I used to write a link roundup to some friends regularly. I’m hoping to get back into the habit, but now for a larger audience!
Karri Saarinen on learning about the local fauna by his Finnish cottage:
I spend part of the summer at a cottage in Finland, by the Baltic Sea. There is all kinds of animal and plant life around, and instead of asking OpenClaw, I bought a couple of books about the local fauna and flora.
Now when I encounter a snake or some other animal, I may already know what it is. And if I don’t, I go back and check the books.
I have small moments like this everyday now, usually less dramatic, that remind me how books have made daily life feel a little more magical.
Karri is the founder of Linear, a software company with some of the best SaaS design out there, and I really like his beliefs on AI.
He deeply understands how to use it, but thinks user interfaces are still important. And this tweet is pretty old (it’s from the spring), but I think it’s aged well: AI can help you shortcut getting something done in a new domain, but it doesn’t replace actually learning it deeply.
In the last few months, I’ve learned to do video editing in CapCut, Premiere, and After Effects. What I’ve found is that while the models are getting better and better at video editing, using them is still like using a hammer where you need a chisel (or, as someone once said, ‘like asking your kid brother to repaint a room.’) In the end, I still had to learn how to use this software to get anything done.
When I think about AI and software, I am reminded a little bit of the sewing machine. I own one and use it for alterations. It’s an extremely effective time-saver, and it can do all sorts of nifty things. But it’s not a panacea. You still often have to hand-sew or at least carefully manipulate a piece of fabric as you feed it through the machine. And similarly, I think SaaS tools will remain important for experts in any domain for a while…
The old world is dying. Occasionally I’m lucky enough to meet extremely agentic people, and Jasmine is one par excellence. But I think it’s telling that she writes about how much joy she found in leaving Stanford, living in the real world, finding an apartment, and being free of the strictures of student life. I think she also might not like institutions very much.
And so in this article, she claims the old world is dying. Every new grad should meet more people, go do something new, work for a startup, ask more questions, and break more rules.
This is a way of working on the West Coast that I really enjoy — that instead of joining an institution, you can create the one you want around you. You can see it in career outcomes. Strong Princeton graduates join Goldman Sachs or Jane Street; strong Stanford graduates… well, it’s hard to generalize, because they all go down their own paths, building their own tiny cultures around them before they subsume the whole world into them.
But I wonder if there’s a bit of ‘this advice is not for everybody’ in this article. Rather than a guide to all new graduates, I would treat it as one for the most high-agency folks. There’s still a lot of alpha in joining a rising institution, playing your cards well get promoted, and getting mentorship from people who have done it before!
Why do Japanese companies do so many different things? Why Japanese companies so often operate as conglomerates that sell all sorts of weird, different things like motorcycles and pianos, when American companies tend to focus on one thing?
David Ok’s argument is that the Japanese system of lifetime employment means that Japanese companies are really run mostly for the benefit of the employees.
They don’t offer high returns to shareholders, have strong unions and few visionary leaders, and are designed to insulate themselves from many outside pressures that might, for instance, require layoffs. Pay for workers, however, is based more on tenure and company performance than individual worker performance. He argues that this made Japan excellent at catch-up growth but also caused it to lag behind anywhere extreme innovation was required.
It’s an interesting argument, but I don’t really have an opinion on this, to be honest. What stuck out to me as I read it was that the J-firms reminded me of Jane Street, a company that roughly operates on the same principles: it trains generalists, is run mostly for the benefit of its employees, and cares more about continuing to exist than returning high amounts to shareholders. I’m not 100% sure what’s going on here. Maybe I’ll write more thoughts a different time…
I love technology, and I love the amazing new devices that Elon Musk wants to bring into the world: brain-computer interfaces, electric cars, spaceships. And yet on the scales of justice, against all the lives they’ve saved or will save are the people he killed by defunding USAID.
I don’t think it’s an easy read, but I think it’s an important one. It’s easy to think of policies only on a global scale. And so I’m grateful for the work done by reporters to find the individual stories, because they hit much harder.
This post on rewriting Bun in Rust is probably meaningless if you are not a software engineer, but if you are, I think it’s one of the best pieces out there on effective coding with AI. There’s still a lot of skill involved! They were successful in rewriting it, but it also took a lot of care to create a system that would let them produce safe, verified code.
I’m reminded a bit of this piece on how LLMs reward expertise, which I also appreciated:
The most important skill in prompting is expertise in the domain you’re prompting for.
It’s about how Terence Tao can pull far more out of LLMs in math than you and I can. It was written before the Navier-Stokes discoveries, but even in those cases I don’t think the LLMs got there without some mathematical guidance — someone had to push the AI into thinking about the right part of the latent space. So far, I don’t find the breathless takes about how AI is automating everyone’s job very convincing. It still seems to me that like it acts as a 10x factor, and that the knowledge of the top domain experts is still really important!