262 karma · joined April 27, 2021
If you say “you are praf” in Romanian it means that you are f*cked, wasted, done for, etc.
Who am I? Senior engineer with 5 years of experience, working within the AI team.
Project: social platform with millions of MAU
Codebase: Typescript, NextJS frontend, express backend, Prisma ORM, monorepo . Hundreds of thousands of lines of code.
My expertise sits in the same domain, language, codebase that I apply AI to.
When do I use AI? All the time. It has significantly sped me up.
How do I use AI? Almost always in a targeted fashion. I line it up like a sniper rifle - I understand my problem, where in the codebase it should act in order to apply it, and what general structure I want. Then I create the prompt to get me there. Before, I needed to do the above anyway, minus the prompt, but then I needed to use my slow hands to code dozens or hundreds of lines of code to get there. Now, the AI does it for me, and it makes me much faster. To clarify, I do not plan everything in the prompt - I let the AI color between the lines, but I do give it the general direction. Of course, I also use AI for smaller tasks like writing SQL queries for example, and it’s saved tons of time there too.
In my side projects, I have a different approach - I vibe code my way through with less “sniper targeting”, because I care less about the beauty of the code organization and more about results. If I do see AI slop, I then ask the AI to clean it up, or redo the prompt with more targeting.
Overall, AI has significantly sped me up. I am still heavily involved as you can tell, but it is a phenomenal tool, and I can envision it taking over more and more of the code writing process, with the right rules in place (we are actively working on that at my work, with md files and rebuilding our FE from the ground up with AI in mind).
Does this sort of answer give you a better idea of how AI helps?
I don't see that as true. Your empathy always needs to be there, but your response is conditional on the situation. "I feel you, but..." and then offer a perspective that pushes the solution forward. Or even better, "I feel you, and..." to remove the defensive wall of your direct report.
I am blessed that I work in a really switched on team. We focus on keep the talent density high, and on open and transparent communication. Practically everyone in our 70 odd team is a killer, so I have not heard any word of complaint from my team members around other teams. But if I did, I definitely would try to fix it via the open and transparent communcation policy.
I highly recommend reading (or rereading) No rules rules..the culture at Netflix. That's the kind of culture that I strive for, and that I want to foster across the organizations that I work with.
What matters: - the model -> choose the SOTA (currently Claude 4 Opus). I use it mostly in Cursor. - the prompt: give it enough context to go by, reference files (especially the ones where it can start delving deeper from), be very clear in your intentions. Do bullet points.
- for a complex problem: ask it to break down its plan for you first. Then have a look to make sure it’s ok. If you need to change anything in your plan, now’s the time. Only then ask it to build the code.
- be patient: SOTA models currently aren’t very fast
I work at a company with millions of MAU, as well as do side projects - for the company, I do spend a bit more time checking and cleaning code, but lately with the new models less and less.
For my side projects, I just bang through with the flow above.
Good luck!
In the course of building my side project (storytveller), I’ve found that newer models tend to do worse at storytelling. I’ve tested basically every model under the sun that is available for use in production, and one stands out - now it may be that others will come along and not do the research that I’ve done to choose the same model, and thus my application will be better than theirs in part because of the AI model I choose.
Having “AI” as part of your application will not matter as much, that I agree with, but having “the right AI” will.
Of course,the user experience will definitely matter as well, as will the marketing and other criteria - another point of agreement with the article. But that does not diminish the fact that, if your application does not involve testing benchmarks, there is a good chance that a model that may not be the newest could still be the best for your particular use case, so you should not just blindly choose the latest shiniest model as this article sort of implies.
The hammer does matter.
But any self-respecting JS full stack dev would want to have a good level of familiarity with the first 4, so some level of experience with them is absolutely a +++
I appreciate the grind though
It is a paradigm shift, yes. And you will know less about the implementation at times, yes. But will you care when you can deploy things twice, three times, five times as fast as the person not using AI? No. And also, when you want to learn more about a specific bit of the AI written code, you can simply delve deep into it by asking the AI questions.
The AI right now may not be perfect, so yes you still need to know how to code. But in 5 years from now? Chances are you will go in your favorite app builder, state what you want, tweak what you get and you will get the product that you want, with maybe one dev making sure every once in a while that you’re not messing things up - maybe. So will new devs need to know high level programming languages? Possibly, but maybe not.
Redeploying takes time. Sometimes you want to disable something quick. Having a way to disable that feature without deploys is amazing in those cases.
That being said, there’s really no need to rely for a dedicated service for this. We use our in house crm, but we also have amplitude for more complex cases (like progressive rollout)
Thank you so much for sharing. It really makes me think what I’d do if I were in your shoes.
For me a big portion would be family, and helping it thrive. And putting my time and energy into some of the crazier ideas I have about how to help improve the world for the greatest number of people.
I wish you serendipity in trying to find yourself, deeply, and in discovering what you truly enjoy.
I’d also say don’t give up on love. The most important part of it all are the people you share your life with. So once you are in a good place, fill your life with them. Keep channels open with those that you know check that box, as you go through this journey of self-discovery that may require more time for yourself.
My free time is dedicated to projects that I believe have the potential to improve the world for the greatest number of people. I wrote a few books motivated by this, and then when I became a software engineer I build a few projects motivated by the same.
Examples include messaiah.ai, consciousness.social, multizoa.com, and dex.thesacred.xyz (though that one may not be functional anymore)
Not saying that they did the job - but that won't stop me from trying. Why I do it is a whole other discussion, but if I'm motivated by this, then there must be others, since I can't be THAT unique.
One of the reasons why I became a software engineer was to be able to bring to life projects that I believe have the potential to lead to "betterment for the average person," so...joke's on you :p