Replying to Spooky Phijkchu #spooky.131 Max, do you think he pays $5/month for this? (edited)
no
rawb
Saturday, January 31, 2026
I need someone with an opinion to weigh in
do you like when transcriptions have uh's and um's in them
I was watching it around this timestamp, and was going back and forth
(I'm changing up my remote runner setup so I can actually process transcriptions faster)
Replying to Spooky Phijkchu #spooky.131 Click to see attachment 🖼️
Like here, it should be included
interesting, I was leaning towards the exact opposite
but again, idk, hence why I'm asking
I'm leaning towards,
Remove:
- Um's
- Uh's
- Oh's
- Ah's
- Like's
I think a 1 to 1 transcription even if there's a lot of um's and uh's is better, because in my mind the transcription is for having written what was said, so having that makes the most sense to me
Replying to Spooky Phijkchu #spooky.131 I'm leaning towards, Remove:Keep
- Um's
- Uh's
(edited)
- Oh's
- Ah's
- Like's
completely agree that Oh's, Ah's, and Like's should not be removed
Replying to legorhin #jazardly does a stenographer keep ums?
Typically yes, but there are exceptions
Sunday, February 1, 2026
Replying to Spooky Phijkchu #spooky.131 Polar, I'm not shilling when I say this I really don't think many people appreciate how overpowered AI are at coding now
my company has an internal version of ChatGPT and i gotta be honest its terrible. i know the latest gemini and Claude are much better than GPT 5, but Ill boil my problems down to the most leetcode-y type problem i can and at best take like a quarter of its code and do the rest my self.
Our internal Open-source GPT-knock-off at work sucks too, we've been lobbying for a while to bring Anthropic on-prem, but no luck thus far
I've been generally pretty impressed at the coding up-tick since the last major models released in October/November
they're not better than people who know the codebase and work in it every day or anything
but I've found they're able to debug and self-test pretty efficiently
Replying to Spooky Phijkchu #spooky.131 I've been generally pretty impressed at the coding up-tick since the last major models released in October/November they're not better than people who know the codebase and work in it every day or anything
i basically only ask it questions on writing efficent pandas dataframe manipulations. it doesnt know shit about lesser-known librarys even when i include the docs in the prompt
I hate that, it feels like a lot of LLMs "cost save" by just regurgitating what it thinks the answer you're looking for is
rather than actually doing and computational work
That's one of the ways clueless students using AI to cheat on tests get caught, when it's giving you the straight Quora answer that it determined was "close enough" to what you were probably asking
Replying to Spooky Phijkchu #spooky.131 I hate that, it feels like a lot of LLMs "cost save" by just regurgitating what it thinks the answer you're looking for is (edited)
it also gets wayyy to verbose and adds unnecessary lines of code. some times i ask it for an algorthm. i have a general idea that it shouldnt be more than ~20ish lines. maybe 40 if you really over comment everything. but it gives me 100 lines and im just like, there is no way this is effecient if it even works
"Hey, I refactored all of your code to add in unnecessary formatting and break the flow of all of your functions"
Replying to Spooky Phijkchu #spooky.131 "Hey, I refactored all of your code to add in unnecessary formatting and break the flow of all of your functions" (edited)
0 chance im letting it touch any written code. maybe next time im bored and use claud for a personal project
oh 100%, introducing an AI into old code in your IDE is just asking for trouble
you need to clean fork anything you're going to start sloppifying
And where would we be if 90% of the projects on Github weren't drowning in random emoji spam
Replying to Spooky Phijkchu #spooky.131 And where would we be if 90% of the projects on Github weren't drowning in random emoji spam
also wouldnt mind if there was an AI bot out there that didnt make Yes Man from New Vegas look like a contrarian
99->97
Without going back to listen to the actual song, it does sound off to me
at a glance
This is the .WAV from bandcamp
the cymbal at the start if nothing else, is just comletely different
wait, I listened later into your version Max, that's completely different
there's a whole solo added in around 2:10
I thought he gave that up
guess he's back in the mines
I wonder if this is just an alt-stem version his brother made for him that he didn't use
or if he AI removed the vocals from the actual song, and this is what it re-created
Monday, February 2, 2026
For Pika
I was checking in on a random caption generation with some new settings, and randomly scrubbed to this
Tuesday, February 3, 2026
Replying to Spooky Phijkchu #spooky.131 Click to see attachment 🖼️
Forgot to watch this until now. Lmao
that part of the video starts at 2:12, after the in-game teaser
this video at the timestamp is from the Kickstarter launch, and when he's all excited and bouncing off the walls
also
it's a little weird remembering that Rob is our age in this video (24)