since the advent of gpt, intelligence has been highly accessible. you have a tool at hand which has phd-level intelligence in any and all subject matter you can think of. where does that leave intelligence? if not intelligence, what is the quality that decides whether an individual would make it(financially) or not? is it just luck? is it just courage? is there any single quality that would contribute to it the most?
i think so.
agency.
agency is defined as the personality trait which makes an individual believe that everything that’s happening to them is a result of their past/present actions(instead of external factors).
if you are high agency:
- you would shoot your shot
- you would never think you are the victim
- you would ask for help instead of waiting around
- you would proactively try to change things instead of complaining
on the contrary, if you are low agency:
- you’d complain about how life is unfair
- you would wait for things to happen to you
- you would bicker, complain and blame it on fate
- you would not take actions and spend all the time planning
this is also very closely associated to developer productivity in 2025. if you are someone who is low agency, you’d delay learning new tools because you don’t “believe in them” or “your way is the high way” and be left far behind in a couple of years.
you can learn any framework, any tool, any language, any principle at a
- basic level < 2 weeks
- intermediate level < 2 months
- top 1% < 6 months
if you do not believe so, that’s another sign of low agency. not having faith in the ability to do something because being unable to do something is a very fair excuse.
we are in 2025, an era where anyone can do anything. designers can code, marketers can design, coders can market and the only thing that’s stopping you to do anything is your brain. the brain which often lies to you to keep you inside your comfort zone.
vibe-coding is a term growing in popularity. its the art of coding with the help of LLMs while having very little to no idea of what’s happening under the hood.
engineers are known to usually frown upon vibe coders because they do not understand what is happening under the hood. hypocritically so, most of the engineers(including myself) also do not understand how the code i write gets converted to machine language and the specific set of instructions that follow and execute the code i’m writing in my IDE.
human beings hate changes. human beings also love to feel superior. both of these things come into the picture when as engineers, the code that’s closest to our heart is being written by a model. it takes away the pride and the ego of feeling smart just because we could produce something just by typing characters while others couldn’t.
but that’s okay.
the very thing that also makes engineers good is the fact that we are used to learning something new every other week/month/job switch and we will learn, accept and come to terms with vibe coding as well.
now if you are a coder, stop reading. that was all for you.
you have an idea but what’s next? how do you actually convert it to a working product?
i’m going to explain to you few simple principles and soft skills to learn how to vibe code that ACTUALLY works.
1. planning: strictly start your first prompt to build any app, feature with “don’t code”, blabber everything you want from the app, the more the better( voice) and towards the end just ask the tool to research and create a plan(phases). go through the plan, suggest changes for each small phase.
2. context: you are not supposed to know the documentation but ai agents do. for every piece of software mentioned, ask the coding tool to “get relevant info from official documentation”.
3. implementation: you have the phases already from 1, now start asking the model to build one phase at a time, if possible test out those singular phases. if not, take it to the next possible testing stage and test it out.
4. bug finding: mention the thing that’s failing, do NOT ask it to solve it just yet. ask it to “research all possible sources from this error and add logs to pin point, make your recommendation” to find why something fails.
5. bug solving: provide every log that you have from frontend, backend and dev tools from browser(cmd+shift+i in mac to see what requests the app is making what are failing) to solve the bug. if it is unable to do so, do NOT prolong the conversation. end it, new thread: repeat.
6. git: create a new github repo from instructions on official page, “git add .” to add all files, “git commit -m {message}” + “git push” to save it on github. that’s all you need to know for the time being. its a lifesaver when ai messes things up. use the graphical interface and not cli.
soft skills:
1. perseverance: iterate, iterate and iterate. there are no shortcuts, you are not going to get the app up and running in the first attempt. do not aim for that.
2. curiosity: be curious about how pieces work together. what failure would lead to what breakage. its always better to have an idea about what’s really happening.
3. low ego: ask for help. the internet is filled with people who want to help you(dm me). try to figure things out but also be ready to ask for help
that’s all there is. don’t give up before you’ve actually spent enough time persevering, being curious and low ego-ing. you can not fail. tools have gotten good enough. you can do it. just
keep goingggg.