just now i was reviewing my own code and making claude do the same. it found an edge case i missed while i was pretty sure my code is perfect.
i felt pretty stupid and scared. stupid cus how could i miss such an obvious case? scared cus am i done for?
i have stopped writing code. most of the reviewing is also done by models(or tools). everything that i used to do is now being done by the models.
for the past 3 years, i have taken pride in writing how to code, how to write good code, how to write code that's readable by others, how to write code that's efficient, how to write code that is elegant and most of all, how to write code that works.
but all of it can now be done by claude and probably better.
is there nothing where im better at than a model? will there be no more senior engineers? would i have to change my career instead? should've prolly become a doctor like my mom used to tell me. ah that's too long though, what else?
there has to be something.
otherwise, i'm f*cked.
here is why i'm not:
writing code isn't typing random variables and functions and making sure it works. its not changing the button size from small to large. its not just knowing how to horizontally scale the software to accomodate more users.
its about understanding the user problems. why do they exist in the first place? what does an ideal solution look like for them? is this software used by old people a lot, if so i should probably increase the button size. is this software life or death for some people, if so i should probably scale before our app goes down.
engineering is decisioning based on understanding.
the implementation has now been extracted but im still working. maybe in a different form where i read customer interviews, think about their experience, design solutions, see which one feels better, let the agent rip apart solutions, understand what it did and roll out the features after testing.
yeah prolly a few more of these functions will be automated in the next few months. there is however something which still saves us:
proximity.
there is a layer before the customer interviews, the linear ticket, the complaint. its presence. the one where just by being besides your users, you get to feel what they feel. which feature annoyed them? what's a page people share with their friends on social media? why are they not sharing it with their friends? is there a problem?
there isn't but there is. there is a problem but you need to have an eye out for it. you need to be present and on the lookout. that's exactly why engineers would always exist.
the kind of engineers that solve problems. if you are too attached to the identity of writing code, there is still some time before it completely takes you down.
you don't need to bucket yourself as a coder.
you don't need to bucket yourself as a designer.
you don't need to bucket yourself as customer support.
you can be anything you want.
try designing a website without knowing anything about figma or colors or fonts. spend a lot of time in it. genuinely design for something that you face. you will see how you can craft beautiful solutions because you can empatheize with the problem.
most of all, get into the habit of learning.
yes knowledge is v easily accessible.
yes chatgpt knows everything.
however, there is still a lot of alpha in the ability to learn. continually learning capability is probably gonna make a lot of progress in models to make them more like us. we should not digress and become models instead.
imma be honest. knowing all of this, it still scares me some days. what will the future look like where anyone can build things? would they even need me? am i just pretending to be good at something that will go away in a couple of years?
yeah i don't know. man, i'll figure it out. that's what engineers do. they figure it out.
you can get a headstart in knowing these answers and always be riding these waves if you keep learning. if you are reading this blog, you are prolly already in the most updated ones about what's going on in the world of AI models.
there is no use worrying though. what can you do instead as an engineer?
- learn how to orchestrate agents better?
- learn how to give proper context/ communicate better?
- constantly try and solve your own problems
- build apps for your family and friends
- read what frontier labs are writing about, if you are not already in them
- preferably try to get into a really good company(good doesn't mean big, think cursor, claude, deepmind)
- constantly learn how to get better and faster in what you are doing
- take out time every week(if not every day) to learn SOTA techniques
above all of them, still follow your curiosity. even if SOTA models are smarter than you, know more than you. follow your curiosity. ask questions. learn things. get better at speaking. create more. consume less.
that's all i guess.
jan 5, a junior engineer vs claude code