Rocks for Jocks
@rocksforjocks.bsky.social
Science podcast with a focus on climate and natural hazards! Available on all platforms.
linktr.ee/RocksForJocks
linktr.ee/RocksForJocks
Why Does Science Matter?
I started with a box of rocks. So why should you care about science?
open.substack.com
November 30, 2025 at 5:55 PM
Looks fantastic! I've been trying to convince my fiancée we should go there for our honeymoon. Maybe this video will tip the scales!
October 13, 2025 at 3:19 PM
Looks fantastic! I've been trying to convince my fiancée we should go there for our honeymoon. Maybe this video will tip the scales!
Really excited to hear about the rocks this season, thanks for sharing, Professor Baxter!
October 13, 2025 at 2:59 PM
Really excited to hear about the rocks this season, thanks for sharing, Professor Baxter!
...Syntax knowledge is already a pretty small deal for SWEs to know about, and specific syntax is not discussed frequently in interviews. The focus is generally geared much closer to understanding ideas rather than parts of a specific language.
(end of rant)
(end of rant)
July 27, 2025 at 8:39 PM
...Syntax knowledge is already a pretty small deal for SWEs to know about, and specific syntax is not discussed frequently in interviews. The focus is generally geared much closer to understanding ideas rather than parts of a specific language.
(end of rant)
(end of rant)
..."Firing all software engineers (SWEs) and data teams because AI can code everything" would only come about if everyone could perfectly explain what they want code to do. (In which case, you are essentially a SWE). The only difference at that point is SWEs know syntax (programming languages)...
July 27, 2025 at 8:38 PM
..."Firing all software engineers (SWEs) and data teams because AI can code everything" would only come about if everyone could perfectly explain what they want code to do. (In which case, you are essentially a SWE). The only difference at that point is SWEs know syntax (programming languages)...
...Would it be a LOT faster if I had just coded it myself at that point? Yes. Would it help me understand and help me fix things in the future IF I coded it myself? Absolutely!
Lesson here: Human coders still have a huge advantage over AI...
Lesson here: Human coders still have a huge advantage over AI...
July 27, 2025 at 8:33 PM
...Would it be a LOT faster if I had just coded it myself at that point? Yes. Would it help me understand and help me fix things in the future IF I coded it myself? Absolutely!
Lesson here: Human coders still have a huge advantage over AI...
Lesson here: Human coders still have a huge advantage over AI...
...so theoretically if you kept asking it to make a scatterplot you liked, an LLM might eventually get it right. Kind of similar to the way infinite monkeys with typewriters would eventually write Shakespeare.
Is it possible to live this way? Sure...
Is it possible to live this way? Sure...
July 27, 2025 at 8:33 PM
...so theoretically if you kept asking it to make a scatterplot you liked, an LLM might eventually get it right. Kind of similar to the way infinite monkeys with typewriters would eventually write Shakespeare.
Is it possible to live this way? Sure...
Is it possible to live this way? Sure...
And I know one possible solution is to retype the prompt in a slightly different way. Give it more details, more guidance, ask the AI to try over and over again until they get it right.
Most LLMs give a slightly different output, even when asked the same question multiple times...
Most LLMs give a slightly different output, even when asked the same question multiple times...
July 27, 2025 at 8:31 PM
And I know one possible solution is to retype the prompt in a slightly different way. Give it more details, more guidance, ask the AI to try over and over again until they get it right.
Most LLMs give a slightly different output, even when asked the same question multiple times...
Most LLMs give a slightly different output, even when asked the same question multiple times...
**BUT what ChatGPT actually did**: change all legend colors to Blue and then hide the now "misleading" legend.
In other words, it ruined the plot, while technically fulfilling the goal of not filtering to different values.
In other words, it ruined the plot, while technically fulfilling the goal of not filtering to different values.
July 27, 2025 at 8:30 PM
**BUT what ChatGPT actually did**: change all legend colors to Blue and then hide the now "misleading" legend.
In other words, it ruined the plot, while technically fulfilling the goal of not filtering to different values.
In other words, it ruined the plot, while technically fulfilling the goal of not filtering to different values.
@sydneydemets.bsky.social thanks a bunch for helping out with this episode!
* Spotify: tinyurl.com/SpotifyDeMets
* Apple: tinyurl.com/AppleDeMets
* YouTube: youtu.be/mWxHEGRQalI
* Spotify: tinyurl.com/SpotifyDeMets
* Apple: tinyurl.com/AppleDeMets
* YouTube: youtu.be/mWxHEGRQalI
The Science of Podcasts with Sydney DeMets, PhD Candidate
Podcast Episode · Rocks for Jocks · 06/16/2025 · 50m
tinyurl.com
June 22, 2025 at 3:17 PM
@sydneydemets.bsky.social thanks a bunch for helping out with this episode!
* Spotify: tinyurl.com/SpotifyDeMets
* Apple: tinyurl.com/AppleDeMets
* YouTube: youtu.be/mWxHEGRQalI
* Spotify: tinyurl.com/SpotifyDeMets
* Apple: tinyurl.com/AppleDeMets
* YouTube: youtu.be/mWxHEGRQalI