DoctorGester

@anastasiak2512 @kskrygan @JetBrainsRider Hard to tell because most of the time I work in the same Unity project.

@DoctorGester @kskrygan Does it happen with any project or with a specific one? What is the type of this project? @JetBrainsRider's support pls help investigate here

We saw this coming early — and stopped subsidizing AI back then. JetBrains has been 100% private from day one. No hype-helicopter money. While others chase VC-fueled hype, we focus on what actually matters: real value and playing the long game.

@anastasiak2512 @kskrygan This is a "normal" situation in Rider for me nowadays when switching between files or typing. I'm not sure what other diagnostic data I can provide.

@DoctorGester @kskrygan If you can share a bit of more details to us (here or in support), we will take a look.

@hd_nvim @mohsen____ LLM mediocrity in action, an actual typescript compiler architected from the ground up could be easily 10x tsgo which had to take a lot of tradeoffs. 1.26x with a full rewrite is a joke.

tsz is an unofficial TypeScript compiler in Rust by @mohsen____ It is now nearly compatible with TypeScript 6.0 and 1.26x faster than tsgo

@valigo @barryrambleson Intel macs were so ass compared to M series

@barryrambleson I had mbp mid-2015 (still have, actually), but it didn't show battery like this lol

I've been doing programming and web browsing on this laptop for 4 hours now, and the battery is still at 100%, what is this sorcery!! Linux and Windows laptops actually could never

@wxrlss @ParipariPanini @DannyMacFinn I do use AI quite a bit because I find it very good at reading code, just not writing. In certain smaller tasks even reading code faster yields a massive net gain. Thanks!

@DoctorGester @ParipariPanini @DannyMacFinn It’s ok man, you have time to learn how to use it efficiently if you ever want to! Agree to disagree, don’t want to argue as you obviously have your own strong opinion on this. Also love auto-battlers, your game looks good, wishlisted.

It's okay, and arguably good, to leave a negative review on steam games that use gen-AI or LLM's for their coding.

@wxrlss @ParipariPanini @DannyMacFinn That wasn’t even in the claim? You just inferred that because of your own insecurities. AI code is “incredibly good and fast” in two cases: 1. You have low standards 2. You told AI exactly what to do or fixed the code manually so much it mostly ceases to be AI code

@ParipariPanini @DoctorGester @DannyMacFinn AI generated code is not instantly “slop” or shit. I know you like to think that and it makes you feel special for not using it, but it can be incredibly good & fast if done correctly, which you can clearly see at any point of dev from Randy’s streams. And the game is amazing…

@getifyX Skill issue

fuck typescript and fuck anyone who condescends those who still prefer JS

@CallMeWhatNot @DannyMacFinn Say something even sillier, I’m writing it down

@DoctorGester @DannyMacFinn In the context of games, you're reshuffling old gameplay ideas. You're not being creative. Your skill trees and board game rules are not interesting. Game psychology in general is really lame and basic. If you don't know cinematography, architecture, human anatomy, don't do games

@CallMeWhatNot @DannyMacFinn Tbh this is not true about coding at all, there is infinite creativity. Code is just not user facing, both AI code and AI art are of shit quality rn and you most likely will notice when code is bad, but humans could have done it too. Art is just immediately apparent.

@DannyMacFinn Because even when you code manually, you're copying, there's limited creativity. It's on rails, an automated process, only a matter of time before 1+1=2. Art is different, a very deep human activity. If you hand it over to an AI, you might as well not bother creating at all.

@wxrlss @ParipariPanini @DannyMacFinn What’s wrong with their statement? I understand a lot about development and it’s true.

@ParipariPanini @DannyMacFinn Tell me you understand NOTHING about development without telling me

@RedBedDread @retr0techie @lonelysloth_sec Not really tbh, I’ve had codex push back on ideas (non aggressively but still, for example it says “I wouldn’t do it in the X way” all the time when listing options). And also I do know better most of the time.

@DoctorGester @retr0techie @lonelysloth_sec > and it said “this is much better”, It's an AI. Your answer could have been objectively worse and it still would have said "you did amazing. You're a wonderful smart cookie."

Claude routinely writes code that takes hours to run, when I look at it, there's something that should be O(logN) but Claude wrote an algo that is O(N^2). I prompt it with something like "your algo is too inefficient, improve it". After hours it comes back with an algo that is like O(N^3). Then I look at it for like a minute, figure it out. When I explain to it the correct algo it always blows Claude's "mind". "WOW! This is so much better." The more I use AI the more impressed I am with my own intelligence. Some day Claude will convince me I reached AGI.

@retr0techie @lonelysloth_sec That sounds inefficient given how quickly agent performance degrades with context fill

@DoctorGester @lonelysloth_sec I don’t use agent md files, I have it maintain about 500k tokens of documentation about large projects thats organized hierarchically, in subdirectories. And then have it read about 200-300k of them before working on any feature.

@retr0techie @lonelysloth_sec Game development: multiplayer programming, core engine systems, core gameplay systems. The only pitfalls is that it doesn’t have taste and doesn’t understand the big picture. This leads to different failures every time. You can’t fix that with agents md.

@DoctorGester @lonelysloth_sec Interesting. What’s your domain? Do you hand it extensive documentation and guides? Do you have it document common pitfalls so it doesn’t repeat them? Because for me these things make all the difference between it being completely useless and ~80% autonomous in prod.

@retr0techie @lonelysloth_sec This literally happens 90% of time. That’s why I don’t let codex write code unless it’s throwaway. It can read code fine, much better than opus, so I use it mostly for reading/research/investigation.

@DoctorGester @lonelysloth_sec Never had this happen before tbh. Usually what I see is that CC fails at optimizing something intelligently, like implementing DP properly, and then Codex just does it first try.

@retr0techie @lonelysloth_sec It is better, but it’s nowhere as good as you say. Just yesterday I had literally the same situation with it as OP, 5.5 wrote the dumbest code on earth, I told it which changes to me and it said “this is much better”, yeah no shit. And then it still tried to ruin it.

@lonelysloth_sec Try it. It’s in a totally different category compared to CC

@MisutaaAsriel @SebAaltonen Found a preserved version if you are curious wrike.github.io

@MisutaaAsriel @SebAaltonen Yeah, my game was fully canvas based

I am debugging why web app is so slow at 6K resolution. Native runs at 700 fps... Turns out that web page had a blur for the background (which was obviously not even visible). These UI blur filters are always extremely slow...

@MisutaaAsriel @SebAaltonen Ah, okay. This filtering only affects subsequent drawing operations. I used it to draw a blurred background with UI with more non blurred UI on top.

@MisutaaAsriel @SebAaltonen I’m assuming you can actually sample pixel data after filtering this way, with css you don’t.

@MisutaaAsriel @SebAaltonen Haven't tried, might just work. It was for a jam game I worked on and I only tested in chrome during the whole dev to then discover it only really works in chrome anyway because both safari and firefox appeared to run most of the canvas drawing in software.

@DoctorGester @SebAaltonen Appears it is an experimental feature flag you can enable starting in version 18. Though I question the point if it's just enabling the same filters CSS filters do, in which case why not just programmatically change the CSS filter?