Just a rando on Intertubes

Joined March 2009
3,864 Photos and videos
Pinned Tweet
5 Dec 2025
AI is amazing!
2
15
6,519
Jun 15
hm. i have to code my calendar entries better if AI analysis is going to be any use.
19
Jun 14
boo.
21
Jun 13
well, i had to try it out to compare (things shift/change). 'looks like i'm sticking with NotebookLM for slide generation for a while.
39
Jun 13
do they have any data? transfer students never needed SAT scores. if they can show that their freshmen admits are now actually doing worse than transfer students, then sure why not go back?
🔥Humanities professors unite w/ STEM faculty, demanding the University of California reinstate the SAT “We call for the UC Academic Senate and the UC Regents to give up the failed experiment of the last 6 years & return to including both the math & the verbal reasoning components of SAT/ACT as part of undergraduate admissions”
55
Jun 12
huh. there's a got4o-transcribe model? maybe starting with this transcription will work better than my current Whisper-based workflow. (manual transcript review is the most painful part of the process.)
1
45
Jun 12
huh. if it does that correctly, that would be a pretty smart comparison.
1
26
Jun 12
yeah, pretty substantial improvement. bottleneck is back to me, *mostly* on how long i can stay focused (over a long manual review, my attention drifts and i can't quite follow the audio and text together, even when no correction is needed).
8
"The victim was able to use Eryilmaz’s phone to call her sister" i find that amazing. if i lost my phone and didn't have access to my online accounts, i have literally no one whose phone number i would remember well enough to call.
A San Francisco jury found a 44-year-old man guilty of raping a woman who mistakenly got into his car instead of an Uber, District Attorney Brooke Jenkins announced Friday. STORY: kron4.com/news/bay-area/sf-m…
1
570
edited all the banked turn questions to recognize a_c = mg sin theta mistake specifically (hopefully didn't introduce any regression; will need a full-semester's testing to know).
1
50
the challenge was, it amounts to swapping tan theta for sin theta in most cases, ... which means small angle approximation basically makes them same (within the 1% or 3% relative tolerance used in most questions). so, i had to carve out an exception (applicable when a student answers with enough significant figures to tell them apart) to handle the "sin theta" answers *specifically* differently, even though it falls within the larger umbrella of correct answer - relative tolerance.
51
huh. this slide isn't *exactly* lecture-faithful, but if re-recording that segment of lecture weren't such a pain, i probably would've recorded so that this is correct (specifically using relative velocity relationship in place of KE conservation). so ... retain?
43
sigh. 'was hoping somehow Edge with Copilot would have less hasslesome agentic browsing enabled, but no. :|
40
i don't think i'm getting rid of ChatGPT subscription any time soon. Google's own tools are so ... so far behind. first is my interaction with Gemini trying to find a student name (if i knew how to spell it correctly, a direct email search would have surfaced it). the second is ChatGPT working with Gmail and Contacts connectors.
67
ooh. yay, LTI 1.3!
1
40
oh, wow. LTI 1.3 just takes over all the connections. it's actually good for me, because that means i won't have to re-do the connections for the remaining assignments (just let LTI 1.3 handle it).
39
huh. suddenly "Gemini Act on Web" started working on my personal computer... but *only* for one of the two profiles (the profile that's my "personal Workspace" account; for some odd reason the account i created with my work email---but not work managed---still doesn't work).
29
huh. 'didn't know you could do this in a locally run version. this might improve my process (i've been using ChatGPT to assemble the sentences, and Whisper has more info to correctly assemble them, if it can be done at that stage).
1
40
is Uber trying to get back to autonomous driving game? what is there to code so extensively?
Dara (CEO of Uber) on their AI spend: "We blew through our AI budget in a quarter, for the whole year. It is forcing us to adjust. We are going to meter headcount increases because to the extent that my engineers are getting much more efficient, their throughput is increasing. There's a cost to that, and it's a significant cost. AI adoption has been occurring in all parts of the business –– whether it's engineers and how they scope projects, how they build, debugging, platform migrations. I'm pushing the teams to fundamentally use the power of AI to rebuild systems and processes from the bottoms up. I do think it's a combination for us right now of encouraging adoption, but then driving efficiency. We're using the more expensive models to explore. Once we scale some of these experiences, we'll look to bring in more efficient models that are more efficient on a token basis or are open source."
48
eh. i think AI is going to be like ship's computer in Star Trek (TNG and on). no one worries about if it's conscious or not, despite being voiced by Nurse Chapel (TOS); it's just very intelligent and very useful.
If we confuse generative AI’s ability to produce text with consciousness, we risk assigning moral responsibility to chatbots—and not to their makers, Ted Chiang argues. theatlantic.com/philosophy/2…
1
570