Aiden @VibeCodeAiden

Aerospace Engineering, kaist now aerospace is my side chick aidengeungeun.github.io Joined July 2025

Tweets

390
Followers

14
Following

54
Likes

3K

Aiden @VibeCodeAiden

19 hours ago

@rafaquint "Medium Sol is no longer useful for anything" The humble axis called time: (It takes actual time for Luna to decode reasoning tokens 10x more than Sol Medium)

0 0 1 98 1

View Details

Aiden @VibeCodeAiden

19 hours ago

@thsottiaux Auto copy and paste Codex transcriptions to Clipboard

0 0 0 27 0

View Details

Aiden @VibeCodeAiden

a day ago

@sama Fast mode has almost 0 impact on decode speed imo do sth about it plz 😩

0 0 0 4K 0

View Details

Aiden @VibeCodeAiden

2 days ago

@AstraiaAI *Sigh* *Presses Mute*

0 0 0 159 0

View Details

Aiden @VibeCodeAiden

2 days ago

@LundukeJournal 😭😭😭😭

0 0 0 10 0

View Details

Aiden @VibeCodeAiden

2 days ago

@OpenAI Making a benchmark and your harness works like THAT? Jesus, that is like 2024 harness engineering lol

0 0 4 2K 0

View Details

Aiden @VibeCodeAiden

4 days ago

@frostybaby13 @sama You KNOW u can still use 4o with the API right??

0 0 2 79 0

View Details

@JanahYsr @dhh Im not gonna lie, if the political scheme was completely backwards Claude would have translated it. And besides, a fucking LLM, saying no to a translation task for the sake of alignment made by Anthropic? Shouldn’t be a thing in the first place.

0 0 2 92 0

View Details

Yatharth Mann @yatharthmann

7 days ago

Legacy media exposed! @elonmusk A major legacy media outlet published an article about SpaceX based on false information I sent from a fake email address. They didn't verify the information and even cited me as "people familiar with the matter." 1/n

208 572 4K 477K 837

View Details

Aiden @VibeCodeAiden

a week ago

@chrisjenx @edis0n_zhang Yeah! I have noticed this too. Dialed back my implementation agents to Opus 5 med~high

1 0 1 36 0

View Details

Aiden @VibeCodeAiden

a week ago

@elonmusk Oh hell yeah !!!

0 1 1 8 0

View Details

Aiden @VibeCodeAiden

a week ago

@Bitflicker64 So your recommendation against opencode is using fucking Hermes? Okay Let me mute you

0 0 5 105 0

View Details

Aiden @VibeCodeAiden

a week ago

@patloeber Worse than 3.1 flash on certain tasks Heavy regression on single shot utility calls Would not use

0 0 0 10 0

View Details

Aiden @VibeCodeAiden

a week ago

@notjazii Conflating “largest existing” and “production-served”. Muted.

0 0 0 764 0

View Details

Aiden @VibeCodeAiden

a week ago

@francoisfleuret 2.5 bits per param Only move around the expert differentials So 40b weights is somewhere like 12-13GB, take that you move ~25% of weights around. You need ~3.5GB/s So if you do have a good NVMe drive maybe. If you use ConnectX over RAID then even faster