@kritikakodes I had the same problem, not anymore because funny enough I built a tool for it. ideacheck is a cli tool that uses LLM to reason and jev to judge the idea so you know if it is worth building or not!
https://t.co/XxxKLuLpbE
Dailies (5.6)
I do not want to be a software engineer anymore.
No, seriously, this is exhausting!
We thought AI would make our jobs easier; I'd argue it's made them 10x harder.
How did I spend 10 hours today just building and testing a CLI tool? And the reason I built it was to stop myself from building things that might have no value.
Funny enough now that it's done, I suddenly have no new ideas to test it with :*
I also want to talk about how I miss the old days of coding. I know we all do, but can we talk about it for a bit?
Sure, AI writes code faster. It researches, finds, installs, plans, and builds things faster, and sometimes better. But I still had to make it rewrite ideacheck's code many times before I got a result I was happy with.
The point I'm trying to make here is this:
"with great power comes great responsibilities"
Writing code fast isn't a blessing; it's a curse. Because you know you can write more, you start trying things out and wandering into spaces you'd never have indulged in without AI. And you never feel fulfilled, because it all happens so fast, and most of the time you're left with a sense of imposter syndrome.
I tried Laya today, rewrote the TUI a couple of times, and still felt like something was missing. Maybe it needs a GUI frontend? should I use kev? is there a better way to hand out context between the judge and the researcher? But what if nobody uses this tool? What did I gain by building it? etc.
Back when coding was manual, even if a project didn't work out, you learned something: a new library, a new language, a new way of solving a problem.
Today it feels heavier, because we're not only spending time, we're spending money too. It looks easier, but it's actually far harder. The wow factor is gone, and you end up with time and money spent and very little experience to show for it.
And time value is something that is subjective; time matters in comparison of what can you do with it. If you have all the time in the world but you can only do few things then that time probably doesn't worth that much does it? However the opposite is true! If you have limited time and you can do so much with that limited time, then you better go all in to try and get the best out of it else you feel unfulfilled.
The latter is how I feel with my time the past few months.
The amount of things I can possibly build is many, but I only have 24hrs a day! after cutting the parts that fuels my mind and body, I only get what 8? 10 hrs at best.
So I feel like I shouldn't be building just whatever feels good, shouldn't build things just because they are trendy, and for sure shouldn't build what I won't use!
That's why I will be the first user for my tool (ideacheck). I will use it every time I get an idea to score it before building it.
Only building the ideas that score 7 or more out of 10.
**
ideacheck is an open-source tool I built that uses LLMs and open-system style model (jev) to score and rank your idea so you know if its worth.
**
Thank you for giving the time to read this post, I hope you got value and it was worth! if so give feedback, maybe tap that heart icon to let me know that you read the whole thing! see you tomorrow.
@KeithZhai yeah well the point is that the default search is horrible like search in an agent loop with the search tool. unless you use a tool made specifically for search. haven't heard of tiny fish is it a tool you are building? and how different it is than something like searchxng?
Daily writings.
Day rating (5.6) worse than yesterday.
Learned:
> Don't use an agent loop for research. One simple search in ideacheck took 70 seconds and about 146k tokens. A search API like SearXNG would cost about 10x less and finish in a fraction of the time.
> Building a native app burns my Mac's CPU and 16GB of RAM. Expo's EAS cloud build takes that load off it. however i have trust issues to share my apple id with them so I'm going to burn my mac to build the gym app.
Did
- Used jev on the ideacheck app.
- Finished reading a hizb of Quran.
Didn't do
- The Telegram scaling video script is still unfinished.
- Still don't understand how distribution works.
- No workout, no reading, and I slept and woke late again.
Next: try the pi harness and try remocn.