So many things to write about. I almost don't know where to start. I'll start with my coffee. I did planning just before this, so the coffee is now 10 minutes old and perfect temperature for drinking. [Takes another sip].
Let's get to some of the juicier thoughts. Yesterday I spent some time promt engineering. I have mixed feelings and some good learnings already. This morning I thought about yesterday when it occured to me that promt engineering has a sort of 'rubber duck' paradigm built in to the process.
I found that the better I described what I wanted or what the problem was, the better the result. I think this is kind of obvious in hindsight, but as an engineer it also gave me chance to skip a step or two. This happened automatically for me after a while. As I was describing, I would do a quick lookup of a variable, or maybe glance at the methods for context. I would also solve some of the problems by desribing how I thought we should solve it. This way of promting gave much better results.
I was also quite surprised at how easy it was to separate things into multiple files. Cursor doesnt feel the need to do this, it is fine with just throwing everything into a single file. This might be fine for a machine, but I needed to get a better overview. It was also easier for me to debug and understand the changes it was suggesting to me. Factorising is a real pain when building sites and apps, and having Cursor do this for me with a single request was quite remarkable.
I also started reading what it was telling me after a while. In the beginning I was just accepting anything it suggested. This worked well when there wasnt much code, but I ended up not really understanding what was happening. I guess you could say it was fun while it lasted. In the end, reading the results and sometimes saying that you don't think the suggestion was good, or even broken helped the process.
So what is my conclusion? Is Cursor worth it? I think it will be in a very short while. The great thing is that I can switch back and forth between coding and having cursor do the tasks I either don't feel like doing or know how to do. I'm looking forward to seeing how my app turns out.
- Trolz
Let's get to some of the juicier thoughts. Yesterday I spent some time promt engineering. I have mixed feelings and some good learnings already. This morning I thought about yesterday when it occured to me that promt engineering has a sort of 'rubber duck' paradigm built in to the process.
I found that the better I described what I wanted or what the problem was, the better the result. I think this is kind of obvious in hindsight, but as an engineer it also gave me chance to skip a step or two. This happened automatically for me after a while. As I was describing, I would do a quick lookup of a variable, or maybe glance at the methods for context. I would also solve some of the problems by desribing how I thought we should solve it. This way of promting gave much better results.
I was also quite surprised at how easy it was to separate things into multiple files. Cursor doesnt feel the need to do this, it is fine with just throwing everything into a single file. This might be fine for a machine, but I needed to get a better overview. It was also easier for me to debug and understand the changes it was suggesting to me. Factorising is a real pain when building sites and apps, and having Cursor do this for me with a single request was quite remarkable.
I also started reading what it was telling me after a while. In the beginning I was just accepting anything it suggested. This worked well when there wasnt much code, but I ended up not really understanding what was happening. I guess you could say it was fun while it lasted. In the end, reading the results and sometimes saying that you don't think the suggestion was good, or even broken helped the process.
So what is my conclusion? Is Cursor worth it? I think it will be in a very short while. The great thing is that I can switch back and forth between coding and having cursor do the tasks I either don't feel like doing or know how to do. I'm looking forward to seeing how my app turns out.
- Trolz