My workflow used to be as most did agents in the command line, whenever bugs happened i tried to enforce verification layers, preferably deterministic to make sure it doesnt happen again. I dug deep into linting, biome and type rules. I tried to make the rules as tight as possible. I also thought about general design patterns, boundaries, enforcing these rules with file names and probabl some other enforcements.
For some of you that might be the obvious consequences and I don’t claim I’m right. I’m just writing down my feelings in that case as they’ve massively switched and I feel very disconnected to the world right now, reading everything that companies do, people post and write.
I have a huge distrust, and I understand that this is a skill issue, but the question still remains. When I don’t have the skill to build a good app, what is my best bet. To hope AI does it for me or to grind and learn?
Re Strauva I didn’t game it. I just used Claude Code everyday 12 hours approx.
Now I’m back in the idea and I sleep much better, I feel master of my codebase and my business domain again and most clearly have a tremendously better mental model about how things work and should work.
I utilize auto complete and auto next edit and I chat with AI in a browser window. I make it a habit to test every assumption fast (custom shortcuts in my idea (alt + gv = generate vitest, creates test file from every function, ctrl + t to run) AI throws at me and in new domains.This gives me a fast feedback loop to learn new stuff and makes it more likely I’ll remember.
Like have you seen cracked coders working, they are like fast apm coders and worlds more accurate then AI.
I try to learn stuff from the bottom up. After a while its faster for me to code it than to ask ai and wait for exploration and have to go through review iterations.
If I can describe the architecture/stack that I want for an app, most AI models can build the desktop app for me. You have to remain conscious of the tools you're using, but AI can produce great boilerplate for this type of rote programming.
> Like have you seen cracked coders working, they are like fast apm coders and worlds more accurate then AI.
I've worked with 10x engineers before, most of them were not Carmack-level geniuses. They simply understood their stack and stayed laser-focused on the goal. 10xers are impressive because they're specialists, and most of them go back to the 1-2x level if you force them to use technology they don't understand.
If you hate AI, don't use AI. If you aren't being forced by your employer, just don't use it.
If you do find a way to get AI to generate a "good app," that won't do anything for your own skill. It sounds like this concerns you. If you actually want to improve your skill you need to grind and learn. The only thing you'll learn using AI is how to use AI.
12 hours a day doing something you hate? That seems like an addiction. And it seems like more work than actually doing the work would be.
i am using --dangerously-skip-permissions with task spec. think this is faster. and it gave me more control actually over architecture and product decision. i think i just dont like reacting to suggestions mid-flow
It all went downhill from the moment they changed Reading *.* to reading (*) files.
I can’t use Claude Code at all anymore, not even for simple tasks. The output genuinely disgusts me. Like a friend who constantly stabs you in the back.
My favorite AI feature at the moment is the JetBrains predict next edit. It‘s so fast that I don’t lose attention and I’m still fully under control.
I haven't explored hash indexes enough. That might be one thing that differs between Postgres and MySQL, because for a long time Postgres didn't have a good story for those.
Like seeing orchestration pipelines, videos, tools or blog posts fills me with literal anxiety, like I feel bad for all these people. They seem like people on drugs to me that think they are on top of the world.
To be clear I’ve been like this as well and I learned my lesson. I’m also not saying that I’m right. These are just feelings that I experience, I could be the idiot.
Today I used AI to help code a feature and it worked pretty well. I am not doing any of this gaslight town stuff, and I went back about 4-5 times with it to make sure we had a mutual understanding -- it's a nice clean patch.
As of the end of the day there was still a bug left, there probably would have been a bug left if I did it myself. Tomorrow i will fix the bug, maybe with some help, and I am on to another ticket.
I treat Junie as a coding buddy (think pair programming) and I don't delude myself that 20 slaves are going to create the Great American Javascript while I sleep. AI coding makes my life better.
I haven’t tried Junie, but I’m using Jetbrains. First time I tried it it didn’t work lol (right at alpha).
I also see how AI makes my life better and I’m very grateful for everything that I could learn way faster through AI. Languages, systems, patterns etc..
My main experience is Claude Code and its suggested code changes..
This is probably driven to be more usable with AI agents, but smaller prs can create more code as they need to enforce more backwards compability, this can also lead to more code or more maintenance work.
Honestly I don’t see the benefit of smaller prs, except driving vanity scores?
Same here. I have now deleted 43k and counting lines of my codebase. There is no point in putting any AI code into production anymore as it almost always uses none or the wrong abstractions.
When you try to throw more agents at the problem or even more verification layer, you just kill your agility even if they would still be able to work
Case in point, just this morning I contributed a one-line change to an open source repo and the CI started failing.
I asked Claude (Opus High Effort) and pasted in all the logs. I went back and forth and it very confidently made over 20 separate changes in the repo, none of which fixed the issue. Eventually I stepped in and figured out it was a versioning issue.
I fear what would happen if I ran “10 agents for 10 days” on this simple issue.
