Remember when, two weeks ago, I told you about my internal struggle when packing for trips? Remember when I said that I always pack too much and yet still envy the carry-on-only people? Well, right after I told you about it, I hopped on a plane and arrived in Mexico… without my luggage. That only arrived a day later.
Remember, when 6 weeks ago, I told you about the black shirt I bought twice and lost twice in San Francisco? Well. I was in San Francisco again and, again, brought the black shirt and — I didn’t lose it. I still have it. But! On the way back, I boarded the plane, sat down, was told to get up again because we had to deplane, waited for two hours, boarded another plane and realized I forgot my hoodie on that first plane.
What I’m trying to ell you is this: if you haven’t, you should subscribe to this newsletter. Only to keep up with the high-stakes lost-and-found drama in my life.
Quinn and I were guests on the Latent Space podcast. We recorded last Monday, in San Francisco, and I think it’s going to be one of the things I’ll tell my grandchildren about: in the 2020s, I was a guest a podcast, in San Francisco, talking about AI, and a coding agent I was building, and, yes, you read that right, the thumbnail says that we were “building the coding agent god” and, I swear to you my grandchild, we never mentioned the word god in this conversation, but it was a lovely and enjoyable conversation and, man, I’m telling you, I had a ton of fun.
Amazing: The Making of a Market Maker. Started off all the things one can find in other rich-business-men profiles too, but then there’s this: “In the corner of the office sat a $3,000 solution nobody wanted to touch. The Olivetti Programma 101 weighed 20 pounds and looked like an oversized cash register. It had a numerical keypad, a slot for magnetic cards, and a built-in printer. […] When Peterffy volunteered to tackle the unused machine, nobody objected.” And then there’s this: “Most clients wanted the same reports that compared securities across various metrics like price-to-earnings, book value, and earnings growth rates. Peterffy wrote programs in Fortran, fed stacks of punched cards into room-sized IBM mainframes, and waited as the machine hummed and clicked through calculations.” And it kept getting better: “By 1976, he commanded a team of 80 programmers—one of the largest financial coding operations in the world—and Jarecki began bringing him into meetings that had nothing to do with software.” Very fun read. Highly recommend it.
Brandon Smith with Thoughts on Testing. Reasonably sure that I linked to this last year, but I re-read it and nodded (presumably again). I’ve said it here before: I did a 180 on tests in the past few years and now they’re just something I might do.
This website is a statement: News Minimalist. What does it do? “Today ChatGPT read 30082 news articles and gave 22 of them a significance score over 5.5.” Trying to filter out the noise of daily news. I’d love the see the same statement, but made from the other side: it should look back at the last four weeks, four months, four years and flag how much of the daily news turned out to be insignificant and just noise that was drowned out a couple days later.
“Nowadays, you can specify the task you want to automate, fire off Codex or Claude Code, check back in 30 mins and have an almost-perfect result. To give you some ideas, some things I have automated or do regularly using LLMs: […]” from You’re not using LLMs enough. I’d love to do more like this, but I have to admit that I get lazy whenever it gets to deployment.
This should be some sort of official personality test: I’m Not a Robot. Hats off, but I only made it to level 5.
How to build your first forward deployed engineering team: “And remember: this is painful. In McGrew’s recent appearance on the YC pod, we counted eight mentions of ‘pain.’ If you master the loop, you can achieve power-law outcomes. But it’s painful and hard to pull off. You can easily veer into customer support or consulting, which dilutes the strategy and balloons your cost structure.”
I’ve never worked with R but I sat next to people working with R and what I heard didn’t exactly convince me to never trying it, but it also kinda made me think that I’ve heard enough. So when I started reading this — If all the world were a monorepo — I thought: ahh, R. But I ended up being fascinated. (And surprised! Only after reading did I realize it was written by my ex-colleague Julie!)
“How I, a non-developer, read the tutorial you, a developer, wrote for me, a beginner”
Fascinating piece of writing: “I ran out of money a year ago, spent the last of my savings on a prostitute in Hong Kong, and became a commie.” It’s well-written, isn’t it, and also odd, strange. I bet that in ten years you can point at it to explain the San Francisco Artificial Intelligence Zeitgeist of the Early 2020s.
And when you do point back, make sure to point them at this piece too: “The AI gold rush has sparked a vibe shift in San Francisco. The city is flush with money again after the post-ZIRP recession of 2022. Cracked 22-year-old coders are telling the world they’re going to ‘solve hurricanes’ and the ‘national debt.’ Lurie is mayor, nature is healing, the technology brothers are back with a vengeance. Take a look—$100 million salaries, glitzy hype videos for fundraises, lavish parties with dress codes—Silicon Valley is swelling with Trump-era opulence—blustery, spendy, and male.”
It made me think of this passage from Hunter S. Thompson’s Fear and Loathing in Las Vegas: “Strange memories on this nervous night in Las Vegas. Five years later? Six? It seems like a lifetime, or at least a Main Era—the kind of peak that never comes again. San Francisco in the middle sixties was a very special time and place to be a part of. Maybe it meant something. Maybe not, in the long run . . . but no explanation, no mix of words or music or memories can touch that sense of knowing that you were there and alive in that corner of time and the world. Whatever it meant. […] There was madness in any direction, at any hour. If not across the Bay, then up the Golden Gate or down 101 to Los Altos or La Honda. … You could strike sparks anywhere. There was a fantastic universal sense that whatever we were doing was right, that we were winning. […] So now, less than five years later, you can go up on a steep hill in Las Vegas and look West, and with the right kind of eyes you can almost see the high-water mark—that place where the wave finally broke and rolled back.” Then again, there’s few other pieces of writing I think of as often.
Lee Robinson: Things I Believe. Very good.
I’ve read this one before, a few years ago, but ended up re-reading the whole thing again: Tim Ferriss with 11 Reasons Not to Become Famous (or “A Few Lessons Learned Since 2007”).
“The U.S. Secret Service has announced that, following a monthlong investigation, it has carried out an operation which resulted in the dismantling of a network consisting of more than 300 SIM Servers and 100,000 SIM Cards at multiple sites across New York and New York City, all within a 35-mile radius of this week’s United Nations General Assembly.” Dude. Look at the photos. It’s “capable of sending over thirty-million text messages per minute, anonymously.”
libghostty is coming. Hell yes.
“It seems pretty well-accepted that AI coding tools struggle with real production codebases. [...] The common response is somewhere between the pessimist ‘this will never work’ and the more measured ‘maybe someday when there are smarter models.’ After several months of tinkering, I’ve found that you can get really far with today’s models if you embrace core context engineering principles. This isn’t another ‘10x your productivity’ pitch. I tend to be pretty measured when it comes to interfacing with the ai hype machine. But we’ve stumbled into workflows that leave me with considerable optimism for what’s possible.” I’ve become skeptical of subagents and I’m also very skeptical of encoding these workflows in hard UIs, but there’s so much other practical, true things in here. Very good.
When I saved this link it said “I reverse engineered San Francisco’s parking ticket system to show you exactly where each parking cop last wrote a ticket.” It still does that. But now it also says “In rare lightning speed, the SF government changed their site within hours of this site going live. I can’t get data from it anymore.” I’ll see you, Icarus.
Please read the Google section in the Steve Ballmer Wikipedia entry: “At some point in the conversation Mr. Ballmer said: ‘Just tell me it’s not Google.’ I told him it was Google. At that point, Mr. Ballmer picked up a chair and threw it across the room hitting a table in his office. Mr. Ballmer then said: ‘Fucking Eric Schmidt is a fucking pussy. I’m going to fucking bury that guy, I have done it before, and I will do it again. I’m going to fucking kill Google.’”
