How much time do you spend on these updates? They don't seem AI assisted but I also am amazed at their depth even at the summary level more or less at the source. I'm just getting through the Elon interview as of today. Do you generally complete consumption prior to the review?
I find them valuable and they seem like a full time job equivalent from the outside.
So I'm currently unemployed (by choice), so I have a lot of free time. In a typical day I will spend probably 2-6 hours reading, then because I'm traveling when I'm exploring whatever city I'm in I'll listen to podcasts along the way, and then I'll spend probably 1-4 hours writing things out. Right now I'm in Asia, so most of the articles for a day are already there when I wake up, which means I can organize them as I read and stew on them as I walk around.
Yesterday was on the higher end, since it was a big news day so to speak. I should probably cut things down, but my RSS feed keeps growing to fill my available bandwidth.
For AI usage, sometimes I'll have a discussion with Claude about an article if I think I might be misunderstanding it, or to check my intuitions about my takes, like if it seems like there's some obvious point or solution which the article isn't mentioning. I'm thinking of using it for spell checking and grammar, but I feel like giving AI the last pass might not be a good idea, since it might make things feel like they're AI written.
I've only made it through ~ half of the Elon interview but I found his arguments to be totally uncompelling. I work in solar and battery development on the project finance side, so perhaps I'm making an isolated demand for rigor, but a lot of what he was saying didn't pass the sniff test. E.g. at one point he claimed that Chinese solar panels are $0.25/W-$0.30/W, which is 3x-4x their actual cost. Whenever Dwarkesh would gently prod him about how expensive it is to launch things into space he would kind of just shrug it off. He failed to answer what seems like THE question "why not just investing in scaling up nat gat turbine production; isn't that easier than revolutionizing space transport?"
Basically, the whole interview felt more like him trying to hype up the spaceX IPO rather than making a cohesive strategy argument. Feels more like a post-hoc justification. "Well, I have an AI company that needs funding and a space company that is about to unlock a ton of funding; let's mash those together and invent a synergy."
I'll probably look back on this comment in three years and feel like an idiot but that's my take right now.
This was also my take and I'm glad Dwarkesh and John didn't let him totally off the hook. I do think Elon's arguments could make sense down the line, but the universe in which we are launching 10k starships a year for space data centers in the next decade to build an AI God is as improbable or faith based as the universe in which there's just a strait up God already
How much time do you spend on these updates? They don't seem AI assisted but I also am amazed at their depth even at the summary level more or less at the source. I'm just getting through the Elon interview as of today. Do you generally complete consumption prior to the review?
I find them valuable and they seem like a full time job equivalent from the outside.
So I'm currently unemployed (by choice), so I have a lot of free time. In a typical day I will spend probably 2-6 hours reading, then because I'm traveling when I'm exploring whatever city I'm in I'll listen to podcasts along the way, and then I'll spend probably 1-4 hours writing things out. Right now I'm in Asia, so most of the articles for a day are already there when I wake up, which means I can organize them as I read and stew on them as I walk around.
Yesterday was on the higher end, since it was a big news day so to speak. I should probably cut things down, but my RSS feed keeps growing to fill my available bandwidth.
For AI usage, sometimes I'll have a discussion with Claude about an article if I think I might be misunderstanding it, or to check my intuitions about my takes, like if it seems like there's some obvious point or solution which the article isn't mentioning. I'm thinking of using it for spell checking and grammar, but I feel like giving AI the last pass might not be a good idea, since it might make things feel like they're AI written.
That makes sense, and is consistent with my interpretation of the level of care put in here. We can tell. Thank you!
Good question Christine and honestly I'm surprised my obscure post was even discovered at all let alone captured and aptly summarized.
You have a uniquely misanthropic yet optimistic perspective on things and write very well. Definitely a top tier blog.
I've only made it through ~ half of the Elon interview but I found his arguments to be totally uncompelling. I work in solar and battery development on the project finance side, so perhaps I'm making an isolated demand for rigor, but a lot of what he was saying didn't pass the sniff test. E.g. at one point he claimed that Chinese solar panels are $0.25/W-$0.30/W, which is 3x-4x their actual cost. Whenever Dwarkesh would gently prod him about how expensive it is to launch things into space he would kind of just shrug it off. He failed to answer what seems like THE question "why not just investing in scaling up nat gat turbine production; isn't that easier than revolutionizing space transport?"
Basically, the whole interview felt more like him trying to hype up the spaceX IPO rather than making a cohesive strategy argument. Feels more like a post-hoc justification. "Well, I have an AI company that needs funding and a space company that is about to unlock a ton of funding; let's mash those together and invent a synergy."
I'll probably look back on this comment in three years and feel like an idiot but that's my take right now.
This was also my take and I'm glad Dwarkesh and John didn't let him totally off the hook. I do think Elon's arguments could make sense down the line, but the universe in which we are launching 10k starships a year for space data centers in the next decade to build an AI God is as improbable or faith based as the universe in which there's just a strait up God already