the market manipulation and insider trading presidency
www.reuters.com/sustainabili...
09.03.2026 16:59
👍 142
🔁 81
💬 2
📌 3
the market manipulation and insider trading presidency
www.reuters.com/sustainabili...
yeah, I think a lot of expectations will need to adjust as things become more plausible …
"no woke shit" haha
someone at the pentagon frantically typing “Claude, open the strait of Hormuz for me, quickest possible strategy, make no mistakes.”
two examples of AI's confusing effects on jobs -
1) claude code is great, so we need less computer scientists. except, using it requires skills to prompt, iterate, and debug.
2) AI can do lawyer things. except, AI makes it so easy to do lawyer things we're going to need more lawyers & judges.
it's easier to build now, so you can build them. or use the classical methods, still pretty hard to build physical devices for instance.
in HCI, prototyping or building the thing was always the part that slowed everything down. as a result, we teach methods intended to help people get feedback before a thing exists.
it's so exciting that building is getting faster, but still need good ideas and ways to validate them. that's HCI.
Proud Boys leader Enrique Tarrio meets with Secretary Rubio
Enrique Tarrio and other Proud Boys were convicted of seditious conspiracy. Trump pardoned them.
Tarrio and his men planned the storming of the U.S. Capitol on January 6, 2021. Tarrio even claimed credit for it, messaging others: “Make no mistake…We did this…”
Now he’s meeting with Sec. Rubio.
what is predictable from the inscrutable kind and amount of data in these models is a question that even Claude Code cannot answer for you.
"what's your dream job in the new AI economy?"
"I want to be the person who chooses what synonyms for 'thinking' Claude should show while it's progressing, instead of meaningful progress bar or other feedback."
Update Todos Discover local PDFs from cs.cmu.edu/~jbigham Look up DOIs for publications missing them Add publisher links (ACM DL, etc.) Update BibTeX with enriched metadata Update templates to show new links Rebuild and verify all formats
redoing my CV and such, so far a fun experience with Claude Code
a map showing airplane routes...
TIL despite the recent Azerbaijani cultural genocide of Armenians (in Artsakh), it's kinda nice that the two countries are not at a state of active war, as it's _the only_ remaining corridor for planes flying between Europe and Asia....
I still am not 100% sure what happens if I set an alarm for 9am tomorrow, .. although pretty sure it’s 9am tomorrow time
i figured out what to do in the 1-3 minutes i'm waiting for claude to finish -- i've started another project, so i can interleave the 1-3 minute waits.
every time i click "allow" on a claude code request without reading what's in it, i think about that guy who died in his (nearly) self-driving tesla.
Commentary: Anyone Else Have Those Weird Dreams Where Sobbing Future Generations Beg You To Change Course?
As of Friday, only seven companies are certified to operate “Highly Autonomous Vehicles” with a driver on board in Pennsylvania: Aurora, Motional, Stack, CMU, Perrone Robotics, Mapless AI and Waymo.
www.post-gazette.com/business/tec...
kind of amazing this idea that "computer science" is going to go away because of AI, but people think that other knowledge work isn't susceptible … in the medium term, it might be just the opposite, you're gonna need armies of prompters to customize the machines to each paper pushing subfield
I usually don’t like the formulation where you need to use references or people as a subject , usually feels lazy, ljke “here is a stamenent [ref]”
Promotional poster for the 8th VizWiz Grand Challenge Workshop. Large heading at top left reads “8th VizWiz Grand Challenge Workshop,” followed by “Call for Challenge & Extended Abstract submissions.” Two highlighted deadline banners say “Challenge due: Friday, May 1 (11:59pm AoE)” and “Abstract due: Friday, May 8 (11:59pm AoE).” A small Denver CVPR 2026 graphic appears on the upper right. Below, text states that the workshop presents four dataset challenges designed with data originating from people with visual impairments. Four illustrated panels summarize the challenges: locating private objects, grounding focus regions, grounding all answers, and tracking hierarchical instances.
ECCV submission in? Ready to compete on the leaderboard and build systems that work better for data originating from people with visual impairments?
Join the 8th VizWiz Grand Challenge Workshop at CVPR 2026.
🔗 bit.ly/3MI4rRF
@dr-g.bsky.social @jeffreybigham.com
#CVPR2026 #Accessibility #VizWiz
30 prompts in and now it's messing up and i have to figure out at least a little bit about how all this slop works so i'll have a chance of debugging it 😂
when AI does all the work, the remaining value of humans will be our 'vibes' … think about that for a minute.
pretty dumb though, since i assume he in some way needs to get the thing to run, and then he's the one creating the conceptual scaffolding around it (it's art! file for copyright). i have created computer programs, i run those programs, they do stuff, they are not sole creators of their output.
100% lol
democracy is broken when the best path to a reasonable senator being elected is an unreasonable man winning the primary
There isd a problem with your Amazon Mechanical Turk account this account has been suspended by the Amazon Mechanical Turk team
hey, what did i do?
in computer use, if i say, "book me a flight to Seattle", and it goes off and does that. first off, i can't iterate because of externalities (it might have already bought it!), and it feels super dumb to iterate on it, might as well do it myself. "yeah, like that one, except make it leave earlier"
yeah, but i don't entirely believe that explanation. i think we humans are more forgiving of iteration in code, because code/debug loops are what we always do.
the other part of it is that people have more unstated opinions about what they really want the computer use agent to do, especially as more information is revealed. in coding agents, kind of just want it to work, and again feels natural to refine w/ another prompt after it does a first pass.
it's not at all obvious to me why coding agents seem to work better than computer use agents.
my theory is that it feels kind of natural to prompt and re-prompt code as you work toward a solution. but, it feels absolutely stupid to prompt and re-prompt UI to buy a product on a web page.