertise so I can do the job higher. While there could also be loads
of software program the place it doesn’t matter that an LLM wrote it, I do suppose that there
will proceed to be big swaths of software program the place it does matter for a human
to put in writing it. There is a lot worth (information & experience) generated by that
course of, which is extremely vital for programs that want reliability and
debuggability. Ultimately, I don’t need my pc’s OS to be vibe-coded, nor
my financial institution’s programs, nor my automotive software program.

I do know higher than to prognosticate, and I undoubtedly don’t suppose anyone
ought to belief my opinion on whether or not my very own job can be eradicated. But I’d guess
that it will depend on the extent my targets & values align with the work that I do.
I’ve been fortunate sufficient to discover a job that aligns with my values: it’s about
technical experience; understanding & debugging programs. So lengthy as that
alignment holds, I’m cautiously optimistic.

---