Waymo reveals it has 70 remote workers managing over 3,000 automated vehicles
And every once in a while I still see articles claiming this means that the cars aren't self driving at all, and actually they're human driven. The whole conversation is just so ridiculous and I'm amazed Waymo has done as well as it has to date.
At some point, when vehicle production can scale to meet demand, the insurance companies and their actuaries will be able to lay a lot off this to rest. "Yes, madam, I understand you don't trust self-driving cars, but are you aware you could reduce your premiums by 90% and also spend more time doing things other than driving?"
AI continue to accelerate and dominate the schedule, which is why this is a bit late, but we do occasionally need to pay our respects to the Goddess of Everything Else.
There’s cool or interesting things everywhere. Also maddenning things. But did you hear, for example, that they’re making some exceptions to the Jones Act?
Table of Contents
Bad News
Seth Burn points out that if Google wanted to avoid fake reviews, the ‘report review’ feature would have an option for ‘this is a fake review.’ It doesn’t.
Apple by default stores your Signal messages in an iOS notification database, which allows authorities to extract them. Signal is asking Apple to address this. In the meantime, if you want to be protected, go to Signal Settings > Notifications > Show “No Name Or Content.”
The gambling industry is indeed out of control, the example here being that Carton filled out a self-exclusion form for online gambling in New Jersey, and a brick-and-mortar casino in Atlantic City used this fact to market to him to come on over.
Yes, this is a big problem:
This can often be true, but at least at Anthropic it seems clearly not true, as Dario Amodei’s strategy memos usually do not leak. Could OpenAI pull off a Manhattan Project from an OpSec point of view? My guess is if they cared they could do about as well as the actual Manhattan Project did, which as we know was infiltrated by Russian spies.
It’s true:
It’s one thing when you work at Jane Street on a trading desk and everyone has to communicate. If you’re at a tech company, there’s no excuse. You’d have to at least double my pay to get me to work in an open office at this point.
I am with Grant Slatton in that I don’t understand why Lyft, Uber, DoorDash and similar services don’t lean harder on customer (and driver) reputation when evaluating claims. And also perhaps look for impossible situations like claimed 3 minute rides.
Also, I mean, this response to the OP is another sign that they are doing the wrong amount of vibe coding over there. Could be too much, could be too little.
Good Advice
A classic. Is he right? He’s kind of right, remarkably often, but not always.
‘Control freak’ also can mean something else. Golden mean and all that. But there’s no reason to assume that the golden mean can’t lie in the fat tail of the distribution.
All four have much to teach those with eyes to see.
Opportunity Knocks
Cate Hall and Sasha Cohen can just write a book, so they did, and it’s called You Can Just Do Things: How High-Agency People Get What They Want Out of Life. Self-recommending, and as usual pre-orders are very good for book sales.
Slutcon 2026 tickets are live. It will be September 25-27 in East Bay, CA. Based on all reports from the first version, this is an excellent product.
Who Judges The Judges
Could we provide judges with better incentives, as Alex Tabarrok suggests via Landsburg, by paying judges a bounty when they release a defendant but charging them fines for later crimes?
It is a fun thought experiment, but you don’t want to go there. It destroys the sacred values of asking the judge to follow the law, to serve justice and temper with mercy, to treat people equally and ignore inadmissible evidence, and to consider broader implications.
Instead, many judges will quickly look to start maximizing expected revenue, at least on many margins. So you shift the question to what that decision might be, and soon that has a huge weight on trial outcomes. You also have to worry that judges will then intervene to get future charges dismissed, lest the judge be fined.
The general principle is that economic incentives are great, but you can quickly crowd out other considerations, so you need to cover all your bases or you likely do more harm than good. You don’t want to turn our courts into a Minority Report situation.
A judge shouldn’t primarily be thinking about the recidivism rate when passing judgment, in most cases. They should be thinking about truth, law and justice.
If you narrowly applied this to some parole decisions, that might be reasonable, but only once you are ready to toss out all other considerations.
I also think there is much wisdom in this:
Indeed. If we were capable of implementing this policy, we could choose a better policy.
Close Socrates
Benjamin Hoffman writes that Socrates is Mortal, explaining his view of Socrates acting in particular situations for particular grounded reasons, who he views as one of the few remaining alive persons in an Athens in crisis, where people still felt obligated to engage with him. He views Agnes Callard’s vision as wanting a Socrates that is timeless, and thus lacks the aliveness, and thus is bad, and that I did not need to spend so many words responding to the fake decontextualized Socrates and winning a lame debate, instead of engaging with the real one.
Hoffman pulls out stories and details about Socrates I hadn’t heard before, and frames his actions there in different lights than either what I’ve read in the books or what Agnes presented, even taken in their original contexts. I felt I did look at the original contexts that she chose to offer, and did not find the good thing.
The Socrates that proposes borrowing money to put refugees to work with wool or funding countersuits sounds great indeed. He’s coming up with solutions to practical problems by thinking in concrete terms. But he also sounds completely different from the one I’ve read about.
Similarly, Hoffman offers an alternative reading of the questioning of Euthyphro, where Socrates is motivated by a need to know how to defend himself in court against charges of impiety, that he has a practical need. One could imagine it so, even if his approach to solving this was higher level than the situation warranted, and would have been better served by raising his own particular example and in focusing on how others would view it rather than trying to solve for an abstraction.
The practical Socrates would know not to try and do too much or solve too general a problem here given limited time, and pivot to his own. His failure to do so makes me think he was following a far more general procedure, rather than trying to solve a particular problem.
Even more than that, the practical Socrates would know that ‘impiety’ was the nominal charge, but was not why he was on trial. No one could even define piety, which was a central point of the dialogue, and this issue was wisely not his defense at trial. Formally, Socrates was actually on trial for ‘impiety and ‘corrupting the youth of Athens’ but everyone agrees that was Obvious Nonsense, an excuse to get rid of him. To the (likely small) extent it was ‘impiety’ it was that Socrates claimed divine guidance, and you don’t need a full definition of piety to know that doing this is going to have piety issues. So Socrates saying ‘I need this information for my own defense’ is also Obvious Nonsense.
Realistically, he was on trial because of some combination of:
These are entirely consistent with Hoffman’s view that Socrates was trying to wake people up and cut through a bunch of bullshit that was dominating public life, but he’s not going to be helped by definitions.
It’s also possible that ‘corrupted the youth of Athens’ was partly literal, as per Plato’s own descriptions in The Symposium. He was not not running a child sex cult, including in ways that might have led to Alcibiades. We presume everyone was basically cool with this, but maybe they weren’t. Combine that with the rhetorical brainwashing skills I go over, and yeah, the charge seems rather accurate to me.
At heart, of course, the Athenian justice system at the time was majority votes, not due process of law, so ‘everybody hates you, sir’ was a valid charge.
I also hold to my read on Meno, contra Hoffman, that Socrates was not actually asking the slave to solve the math problem in a meaningful sense.
So essentially, I am all for the importance of the aliveness characteristic that Hoffman is pointing at here. I simply don’t see that character coming through in my (admittedly limited) experience of the source material.
I also notice this reaction to my writing up of the review of Open Socrates is helpful, I am glad to get it, but makes me sad, as it indicates that in this way I missed the mark – I was trying to use winning that ‘lame argument’ as a jumping off point or baseline to do things that were less lame, but that seems not to have come through to him, and yes of course it is too long. So that was disappointing.
So far I have not seen a response from Agnus herself, which is also disappointing.
I’m glad I ran the experiment of writing my review of Open Socrates (part 1, part 2). Even though it was a ton of work and didn’t get the reception I hoped. I know at least some people got a kick out of it, at least, but that definitely isn’t good enough here.
Still, I had a lot of fun and interesting thoughts in the act of writing it, and I learned a lot, including by watching it not work. If none of your ambitious posting falls flat, you don’t do ambitious enough posting.
I suspect there’s a lot of good material in there that would do better if it I cut the parts that didn’t land, and it was written in a sequence form that no longer referenced the original. I may try that project out, perhaps as a LessWrong-only series.
Also, there’s this:
While I Cannot Condone This
Birth Tourism, as in ensuring that your child gets birthright citizenship, has highly positive selection for future citizens that will have high value, but that depends on the Levels of Friction and costs involved in doing it. Thus, you want to make this doable, but you do not want to make it too easy.
You are not being watched by alien drones. But it is a fun exercise to ask, as Tyler Cowen does, how you should change your life decisions if you are being watched by alien drones.
Tyler Cowen suggests, uncharacteristically, that you would want to slightly lower your level of ambition, in case the aliens kill us, or cap human achievement, and because this should update you towards humans having lower marginal product.
Jehan Azad responds that instead this shows incredible physics is possible, so who knows what is possible, and AI is less likely to kill us and we are in a competitive environment, so it’s time to build and get more ambitious.
I award this debate to Jehan. You would want to get more ambitious. I think ‘physics does not bind the way you think it does, and the aliens exist but are not grabby and presumably did not all die to their AIs’ should be a whitepill to get to work.
The famous Netflix culture slide deck. There is room for a small number of companies to work this way. I’d have loved to give this a shot when I was younger.
Good News, Everyone
A platonic friend matchmaking service is a great idea, as is ‘hold events to help make friends and charge people money,’ and none of this should be mocked. I look forward to AI-empowered versions.
You do not need to fully be heroicsmaxxing (as in, going above and beyond to solve problems) and there are times when heroics do harm by masking the problem, but you absolutely need to be rewarding or expecting heroics if you’re doing something that matters. The right amount of heroics is highly nonzero.
My favorite anecdote about this will always be when I was sat down during a corporate annual meeting, along with many similar others, and we were told ‘the age of heroes is over.’ That was the exact second I mentally checked out of most (but not all) aspects of that job and I thought about how to be a hero somewhere else.
Violence Is Never The Answer
If you needed a demonstration of the fact that the endorsement of violence mostly is completely unrelated to AI, the recent interviews and discussions in The New York Times and involving Piker should make this extra clear.
There is an entire utterly distinct faction that thinks that it is some combination of understandable, sympathetic and praiseworthy to steal and to shoplift, and also to commit ‘social murder,’ and is being ‘platformed’ to discuss this, as they would say, in The New York Times. And then those people will say ‘90% of people on the streets of our cities would agree’ and find such views unremarkable.
For Your Entertainment
Scott Sumner Q1 movie reviews. Like everyone else his favorite new film is Sirat.
Cool facts about The Pitt, from an interview with Noah Wyle. I didn’t understand the impact of the camera angles in particular until Noah pointed it out here but yeah it works so well. I had definitely noticed the impact of going in real time, and that it was competence porn, and I am usually in for some good competence porn.
If I had to make one change to The Pitt, it would be creating an interactive or enhanced version of some kind that allowed you to use the show to learn about medicine for real – I feel like I’m just on the edge of actually learning stuff worth learning, and would love to do so if it could be done without pausing to chat with Claude. I could totally do that in theory, but in practice I won’t.
Speaking of competence porn, Andy Weir, author of The Martian and Project Hail Mary, pitched a Star Trek show, but was shot down by Paramount, who is not interested in producing actual Star Trek. Weir is correctly responding by pointing out that Star Trek is no longer Star Trek, while taking the generous position of being willing to allow Enterprise and Strange New Worlds and not even be mad at Lower Decks. But yeah, at minimum the rest can totally go. My choice of cutoff for canon would be to count Enterprise and stop there.
Remember this Star Trek competence porn, where Data explains how the professional chain of command works? Yeah, we need more such competence porn that lays out how to Do Competence and set high expectations. Star Trek also has a bunch of other things it wants to do, but it also needs to do this.
Matt Yglesias is of course correct that movies are too long (and rotate too fast). Most of the time if your movie is over two hours that is you making a mistake. Distribution issues used to help us more with this, and then it stopped helping.
Are there exceptions? Of course, when it is justified by all means go long. Kill Bill: The Whole Bloody Affair is probably my favorite movie and it’s four hours long, but you have to earn it. Whereas if you get done in 90 minutes, and give us a short ass movie, we can forgive quite a lot.
Someone in New York City really should open a full ‘30 years ago’ theater, which just plays things from 30 years ago and advances one week at a time, movies are minimum one week and then rotate when they stop selling, maybe you round out some stuff in the early year lull with broader greatest hits.
The cost of Netflix Standard plans has indeed been rising in real dollars over time.
It is still an amazing deal there are things you want to watch, and yes you should pay to avoid the ads. If you’re cheap, don’t suffer through ads, instead rotate between services and only have one active at a time.
I’m actively disappointed by this trailer for the new Street Fighter movie in several ways, including that this wasn’t AI even though it totally could have been, and also no one cares about Ken what are we even doing, and in general it’s going to be terrible on every level, although I’m going to watch it anyway cause why not. I’d be shocked if the old version isn’t ten times better.
Gamers Gonna Game Game Game Game Game
Dan Felder is running a design bootcamp on holistic game systems. It starts the week of May 30 and early bird enrollment closes on April 30.
Civilization, the original really was amazingly good times. Alas, that doesn’t make it easy to go back. You could maybe go back to 2, realistically it’s 4 or 6. I still haven’t played with 7.
How are prediction markets changing the music industry? Rolling Stone’s answer is essentially that they aren’t. Everything is as before, except there are some people betting on outcomes at Polymarket and Kalshi. Okay, then, sure.
What are the best resource systems for TCGs?
My opinion is that lands are the best system if and only if you can get away with it.
As in, Magic: The Gathering is a vastly better game, and remains the gold standard, because it has successfully grandfathered in the right to use lands. Players accept it, along with the cost that there will be many non-games with mana flood and screw, in order to get all the good things that go along with it.
Alas, when making a new game now, you basically don’t get to do that. Players won’t put up with it in a tabula rosa. So you’re basically forced into some form of one resource per turn, or alternatives that are worse.
Emergents, the game I created, used one per turn, with special rules for building faction resources based on what card you played or tried to play via wildcard, plus having temporary access based on the top of your deck. I think it did a great job of creating interesting decisions where you had to plan your game out and providing variance. It wasn’t as good as lands, but you don’t simply get to have lands.
I think at this point I’m with Sam Black that I swap 4 and 5, even knowing that no resources can go rather horribly wrong when you get it wrong. As fun and interesting as it is in theory, and I’m the one who invented ‘play the cards face down and then cast the spells by turning them face up’ system in Versus, forcing the face down decisions that often want to use the most expensive cards is hell on a lot of players and often takes forever. In the end it simply isn’t practical.
Hire gamers who are good at games, they are good at life, as in this study that good players of Civilization V do well in business school student experiments. Guild leaders are great picks, as are high performers in MOBAs, and my favorite pick is players who excel at Magic: The Gathering.
One exception found here is you don’t want FPS winners. It’s not a strike against you, but it’s not an argument in your favor either.
I’ve Got The Magic In Me
Tranquil Domain put out quite the video about my old days on the Pro Tour, including clips from a new interview I did. I had a lot of fun watching it back, even if it gets a bit excited at times.
One correction or clarification that I was explicit about in the interview, and that I simply cannot get to stick, is that the designer of the main deck of The Solution was my British teammate, John Ormerod, including the inclusion of Crimson Acolyte. That was a brilliant move that I doubt anyone else would have found, and his proposal for the main deck did not change throughout testing.
I did build the sideboard, including searching every card and finding Pure Reflection.
I love talking about cEDH (competitive commander) but wow the timing situation is rather out of control. Sam Black thinks the problem is that the format is being warped by Rhystic Study, and it needs badly to be banned.
What is going on? Oh, it’s amazing. In theory. Not so much in practice.
As Sam also adds, the tournament he is reporting from was well-run and was fun and a unique experience. But that doesn’t mean you would want to do it again.
The core problem is that if enough people are playing Rhystic Study this way, either you follow suit and play mostly for the (multi-directional) mirror, or you lose, period.
The correct answer is that Rhystic Study is a rather silly card in a four player game even when it is not competitive, and regular Commander should respect the situation and ban the card even if it wouldn’t otherwise rise to that level. cEDH is willing to roll with the Commander banned list even in places where it is silly in a competitive context, in both directions, so you should be willing to let them have this one.
If regular Commander does not do that, then yes, I think banning the one card in cEDH is totally fine. If you’re throwing that card into a normal deck it’s no skin off your back to swap something else in. If you’re building around it, well, stop it.
I don’t think ‘fix the tournament design’ is a reasonable answer here. You shouldn’t have a strategy that dominates except that it makes tournaments take 5 more hours, even if that were to make it fair again. There are infinite cards. You can go without one of them.
I Was Promised Flying Self-Driving Cars
Waymo launches commercially in Dallas, Houston, San Antonio and Orlando, and starts the process moving in Chicago and Charlotte, then later launched in Nashville.
Waymo raises $16 billion at a $126 billion post-money valuation, most of it from Alphabet, which Brett Winton notes values it at ‘~$40 million per robotaxi.’ That will then decline by 7% per month.
Waymo introduces the Waymo World Model, built with Genie 3, to train itself to deal with unusual situations like elephants.
Waymo reveals it has 70 remote workers managing over 3,000 automated vehicles on over 400,000 rides per week, or ~2% of total ride time. That’s still a relevant ‘last mile’ but is down to a small portion of total costs. If I was Waymo I would move those 70 jobs to America. You’d be paying a premium for mostly the same work, but the PR benefits would be orders of magnitude larger than the cost.
It’s a battle of
rent seeking mafiosoteamsters against Waymo over an attempt to legalize self-driving cars in Maryland.Washington DC is stalling against Waymo by claiming to be waiting for a pointless report due in 2022 that would take less than a month for one person to write.
The public comments are out for blood in Portland, full of claims of safety and worry about Uber jobs. Public comment meetings made sense back before we had telecommunications, now by all reports they are hellholes.
Bloomberg’s David Zipper argued “We Still Don’t Know if Robotaxis Are Safer Than Human Drivers.” That’s completely bonkers, so Kelsey Piper got out a big hammer to explain that yes, Waymos are staggeringly safer than human drivers, as in they have a 80%-90% lower risk of serious crashes even when among human drivers.
Zipper says we don’t have the ‘raw data’ on Waymos, but we very much do, and Kelsey goes on to identify various statistical errors or slights of hand, depending on your perspective. This is all very settled.
And yet we have various politicians, here DC mayoral candidate Janeese Lewis George, opposing Waymo due to supposed safety concerns. Within reason I think mayoral races where this is in doubt justify one-issue voting.
A Tesla Model S drove itself from Los Angeles to New York with zero disengagements. Full reverse cannonball run.
There’s still a difference between full self-driving (FSD) that can take you across the country, and the point when you can sleep while it drives.
A Waymo moving 17mph hits the breaks instantly upon seeing a child step in front of it from a blind spot, hits the child at 6mph and dialed 911. If a human had been driving, the child would likely have been struck at 14mph and be dead.
What did some headlines call this, of course?
This was another good time to notice that almost all the AI Safety people are strongly in favor of Waymo and self-driving cars.
Senator Ed Markey is going after Waymo on the basis that they employ people in The Philippines ‘8000 miles away’ to ‘guide’ the cars, as if that is dangerous. Such opposition is what is dangerous, but also this seems like an unforced error by Waymo. The money you save is not worth the talking point, and Waymo should bring those jobs back to America.
Waymo has a better ratio of agents to vehicles than its rivals, with 43 vehicles for every remote agent.
Humans don’t typically want to drive the Ubers for long, or as a full time job. Weekend morning shifts pay decently, here about $100 in Texas, and weekday evenings $35, but pay at other times is worse. And reports I’ve seen are that turnover is extremely high.
Sports Go Sports
The Trump administration tries to set broad NCAA policy by fiat, as in Executive Order, demanding a five-year eligibility cap, one free transfer, national agent registry, medical care protections for athletes, women’s/Olympic sport protections and a ban on NIL collectives it calls ‘fraudulent schemes.’
This is not how our government works, but Trump would to just declare things, so he’s trying to threaten NIH or other funding to force the universities to do what he wants, even when what he wants has been ruled illegal by courts and doesn’t actually have a working legal definition or plan to deal with the existing court rulings. He just thinks he can say ‘implement these things or else I will cut your funding, even though the courts probably think that is illegal, I don’t care,’ and sit back.
The good news is that there seems to be momentum behind passing something, and everyone smiled about the order. The bad news is that all of that is meaningless.
How did we end up with a legal system where there is no punishment for repeatedly issuing orders that you yourself know are illegal, other than ending enforcement of those illegal orders after someone sues, thus allowing this to be used as leverage?
Shrug.
What should the NCAA committee have done with Miami OH, which had a perfect regular season despite facing hard opponents or being, according to gamblers and those with eyes, not all that good at playing basketball?
I agree with Seth Burn, my good friend and inventor of WAB, that this illustrates that you are better off if you pick a metric in advance and then stick by it. If you do not like WAB, you need to offer something better. And you definitely shouldn’t care about who is injured or is coming back from injury. Your record is your record.
The counterargument is that any fixed system can be gamed. You want to care about margin of victory, but be able to penalize teams that run up the score, or do various ugly things. You want room to do things that are cool and fun for fans on the margin, or very good for business, even if they’re not technically correct.
In particular, you want to be able to care about who is actually good and what the gamblers think, but for various reasons you can’t actually bake that into explicit criteria. So I am sympathetic there. It would actually be cool to have the final 2-4 bubble slots in the play-in games be ‘the gamblers picks’ now that gambling is legal, and do it via conditional game odds against whoever their opponent would be. So much fun. But I presume we can’t have nice things.
The bottom line is simple. Does anyone in MLB or NFL or NBA or NHL think it’s unfair when the ‘worse’ team gets in because they won more games? No, of course not, they know the rule. That’s sports. So have a rule, and then follow that rule.
In any case, so long as you avoid egregious errors, it’s a small mistake to let a team in ‘on the bubble’ over another bubble team, because they’re not that likely to win.
NJ Transit is going to jack up ticket prices to $100 during the FIFA World Cup. Not prices for the matches. Prices for the trains returning to New York City, up from the usual $12.90. If one must do demand destruction on public transit, best to do it by price, but one would hope there are ways to avoid falling short on supply.
Robot Umps Now
Robot umps now? Robot umps now. All signs are that ABS, the system where pitchers and batters can challenge umpire calls, is working great.
Umpire shaming is fun for the whole family and I agree with Derek Thompson that more leagues should do it. I especially love the ‘tap the helmet and start walking to first base’ move.
I actually think this could be a great compromise. In most situations, you still get the human umpire and their ability to make different games unique, adjust the strike zone to the count, slightly favor the home team and create a level of randomness and uncertainty, and reward pitch framing, all of which is part of what makes baseball interesting. But, when the error is sufficiently egregious or high enough stakes, we can get it corrected, and being able to know the strike zone and when to challenge becomes another skill and discussion to keep things interesting.
Ultimately, yes, robot umps now, but I am conceptually enjoying this middle zone. Alas, I’m not actually enjoying it, in the sense that I am not watching any baseball, because I don’t have that kind of time.
Polymarket becomes the exclusive prediction market exchange provider of MLB.
The NBA Needs A Redesign
How should we fix the NBA draft and associated tanking? Nate Silver has a self-described radical (April Fools Day) proposal. Obvious first revision is to break ties for max bid players via whoever has least recently won a max bid player, rather than having teams draw lots, obvious second revision is remove the hoarding tax and also remove to max bid, third is to revise the penalty for missing playoffs repeatedly to be a cap on total gains.
As usual, the actual league is considering something simpler and easier that they can feel they came up with, in this case getting rid of rewards for records entirely and just starting everyone with 100 draft credits each year. That’s obviously second best in various ways but actually might be flat out better than Nate’s proposal. Why reward losing and punish winning at all? If you do a better job you get to win more.
There is also a compromise between these two proposals. Teams still want to make the playoffs if they can do so, so all you have to do is give all teams that don’t make it to a play-in game 100 points, and then assign teams beyond that fewer points. That still gives you some amount of self-balancing, and if you didn’t drop that number off too much too fast then presumably no one would tank.
How sports team math works:
One can model this as ‘people want to be associated with successful and popular futbol franchises, sufficiently so that they can afford to always lose money.’
Zac Hill and Matthew Yglesias point out the problem goes beyond tanking, even if you love basketball.
Derek Thompson is even more brutal. If you wanted to make a meaningless regular season, could you do better than the NBA?
The NBA has much work to do, and not only in the draft and the three point line.
The playoffs are good times, but at the expense of the regular season. You need to raise the stakes of the regular season, by increasing the value of seeding, at minimum.
Government Working
The BBC is alleging that some in the Trump administration ‘may be’ insider trading on events surrounding the war in Iran. To which the response of anyone paying attention is, well, no shit, Sherlock.
Scott Alexander is correct that Orban Was Bad, Even Though We Don’t Have A Perfect Word For His Badness. There is a continuum between free and fair democracy and full dictatorship. He was not maximally a ‘strongman,’ ‘autocrat’ or ‘dictator,’ and but neither was he entirely not those things. Him losing an election and giving up power does not retroactively make him not at all those things, and does not mean there was no serious threat to democracy there. It certainly does not mean there is no serious threat to it elsewhere and I cannot read such claims as unmotivated.
TSA lines got completely out of control for a while. Here is a video for one in Houston, where the wait was several hours at least. The problem is that there are the same number of flights and less throughput, which means the line keeps getting longer until you get substantial demand destruction. That doesn’t happen until the wait is hours long, or you otherwise make the experience sufficiently worse to drive people away, such as putting ICE agents around the airport.
Yes, of course there are those saying openly that ICE agents at the airport are a test run for ICE agents at polling places.
The FCC made all routers covered devices, so if you want to use a foreign one you now need to go through a ‘conditional approval’ process to show a given one does not pose unacceptable risk. Luckily this does not retroactively ban existing routers or approved models.
FCC Chairman Brendan Carr made this reprehensible statement, which is part of a broader pattern of threatening the media for expressing views he thinks Trump dislikes.
This one isn’t even on Mamdani, it predates him but presumably is about to get worse: New York City spending per homeless person rose in six years from $28k, already quite a lot, to $81k, which is patently absurd. That is vastly more than would be required to nicely house and pay all the living expenses of such folks, and it ignores secondary costs.
I try not to opine on such matters, but there are obvious implications here.
The obvious problem is, who controls the US Postal Service? What happens if they are directed to not deliver the ballots until after election day, or otherwise prevented from doing so? Alas, this no longer seems like such an outlandish question to be asking. And there is no obvious limit to how long they could stall this process out for.
Washington State fully bans noncompete agreements.
Did you know our Federal government spends ten times as much per person on our elderly as it does on our children? Please step telling me there is no way we could do anything about declining birth rates or to help working families.
In case you were holding out hope for Mamdani being a secretly good mayor, not only is he allowing Waymo testing to halt, he is for real about to follow through on government owned grocery stores. I am going to make the easiest prediction ever and say the grocery store, if it happens at all, will be a money pit and go absurdly badly.
One reason it is so easy is that this is already a money pit that has gone absurdly badly, even at the planning stage, my lord:
A quick Claude analysis says that $30 million is the upper range of costs to open a full 15k+ square foot supermarket in a central location. This proposal is for 9,000 square feet in Harlem, on free land, which Claude estimates ‘should’ cost $6 million at most.
He’s saying $70 million for five stores and spending $30 million of it on the first one even if it all goes as planned.
Also, there are already five grocery stores within a two block radius (!) of the new store, so what the hell is even the point if this isn’t graft?
Oh, and it turns out the whole thing is justified by an invalid statistic that compares grocery store spending without controlling for purchase type, and uses it to argue about changes in the price level. Instead of, you know, directly measuring the price level, which we could easily do.
Stop pretending Mamdani isn’t terrible.
For those who don’t realize, nuclear brinksmanship is incredibly dangerous, and the fact that it ‘usually works’ does not change that, with at most notably rare exceptions, you shouldn’t do it. ‘The world did not end when I did this thing with a small chance of ending the world’ does not mean you weren’t acting crazy.
Levels of Friction
Yes, when you put it that way it sounds dumb.
The two regimes are inconsistent in theory, but in practice this might be reasonable.
If you allow direct investment in OpenAI and other similar companies, this will seem like a normal investment thing one might do, and will be pitched as such. Investors will put millions of dollars into this.
Whereas if you bet on the OpenAI valuation at Polymarket, you know you are doing something inadvisable, risky and weird, and the liquidity is not so great. It is a bit of fun, a gamble. Yes, you can blow yourself up here if you want to, but you gotta want to.
Jones Act Watch
The Jones Act got selective waivers due to the Iran war.
Unfortunately the combination of ‘temporary and could be reversed any time’ with ‘only for narrow purposes’ means you can’t invest in logistical changes and thus the wins will be relatively minimal, which is why Balsa Research isn’t harping on it. But yes, every little bit helps, and there were good anecdotes even before we looked to expand this to up to 90 days.
Some people were very not happy about that. They were ‘deeply concerned’ that the waiver would be ‘abused’ by foreign shipping companies to take items from one port and then, at lower cost, ship them to another port. The horrors.
They promise to be ‘watching closely.’ Please, all of us, watch closely.
This would be one of many benefits.
That was before we even had an official announcement that waivers were available, and it included NYC-Hawaii.
Indeed. There is a small handful of those capturing rents, who have convinced everyone else they can wield power through union solidarity. But the number of actual beneficiaries is tiny, and the number of people hurt is large, including vastly greater numbers of unions and union members. You can just repeal things.
And you know what? Trump is smart enough to notice this one, too, and is considering extending the exception indefinitely. That’a a great idea, and I have an idea that would be even better, which is to expand the exception indefinitely, too.
That’s what you get with no notice, and no ability to make long term plans. It would be a much bigger effect, across all eligible shipping, if the waivers were permanent.
We are now looking at an up to 90 day extension. At some point, one wonders if it starts to have legs.
The American shipbuilders are mad, and urgently call for ‘preservation of the US shipbuilding supplier ecosystem,’ to protect our sovereign manufacturing capabilities (which largely no longer exist), because you see the United States ‘relies heavily on free-market forces’ via banning everyone else from its markets, whereas other nations ‘protect sovereign manufacturing capabilities as a matter of strategy.’
Why do LNG ships cost four times as much to build in America as they do in Asia, if you are lucky enough to get the American one on time and on budget? One problem is that steel is twice as expensive, mostly due to tariffs, although that obviously can’t explain most of it.
Farmers are among those who hate the Jones Act, since it raises their costs, and farmers do not have the ability to not notice this fact.
Technology Advances
Google will finally allow you to change your GMail address. You get three changes lifetime, one per year, and previous addresses remain as aliases.
Laptop stands are substantial upgrades for many people, in terms of working in a cafe comfortably, as is bringing a full keyboard and mouse. My recommendation continues to be a full desktop, whenever possible, with at least two large monitors.
Variously Effective Altruism
NPR gets two gifts totaling $113 million. There is weird decision theory, where you don’t want to reward the government no longer paying for NPR by spending your own money paying for it, but actually you should anyway if you value NPR because the people cutting it actively want NPR to die, similarly to PEPFAR and USAID.
Tyler Cowen joins Peter Thiel in saying ‘people who promised to give half their money to charity are going back on their promise, and that is good.’
I think this is a no good, very bad position and if you have it there is a very serious problem with your philosophy, decision theory and approach to life.
I also think this goes hand in hand with Tyler Cowen claiming that our intervention in Venezuela was good as well so long as in this particular case people in Venezuela were better off, despite how much our actions broke norms and set terrible precedents.
Indeed, I would say we are already seeing, on a vastly larger scale, the consequences of what happened in Venezuela becoming a precedent and setting expectations. It’s ugly, even if you are foolishly a Causal Decision Theorist and can only look at forward consequences. Localized Act Utilitarianism is foolish.
I can totally understand arguing that it is a mistake for wealthy people to give away half of their money. You can only give away so much money efficiently before you face decreasing marginal returns, and it requires large investments of time to do well, and also potentially makes you a target. If you feel you must give away [$X] no matter what you lose your quality controls. The apparatus risks capture by ideologues or thieves and has poor legal structures that don’t protect you. Investing your money in for profit enterprises that do pro social things is arguably a far better way to deploy tens or hundreds of billions, or perhaps you should be doing out of the box things like buying and releasing major medical patents.
I have many disagreements with Elon Musk, but his plan of ‘I want to use my money to build great companies and make everyone wealthier and go to Mars’ is totally valid. My differences are more about whether projects like xAI in particular are wise ideas, or in his non-business actions.
I have not signed such a pledge, and do not plan to give half my money away. It’s fine.
However, if you make a pledge to give away money, then you should honor that pledge.
And when people break pledges, not because they need to but because they choose to, after benefiting from the halo of that pledge, you do not say ‘good’ because that is a terrible, no good thing to do. If you either do this, or you praise this, then you bring shame upon yourself and upon your house, and you weaken the very foundations of our society.
This is indeed exactly the kind of logic that leads to Sam Bankman-Fried, the argument that if you can ‘do good’ by breaking your promises then you should do so.
There is a case for asking that question, and there is a case for not asking it.
Tyler argues that the great philanthropists of past ages, like Rockefeller and Carnegie, spent when they saw opportunity and people met standard criteria, rather than up to a threshold, and did so with less overhead.
I love that approach where it makes sense, but it has nothing to do with whether one should honor pledges.
Also, you could honor your pledge that way in a simple fashion: You do that, and then use any leftover funds in other ways that scale, in the worst case you Give People Money, or donate it to the Treasury.
If you’re screaming ‘you can do better?’ Good. Do better.
What about ‘you should fund rationalist or EA things’?
The problem is that those things take a lot of investment to understand, if you want to differentiate the good from the bad, unless you find an agent you can trust at this scale, and also the room for funding most options rapidly becomes an issue when you’re talking about 11 or more figures.
That said, if someone is looking to move 11-13 figures to do good, yes, I would be willing to help out for a remarkably small percentage of that, especially if we are willing to not let the perfect be the enemy of the good and I get to use a portion of it for some aspect of AI notkilleveryonism related causes, but I’m happy to confine that to technical research if necessary.
Elon Musk claims Starlink has done more to lift people out of poverty than any NGO so I thought for fun I’d have GPT-5.4 do some comparisons. ChatGPT doubts Starlink even breaks the top 300 corporations in terms of poverty reduction (its top picks? Foxconn, Apple, Walmart and Samsung.) Starlink really is not that special on this one, but also capitalism is what solves this, and only maybe 10-30 NGOs have done more than Starlink. That seems plausible to me.
Scott Alexander recently argued against the concept of telescopic altruism, in ways I found highly unconvincing. Ben Hoffman points out that to succeed, telescopes need good lenses, and offers us a lot of intuition pumps for what is typically wrong with those who purport to focus on helping distant others, and some of the reasons why by default it makes sense to first put your own local house in order, and some of the failure modes that happen if you don’t do that. You will probably disagree with some of his examples, as do I, but do not let this distract from the clear larger message.
Copious Free Time
If you’re single and childless and you say work means you have no time for socializing, then either you’re approaching 80 hours, you have a serious health problem, some massive other mandatory time sink weighing on you, or you’ve made a choice.
This is one way to realize that you have a serious health problem, in which case your focus should be on fixing that.
The Lighter Side
If you ever make a mistake, join me in remembering this mistake, and then feel better.
Note that headlines are not like articles, so it’s more like 1-2 people have to make the mistake than 3-6 people would in the article body.
Marginal Garfield is about as good as one could hope it to be. Rationalist Garfield is a tier below where one could hope it to be, which is still good enough that my kids laugh and I can use it to try and explain concepts.
Suspicious?
It is April 2026 and Elon Musk is still attacking Nate Silver based on the prediction that Hillary Clinton was 70% likely to win the 2016 election. My lord.