Dylan Vener
@dylanvener.bsky.social
They always do a great write up, but this is unfortunately a nearly identical outage to the giant Google one in June. People like to blame vibecoding, but configuration changes have been one of the largest causes of outages ~forever for the hyperscalers.
November 21, 2025 at 5:50 PM
They always do a great write up, but this is unfortunately a nearly identical outage to the giant Google one in June. People like to blame vibecoding, but configuration changes have been one of the largest causes of outages ~forever for the hyperscalers.
I think it's the same problem. You can make it praise Elon, but it's very hard for it to only praise Elon subtly. Same with racism, their goal is that it it's reasonable most of the time, so when it tells people about immigrant crime rates it sounds reputable. Instead: MechaHitler, uber-Trans Elon.
November 21, 2025 at 3:07 PM
I think it's the same problem. You can make it praise Elon, but it's very hard for it to only praise Elon subtly. Same with racism, their goal is that it it's reasonable most of the time, so when it tells people about immigrant crime rates it sounds reputable. Instead: MechaHitler, uber-Trans Elon.
Explaining to people that the hardest technical challenge of our times is making a racist robot. Empirically, reality literally does have a liberal bias
November 21, 2025 at 2:09 PM
Explaining to people that the hardest technical challenge of our times is making a racist robot. Empirically, reality literally does have a liberal bias
I would interpret him exactly the opposite, he’s not calling anything magical or inexplicable, he’s saying we can’t currently explain it! We are still extremely actively researching what these models are actually doing. www.anthropic.com/research/map... is a bit in the weeds, but very interesting.
Mapping the Mind of a Large Language Model
We have identified how millions of concepts are represented inside Claude Sonnet, one of our deployed large language models. This is the first ever detailed look inside a modern, production-grade larg...
www.anthropic.com
August 8, 2025 at 2:30 PM
I would interpret him exactly the opposite, he’s not calling anything magical or inexplicable, he’s saying we can’t currently explain it! We are still extremely actively researching what these models are actually doing. www.anthropic.com/research/map... is a bit in the weeds, but very interesting.
We understand how birds and planes fly, and in particular how the mechanisms are different, much better than we understand how humans and AIs generate language. The people saying they’re not conscious, or who say they are, both have no clue how to distinguish them, because we don’t know!
August 8, 2025 at 2:23 PM
We understand how birds and planes fly, and in particular how the mechanisms are different, much better than we understand how humans and AIs generate language. The people saying they’re not conscious, or who say they are, both have no clue how to distinguish them, because we don’t know!
Penicillin was straightforwardly miraculous, as are mRNA vaccines and airplanes. Relative to what was possible to do with computer 5 years ago almost everything these models can do is miraculous. Even just viewed as a text prediction model, it’s *crazy* what it can do.
August 8, 2025 at 2:09 PM
Penicillin was straightforwardly miraculous, as are mRNA vaccines and airplanes. Relative to what was possible to do with computer 5 years ago almost everything these models can do is miraculous. Even just viewed as a text prediction model, it’s *crazy* what it can do.
I actually really liked Asteroid City, and I’m not ashamed.
June 2, 2025 at 10:01 AM
I actually really liked Asteroid City, and I’m not ashamed.
If this isn’t rhetorical: yes! I loved the hints of the cases they took between the two books and would really enjoy seeing a glimpse of those more small scale cases.
I’d love opportunities to see more of the world and how it works.
I’d love opportunities to see more of the world and how it works.
May 28, 2025 at 1:59 PM
If this isn’t rhetorical: yes! I loved the hints of the cases they took between the two books and would really enjoy seeing a glimpse of those more small scale cases.
I’d love opportunities to see more of the world and how it works.
I’d love opportunities to see more of the world and how it works.
Apologies if you’ve said elsewhere, but what film did you use for the green and pink building? It’s a great shot, I love the pastels. Reminds me of (intentionally) over exposed Portra 160.
December 1, 2024 at 3:51 PM
Apologies if you’ve said elsewhere, but what film did you use for the green and pink building? It’s a great shot, I love the pastels. Reminds me of (intentionally) over exposed Portra 160.
It’s going to be very weird when some people consider them human and other don’t, and what those tipping points will be. Probably not going to be long before people take p-zombies a bit too far
November 27, 2024 at 5:53 PM
It’s going to be very weird when some people consider them human and other don’t, and what those tipping points will be. Probably not going to be long before people take p-zombies a bit too far