I watched almost 80 videos from Tesla's robotaxi launch, including a 4-hour livestream. It went better than I expected! I also finally wrote about the scaling paper Waymo released last month that Tesla fans say proves their point. It's actually more complicated than that.

I took a close look at the 38 most serious crashes Waymo experienced between July 2024 and February 2025. Almost all of them involved a Waymo following the rules and a human-driven car... not doing that. 38 crashes might sound like a lot, but those crashes occurred over more than 30 million miles of driving. If humans were driving those miles we would have had several times as many crashes.

I got 19 subject-matter experts to participate in a blind "taste test" between OpenAI and Google's Deep Research products. More than half said it would take 10+ hours to produce something as good as the OpenAI response. Only three people said that for Google. Seven out of 13 experts said OpenAI's response was at or near the level of an experienced professional. Ten compared it to an intern or entry-level worker. People were not as impressed with Google's responses.

Almost all jobs involve performing some set of tasks, but it doesn't follow that a robot that could perform all the tasks would be able to replace workers doing that job. Many jobs depend on human characteristics that will be difficult or impossible to reproduce using AI. For example, some jobs just inherently require person-to-person contact. Nobody is going to want a robot nanny no matter how good it is at changing diapers.

Generative AI models produce stuff like this and it's a bigger legal vulnerability than a lot of people in the AI community want to admit. I'm excited to publish this copyright explainer I co-authored with @grimmelm. understandingai.org/p/the-ai-commu…
A lot of people in the AI world have an intuition that if it's legal for individuals or academic researchers to learn from copyrighted material, it must be legal for OpenAI and Microsoft to do it on a commercial scale. This is not necessarily correct.

The inconvenience of 30 minute charge times with electric cars is magnified by the sparseness and low reliability of the chargers. You not only have to go out of your way to find a charger, you also need to stop earlier than you would otherwise in case a charger is broken. This should steadily get better as EVs gain market share. By 2040 it’ll be seen as a business requirement for restaurants to have chargers in their parking lots.

I see two broad views about AI risk. One view (held by Kurzweil, Bostrom, Yudkowsky, etc.) holds that sufficiently powerful intelligence is sufficient for world domination, and as a result the first superhuman AI is likely to be the last superhuman AI. In this view alignment is an incredibly important problem because if the first superhuman AI is misaligned we won't have a chance to try again.

I think a lot of liberals underestimate how much this kind of behavior erodes the norms of social trust and solidarity that make societies work well.

https://twitter.com/herandrews/status/1617875151785431050

In functional, high-trust societies you don't need much formal enforcement of the rules by law enforcement because most people comply voluntarily based on shared social expectations.

This seems like a good day to tweet out some of my favorite articles I've written over the years criticizing Elon Musk. A thread... Back in 2016 I did an interview with my favorite Tesla critic, @Tweetermeyer, about Tesla's quality problems. I will say that the Model 3 has worked out better than critics expected. vox.com/2016/6/9/11880…

On #WITHpod, I said that giving parents cash rather than subsidized child care would help men be stay-at-home dads. @chrislhayes described this as "basically a conservative position," noting Mike Lee's proposal for an expanded child tax credit. podcasts.apple.com/us/podcast/dad… I don't think he's wrong that this is how the politics have worked out, but I think it's unfortunate because a giving parents cash when their kids are very small seems like the better policy on the merits. fullstackeconomics.com/p/giving-paren…

I made 9 charts to show the confusing state of the US economy right now. Here's a thread... Private sector data shows rent inflation falling, while the CPI still shows it rising. This is mainly because the former reflects "spot rents" paid by new tenants, while the CPI includes tenants close to the end of their leases.

I did a guest post for @jbarro's newsletter about regulating cryptocurrency. joshbarro.com/p/its-time-to-… For the last decade the crypto industry has been following what I think of as the Uber playbook: their products arguably weren't legal but they launched them anyway, counting on the fact that decentralization would make it difficult to enforce the law.

The Journalism Competition and Preservation Act is an ugly piece of legislation. It's framed as a framework for negotiation, but that process is so rigged against the tech giants that it will basically force Google and Facebook to financially support local news outlets. The negotiations are supposed to be over the terms by which tech giants will "distribute" news organizations' content, but for the most part the tech giants don't distribute their content. They just link to it.

Facebook Marketplace's anti-spam capabilities are astonishingly bad. Only one of these 8 "buyers" for my laptop has a profile that looks like a real person in the DC area. Most appear not to even live in the US. Man, a ton of people want to buy my laptop! Inconveniently almost all of them seem to live outside the United States.

Recently I’ve been thinking about the parallels between Elon Musk and Steve Jobs, another highly successful businessman who had a reputation for being a huge asshole. Apple employees reportedly avoided getting into an elevator with Jobs because of stories of Jobs asking employees what they were working on and then firing them on the spot if he wasn’t impressed with their answers.

Thread! Tech companies have started laying off workers at a time when the broader economy is still adding jobs. I think this represents a significant and possibly permanent shift for the industry. fullstackeconomics.com/p/the-end-of-s… Until this year, tech companies tended to prioritize growth over profitability. This worked out great for Google and Facebook. Not as well for Uber and Lyft. Quite badly for investors in WeWork.

Last fall we surveyed Full Stack Economics readers about which topics they wanted us to cover more or less. Cryptocurrency was by far the least popular topic, with twice as many people asking us to cover it less rather than more. I actually think this is a structural problem with the way the news industry covers crypto right now. If you are a reporter who thinks crypto is probably nonsense, or your audience thinks that, there's little upside in developing deep knowledge of this stuff.

Having Kevin McCarthy be speaker with a 219-member caucus seems like both the funniest possible outcome and the worst for the country because I’ll let the gop’s nuttiest members endlessly torment McCarthy. Democrats are the party of government doing stuff, which means narrow democratic majorities empower the most moderate Democrats. Republicans are the party of government not doing stuff, which means narrow republican majorities empower the most extreme Republicans.

There are two companies offering sidewalk robot delivery services on campuses in the DC area. I tried them both with very different results.
fullstackeconomics.com/p/the-robot-ta… This robot is operating at Howard University here in DC. It was made by Kiwibot. It's cute but it had a lot of trouble getting to its destination. It frequently stopped despite nothing being in its way. It retraced its steps a couple of times for no apparent reason.

In his new book Superabundance, Gale Pooley and Marian Tupy divide resource costs by GDP per hour worked to get an index of "personal resource abundance." Here are abundance charts for 50 widely-used resources. Higher numbers mean greater affordability. amazon.com/Superabundance… Here's the same chart broken down by countries rather than individual resources. It shows the average abundance across the 50 resources in the previous chart.

I went through all 1,566 pages of the 1980 Sears catalog to find items I could easily compare with present-day products. Almost every item I looked at saw significant declines in the hours of work required it buy it. fullstackeconomics.com/24-charts-that… This Craftsman hammer with a fiberglass handle cost $9.59 in 1980. At a typical wage of $6.86, that's about 1.4 hours of work. Today you can buy a comparable Craftsman hammer on Amazon for $13.98—0.5 hours of work at a typical wage of $27.33.