keyosc.bsky.social
@keyosc.bsky.social
I also love how no one even has a good view defeating the purpose
October 28, 2025 at 6:20 AM
try Claude Code if you like! it's quite helpful for code
August 28, 2025 at 1:49 AM
Respectfully, this is a bit of a misconception :) llms are really no worse than netflix even after factoring in the cost of training
open.substack.com/pub/andymasl...
Using ChatGPT is not bad for the environment
And a plea to think seriously about climate change without getting distracted
open.substack.com
August 1, 2025 at 10:21 PM
Respectfully, it's a bit of a misconception that LLMs are any worse than, say, netflix. That's okay; there are plenty of other valid criticisms of them.
open.substack.com/pub/andymasl...
Using ChatGPT is not bad for the environment
And a plea to think seriously about climate change without getting distracted
open.substack.com
June 20, 2025 at 6:08 PM
This is a bit of a misconception :) but there are plenty of other valid criticisms of how LLMs are being used.
open.substack.com/pub/andymasl...
Using ChatGPT is not bad for the environment
And a plea to think seriously about climate change without getting distracted
open.substack.com
June 16, 2025 at 2:08 AM
Andy Masley who wrote this article, and that cheat sheet, claims to have amortized the cost of training into his calculations
open.substack.com/pub/andymasl...
Using ChatGPT is not bad for the environment
And a plea to think seriously about climate change without getting distracted
open.substack.com
June 2, 2025 at 10:59 PM
can you elaborate on how it's being used to harm people? genuinely curious
May 29, 2025 at 7:45 PM
Seems to line up with the one I posted yeah. They found 3300 joules ≈ 1 Wh for Llama 405B, while the first one calculated 3Wh for ChatGPT.
May 29, 2025 at 5:38 PM
The environmental cost is a bit of a misconception - it's really no worse than streaming netflix or anything else.
open.substack.com/pub/andymasl...
Using ChatGPT is not bad for the environment
And a plea to think seriously about climate change without getting distracted
open.substack.com
May 29, 2025 at 5:19 PM
the vast majority of information from most sources is discarded, just by nature of the size of the LLM vs. its training data. I agree that in an ideal world it shouldn't be able to exactly reproduce anything
May 29, 2025 at 4:51 PM
it's very much open to interpretation whether distilling and transforming a huge amount of information into a different form is in fact stealing. collage is fair use, after all
May 29, 2025 at 11:34 AM
I find it quite useful for coding
April 26, 2025 at 2:23 AM
I see them as power tools, which you need to know how to use, and won't accomplish anything (or make things worse) if you don't. In the right hands, an LLM is definitely a productivity boost - write this function, etc.
January 10, 2025 at 8:55 PM
The one time I went skiing, the nice trail lady had to give me a ride down the mountain on the back of her snowmobile because I was taking too long and the slope was closing
January 3, 2025 at 9:00 PM
Sound bars are not great speakers. Just get an external amp and a couple real speakers - they didn't have to be huge. The Topping brand makes great, compact amps - you don't need a huge receiver box like the old days.
January 3, 2025 at 2:09 AM
Someone tried to reply with another article, then blocked me. So this is my reply to them.
January 3, 2025 at 12:02 AM
(Reading menu) Are you willing to stand by your remarks?
January 2, 2025 at 11:16 PM
It says right there in the link you posted, "a single Google search takes 0.3 watt-hours of electricity, while a ChatGPT request takes 2.9 watt-hours."
January 2, 2025 at 11:12 PM
I'm not sure why you think you know more about this than me, who actually uses it, since you were just saying you weren't even aware of any good uses of it.
January 2, 2025 at 11:10 PM
Not really. A typical LLM query uses about 10 Google searches worth of energy. That's worth hours (not 15 minutes) of productivity.
www.goldmansachs.com/insights/art...
AI is poised to drive 160% increase in data center power demand
www.goldmansachs.com
January 2, 2025 at 11:09 PM
You realize you need the expertise to actually tell the LLM what to do. You don't just set it going like a wind up toy.
January 2, 2025 at 7:43 PM
You might as well tell a carpenter that power tools do their job.
January 2, 2025 at 7:37 PM
Huh? I do write code for a living, it's been my career for 10 years. I use LLM tools to speed up the process of writing some code snippets. They would be useless without my expertise of how to use them.
January 2, 2025 at 7:23 PM
A compatibility layer for steam that allows most games to run well. Check it out here. davidotek.github.io/protonup-qt/
ProtonUp-Qt
Install and manage Proton-GE and Luxtorpeda for Steam and Wine-GE for Lutris with this graphical user interface. Based on AUNaseef's ProtonUp, made with Python 3 and Qt 6.
davidotek.github.io
January 2, 2025 at 7:07 PM