Dan Weaver
@danweaver.bsky.social
Minneapolis via Taiwan. Working on solar construction and making puzzles
I think issues with standardized testing are used by schools and some parents to justify a focus on non-academic skills and instilling good values. Good intentions all around but bad results.
September 22, 2026 at 9:18 PM
I think issues with standardized testing are used by schools and some parents to justify a focus on non-academic skills and instilling good values. Good intentions all around but bad results.
My kids are in public school in Minneapolis and I’ve heard that standardized test scores don’t measure what’s important multiple times. That and “oh don’t worry your kid is doing fine.” It’s frustrating because it ends up feeling impolite to talk about a whole range of topics in a normal way
September 22, 2026 at 9:10 PM
My kids are in public school in Minneapolis and I’ve heard that standardized test scores don’t measure what’s important multiple times. That and “oh don’t worry your kid is doing fine.” It’s frustrating because it ends up feeling impolite to talk about a whole range of topics in a normal way
[channeling my inner Gary Marcus] to the extent they work, we don’t know why. we’re releasing products full of capabilities customers will be the first to try. The smarter LLMs get, the subtler their flaws. Even the smartest models make mistakes a five year old wouldn’t.
September 16, 2026 at 8:06 PM
[channeling my inner Gary Marcus] to the extent they work, we don’t know why. we’re releasing products full of capabilities customers will be the first to try. The smarter LLMs get, the subtler their flaws. Even the smartest models make mistakes a five year old wouldn’t.
Yeah. The definitions of “works” I find interesting is “we’ll be able to continue feeding larger neural nets larger amounts of data and they’ll continue to get more capable + aligned.” Plenty to debate there I think.
September 16, 2026 at 7:43 PM
Yeah. The definitions of “works” I find interesting is “we’ll be able to continue feeding larger neural nets larger amounts of data and they’ll continue to get more capable + aligned.” Plenty to debate there I think.
(And I very much hope she becomes skeptical of stats like that as she gets older…)
September 16, 2026 at 6:27 PM
(And I very much hope she becomes skeptical of stats like that as she gets older…)
Oof. I can understand a bit of skepticism then. My daughter’s public school teachers here in Minneapolis seem quite anti AI on the whole. Last year her teacher told her AI has been shown to be wrong 52% of the time. So she’s highly skeptical but I try to encourage a bit of curiosity too.
September 16, 2026 at 6:23 PM
Oof. I can understand a bit of skepticism then. My daughter’s public school teachers here in Minneapolis seem quite anti AI on the whole. Last year her teacher told her AI has been shown to be wrong 52% of the time. So she’s highly skeptical but I try to encourage a bit of curiosity too.
It’s maybe my favorite quadrant of people to follow. I put Gary Marcus and David Deutsch here. I.e. the argument is that LLM tech will hit a wall of some kind. It’s not really intelligence and we should build it differently.
September 16, 2026 at 5:30 PM
It’s maybe my favorite quadrant of people to follow. I put Gary Marcus and David Deutsch here. I.e. the argument is that LLM tech will hit a wall of some kind. It’s not really intelligence and we should build it differently.
For coding, tools and models got significantly better about nine months ago in my experience. I’m curious where you’ve seen staff trying to do the work of lecturers. Are they using AI to write papers or something?
September 16, 2026 at 5:24 PM
For coding, tools and models got significantly better about nine months ago in my experience. I’m curious where you’ve seen staff trying to do the work of lecturers. Are they using AI to write papers or something?
Yeah, only on Bluesky will you see people discussing whether to boycott their local taqueria because they suspect the new welcome sign was made with AI.
September 16, 2026 at 4:55 PM
Yeah, only on Bluesky will you see people discussing whether to boycott their local taqueria because they suspect the new welcome sign was made with AI.
Opinions about valuations map to the “does it work” axis. But I think current valuations are driven by the high end
September 16, 2026 at 3:30 PM
Opinions about valuations map to the “does it work” axis. But I think current valuations are driven by the high end
I can respect suspicion given the circumstances of its development. But I’m curious what happens if the world’s most conscientious people completely opt out.
September 16, 2026 at 3:18 PM
I can respect suspicion given the circumstances of its development. But I’m curious what happens if the world’s most conscientious people completely opt out.
At the risk of overcomplicating it, I think influencers fall into four quadrants, with “Whether it works” and “whether it’s net good for society” on independent axes. The most outspoken BlueSky users are in the “doesn’t work, not good” quadrant.
September 16, 2026 at 2:30 PM
At the risk of overcomplicating it, I think influencers fall into four quadrants, with “Whether it works” and “whether it’s net good for society” on independent axes. The most outspoken BlueSky users are in the “doesn’t work, not good” quadrant.
Maybe in the same way you respect Applebaum I enjoy your work but this opinion seems pretty left bubbled.
September 5, 2026 at 5:16 PM
Maybe in the same way you respect Applebaum I enjoy your work but this opinion seems pretty left bubbled.
Didn’t you gain some notoriety by arguing that the left was not shunning people on flimsy pretexts back in 2021? Would you say that’s still true today right here on Bluesky?
September 5, 2026 at 5:06 PM
Didn’t you gain some notoriety by arguing that the left was not shunning people on flimsy pretexts back in 2021? Would you say that’s still true today right here on Bluesky?
Oof, yeah stories like that make me want to live in a van by the river. How do we develop an inoculation to sycophancy??
September 4, 2026 at 12:02 PM
Oof, yeah stories like that make me want to live in a van by the river. How do we develop an inoculation to sycophancy??
As for psychosis, I won’t argue for my own or anyone else’s sanity. I’d say it’s similar to other tech in that you have to use it in a healthy way, but it poses different challenges than social media. Sometimes it feels gross or nauseating. I respect choosing to abstain.
September 4, 2026 at 4:29 AM
As for psychosis, I won’t argue for my own or anyone else’s sanity. I’d say it’s similar to other tech in that you have to use it in a healthy way, but it poses different challenges than social media. Sometimes it feels gross or nauseating. I respect choosing to abstain.
I should have been more specific. What I meant was that for a thing you could do with AI three years ago, the same thing can be done now 10-100x more electricity efficiently on current hardware and smaller models. But usage is outpacing efficiency gains, so you’re right to be concerned.
September 4, 2026 at 3:33 AM
I should have been more specific. What I meant was that for a thing you could do with AI three years ago, the same thing can be done now 10-100x more electricity efficiently on current hardware and smaller models. But usage is outpacing efficiency gains, so you’re right to be concerned.
Yes this! It makes plausible solutions sound like not enough.
September 3, 2026 at 5:56 PM
Yes this! It makes plausible solutions sound like not enough.
Framing it as good vs evil rather than unfair makes lefties feel they need cosplay being one of the “good guys” who are not rich. So if you propose something boring and practical like closing enormous tax loopholes the response is “no no that doesn’t go nearly far enough”
September 3, 2026 at 5:35 PM
Framing it as good vs evil rather than unfair makes lefties feel they need cosplay being one of the “good guys” who are not rich. So if you propose something boring and practical like closing enormous tax loopholes the response is “no no that doesn’t go nearly far enough”
What should they do instead of telling us it might be a bubble?
September 2, 2026 at 1:04 PM
What should they do instead of telling us it might be a bubble?
Debates about whether VPs manage other people, as if anything whatsoever hinges on that. Ugh.
Some of these are bots maybe? There are a lot of real assholes, but it would be so easy to troll a platform like Bluesky which is full of people who think they’d know AI writing if they saw it.
Some of these are bots maybe? There are a lot of real assholes, but it would be so easy to troll a platform like Bluesky which is full of people who think they’d know AI writing if they saw it.
September 1, 2026 at 5:15 PM
Debates about whether VPs manage other people, as if anything whatsoever hinges on that. Ugh.
Some of these are bots maybe? There are a lot of real assholes, but it would be so easy to troll a platform like Bluesky which is full of people who think they’d know AI writing if they saw it.
Some of these are bots maybe? There are a lot of real assholes, but it would be so easy to troll a platform like Bluesky which is full of people who think they’d know AI writing if they saw it.
I’m hearing “why are republicans focusing on a few kids on sports teams” and “a few kids on sports teams is equivalent to interracial marriage, we hate anyone uncomfortable with any detail”. Messaging works better if you show some empathy/respect for recipient.
September 1, 2026 at 1:10 PM
I’m hearing “why are republicans focusing on a few kids on sports teams” and “a few kids on sports teams is equivalent to interracial marriage, we hate anyone uncomfortable with any detail”. Messaging works better if you show some empathy/respect for recipient.