2025-09-12 23:05:00
I guess the fact that I'm back means that y'all didn't hate this. If you do, tell Uri and Jehan!
Matt Holden and I are doing a YouTube show about building with AI called --dangerously-skip-permissions. The first episode is “How did we get here?”. Matt and I have been having 1:1 conversations for more than a year now about what tools we’re using and how we’re using them… and now we’re having those conversations in public. I especially enjoy the way that Matt is able to connect what’s happening with LLMs today with previous eras of computing innovation. Give it a listen if that’s your thing!
Matt Holden on Markdown coding
OpenRouter has market share by LLM model. Interesting and unexpected in some ways!
On fact checking with AI. I really enjoyed this one. I have a draft blog post in my head called “Vibe Craft: How to do serious work with AI” but every time I try to write it, it falls flat. This is spiritually related to that.
Office building visits are up among people that live less than 5 miles from their office. As someone who made major life changes during the pandemic, I feel the pang of regret.
“A soldier will fight long and hard for a bit of colored ribbon.” — Napoleon
“Everyone has a plan until they get punched in the face.” — Mike Tyson
Get the original Good Tokens and other excellent items at jdilla.xyz
2025-09-10 22:15:45
I'm convinced that some amazing ideas never catch on because they have a terrible, unmemorable name. One of those ideas is the (principle underlying the) Wason Selection Task. If you can think of a catchy name for it please send an answer by email or comment, I am desperate for this.
Most of the people in my life are conflict-avoidant (I suspect it's most people in the world, honestly, but there's a chance this is just an example of self-selecting groups).
Accordingly, it's relatively rare to hear anyone directly contradict anyone else on a matter of importance – if someone passionately exclaims X, mostly other people will not-disagree out loud, even if they passionately disagree internally.
The upshot is that the people with the most confident views can't easily tell how popular or unpopular their opinions are with everyone else; I think often they go around feeling gobsmacked that 1) everyone agrees with them about X, and yet 2) the world still doesn't understand.
Of course, there are also lots of things that everyone in the group does agrees about – "pizza is delicious," "Jehan is handsome" – so it's not like "nobody contradicts me when I passionately advocate X" is a sign that people necessarily disagree. But I still think there's signal to be found here, by trying to figure out which things you go into conversations assuming everyone will agree with, and questioning whether they might feel differently if you came in less strong.
A nice thing that is easy to forget to do: when someone does (or is) something nice in your life, think about who introduced you to that person and send them a note.
E.g. A few times now I've connected two people who live in the same city, and discovered only years later that they became good friends, sometimes without even knowing they even met up in the first place. And I'm sure I've similarly forgotten to thank people for an intro that led to a friendship, I don't think it happens unless you actively try to remember to do it. But it's so nice on the receiving side to hear "by the way, me and Jo hang out all the time now, thanks for putting us in touch!"
One additional benefit of sending a note to thank someone when a different person does something nice in your life is that it gets through the defences of the Person Who Doesn't Easily Take Compliments – if you tell them they're great they'll deflect, but if you thank them for an intro to someone great they can more easily accept the compliment.
2025-09-08 22:15:12
There's a lot of ways to (potentially) ruin your life at 18: addiction, bad marriages, get pregnant without wanting to, commit a crime that gets you locked up for decades.
I think a lot of us have a very uncomfortable relationship with the rights of 18 year olds. It's our societally-defined cutoff for Adulthood, and modern western individualism fundamentally rests on the premise that all adults are equal (except for some exceptions).
And yet, ~everyone over the age of 19 thinks that their 18 year old self was dumb and made very bad decisions.
I hear people making all kinds of ad-hoc arguments about this, often citing the claim that our brains aren't fully formed till 25, though to my knowledge that's a made-up cutoff at an arbitrary point on an asymptote.
The arguments are usually very selective, claiming that we can't hold 18-25 year olds responsible for X, without asking what that implies about their ability to Y.
Anyway.
When I think about the ways that people I know personally have seriously and negatively impacted their lives for decades, a lot of it comes down to signing contracts at age 18.
Some of them committed themselves to government-sponsored programs that irrevocably tied them to terrible jobs for a decade or so; many many others got into hundreds of thousands of dollars of debt, often for graduate degrees that have proven almost-worthless.
My friend group is unfathomably selection-biased, and shouldn't count for much. But I recently stumbled on a personal finance podcast[^1] where the (generally working-class) guests describe their financial issues.
From the few episodes I've listened to, the original sin for many of them is a massive student, car or medical loan they took out at 18.
I'm not sure where this leaves me. I understand that if you ban long-term-contracts for 18 year olds, the alternative isn't magical free money for 18 year olds, it's that they wouldn't be able to get that house/car/masters degree at all. It's obviously illiberal (in the classic sense) to prevent people freely agreeing to contracts, and I care a lot about classical liberty.
And yet.... something sits uneasily for me about allowing someone who days or months previously was not allowed to do basic things without their parents' permission suddenly being able to commit themselves to obligations that can haunt them for longer than they've yet been alive.
Could we have a phase-in? Could we cap how much you can borrow at 18, or how many years of your life you can commit yourself to something, and then raise it a bit every year till your mid-20s?
I don't know, I just don't know.
[^1]: I'm not naming this show because I partly hate it. The host is aggressively rude to the guests, which I guess could be justified if it functioned as a wake-up-call for the guest, but seems like it's mostly for the audience's amusement. He also reads ads for financial services companies that I think are a terrible idea for a financially un-savvy audience.
2025-09-05 23:04:17
Hello! I'm James, Uri's friend and Atoms vs. Bits's biggest fan. Uri and Jehan have graciously invited me to share my weekly-ish list of things that have caught my attention. Let us know how you like it!
Replacing lawns with wildflowers 💐. When I’ve made it, I won’t tell anyone, but there will be signs.
Cate Hall on how to increase your surface area for luck, which is one of the biggest things I learned from Henry Oliver’s book on Second Acts. Cate is quickly rising up the list of people whose work I rush to read. Along the same lines: How to Get Ahead in DC.
“Even the context has context”. Wherein Soren blows my mind and sells me on decentralized edge intelligence.
Should I have kissed her? Some how I missed this one in August of 2022. It’s my favorite type of Uri post.
How can you not love this? A 3,700-year-old Babylonian clay tablet holds a trigonometric table more accurate than any today. Humans are amazing and beautiful.
Drones are downing helicopters.
Yucca man. I’m a sucker for “does this Bigfoot like creature actually exist” stories (see season 1 of the Wild Thing podcast), but this one also has so many great Southern California places in it. Like taking a mini vacation.
Nuclear batteries. “A 157W Voyager-based RTG that launched in 1977 will produce about 88W today.” The clean up problem seems insurmountable.
Noah Smith, Dan Wang, and James Cham talk about Dan’s new book Breakneck.
Why Swiss Kids Walk to School Alone. This is one of the things that made me fall in love with Switzerland. They do this as 5 year olds! Part of it is safety but part of it is teaching agency. The walk to school is a part of the education. This should be our aspiration for American neighborhoods.
Your idea sets the ceiling for your videos potential and other good advice from Paddy Galloway.
Chips on shoulders put chips in pockets.
What’s the steel man case for formality?
What does our society overemphasize now in a way that will seem silly in 25 years?
The secret to engineering is embracing that getting new errors equals progress.
2025-09-03 22:14:06
When people ask me if I play chess, I usually say "I know how the pieces move." And I think this is much more helpful than e.g. "I'm not very good", which could cover a very large range of abilities depending on modesty: "I know how the pieces move" succinctly conveys 1) I am able to play chess with you 2) I don't know any strategy.
I often want similar phrases for other activities, ways to unambiguously convey how good/bad at something I truly am, without the added layer of self-presentation questions. I suppose in this case it's easier because I'm bad at the thing: "do you play chess?" "yes my ELO is 2000" would also convey true and precise information, but is harder to say without sounding like a brie.
Organizing group-trips is incredibly costly for the one person who organizes, both in terms of time/effort/energy and even financial risk or carrying costs. In theory you could split the work between multiple people but in practice this never seems to work. It's also repeatedly the same person in each group who does the trip-organizing.
How would you feel if that person charged, say, 20% of the trip cost as a tip? Friend-of-the-blog H says everyone would hate this, and I assume he's right. But the alternative is not having trips at all, because nobody is motivated enough to organize them.
2025-09-01 20:07:57
The most important thing I learned from weightlifting is how many of my constraints are fundamentally in my head.
If you had asked me beforehand I would have said that "how much weight can you lift today?" is a physical fact you can nudge the tiniest bit up/down through willpower, but it's largely a story about the mechanical ability of your corporeal form.
Until one day I set up my rack to squat 240lb, and did a few reps thinking "damn that was harder than it should have been, I'm out of shape :(", and then realized I had counted wrong and squatted 270lb by accident.
I know for a fact that I could not have lifted it if my brain knew there was 270lb on the bar, and I don't mean that in a conscious psychological sense, I mean my brain would not have allowed my knees to bend if it had heard the number 270 when approaching the bar. But it turned out my body was perfectly able to lift that much.
I'm not sure exactly how this works, psycho-somatically: maybe your brain sets conservative limits on what it allows your body to do, maybe to protect you from injury? But it's wild to know know that my body is capable of vastly more than my mind will usually let it do.
(I have no evidence at all for this but I wonder if those stories about e.g. a parent lifting a car single-handedly to save their child is partly about this, their brain lets them use their full body power because of the emergency.)
I don't want to over-generalize from this insight, there are still plenty of things in the world that are not just brain-constraints, but having even one such vivid example of your brain vetoing something changed how I think about both brain-body connection and the self-imposedness of my limitations.
There's a smaller version of this that happens super regularly, which is: I don't feel like doing the last reps or sets of my workout, and sometimes I let myself just be lazy, and other times I buckle up and do it anyway.
And at some point all your progress in lifting comes down to how often you're able to make yourself do the slightly uncomfortable last reps/sets. It is very hard to do it, and it makes all the difference (even if you're 80/20ing the lifting in the first place -- there's a huge difference between 80/20ing and 50/10ing).
Lifting really highlights the classic tradeoff between doing things in groups and doing things alone: it's really hard to make yourself do the final difficult bit by yourself, whereas if you're around other people they can shame you into pushing through even when you don't want to. Working out (or simply working) with other people is inconvenient and expensive and annoying, but if they can make you do the hard bits that ultimately compound into your progress then it might be worth it.
Weightlifting has this unusual and lovely property that there's numbers and you record them and they go up.
This isn't true forever, you can reach a level where it's really really hard to progress, and there are bumps along the way even before that. But fundamentally, for a pretty long time, you can just keep showing up and doing the thing and watch your number get higher without any supernatural intervention.
And so few of the things I do have this clean linear relationship between effort and outcome that it gets hard to remember this. But weightlifting gives a much more visceral, grounded intuition to the philosophy that 90% is just showing up (consistently for a very long time).
When your progress plateaus, you need to find (and fix) your limiting factor. Weightlifting is full of surprising limiters, e.g. grip strength – I would never previously imagined that you could add 10% to a deadlift by improving your grip, but you can.
It's maybe less surprising than it first appears that it's these weird auxillary things that end up breaking your plateaus, because your training is focused on the Big Things to begin with, and overall that's where most of your progress comes from.
The plateau happens because you've got your Big Things to the point where they're no longer your limiting factors, and then you need to find smaller things to fix to make progress again.
It's much much easier to re-acquire muscle you had previously than to gain it for the first time. I'm not sure what to do with this information, except that I wish I'd started lifting much earlier in life, when mucle-gain is easier. Maybe it argues for doing some intense bootcamps (for various skills) a few times in life, so you can get to a skill ceiling that you will later find it easier to recapture? (Maybe you have to train for a couple of years and then do the intense bootcamp to really get the most out of this, I don't know).