You say that these numbers mean dial it down. I say they mean dial it up. You haven't gotten through. There are people you haven't persuaded yet. These number mean dial it up. Otherwise you're like the French radical, watching the crowd run by and saying, "There go my people. I must find out where they're going so I can lead them."

I ran into a great post from Karri of Linear on product management. Most of what he says warrants further discussion and digestion, but I wanted to chime in on the rhetorical question which prompted his lovely response:

Consider this: if “talk to customers” is the biggest secret to product success, then why aren’t more products successful? Why are so many founders unsuccessful? What explains PMs who’ve been talking to customers 5 times a week for years, without ever making products that win?

"Talk to customers" has become, to me, the same kind of truism as "charge more" and "users don't care about your codebase" — easy to deploy and sound clever in doing so, but dangerous in their banality.

At risk of deploying a metaphor (c.f. The taste of beer), customer feedback is something like a GPS: faster than driving aimlessly, slower than knowing the streets like the back of your hand. Especially in commodity/parity-style industries (like product management tools in Karri's case, or email in mine) customers make for good historians but poor futurists, and certainly they won't do the hardest and most important job of identifying your leverage points for you.

None of this is to say you shouldn't talk to customers: you should! But it should be neither the first nor the last step in your process: if someone needs to talk with people to figure out what to build next, it means they have insufficient vision; if someone needs to talk with people to figure out if something is truly ready for GA it means your org has insufficient conviction and process.

Customers are great at informing a product: reminding you of edge cases, blind spots, prior art, and so on. But there are many incredible things that can be built without information, and the annals of history are swamped with products that had very well-informed backlogs and yet no leverage or power to stop their own demise.

It is rare and difficult for a long-running series to reverse a downward slide. There are simply too many sources of gravity: writers run out of obvious plot lines and have to start re-hashing, Flanderizing, and escaping the confines of the show's existing logic; actors change, and audiences grow tired of what was once novel; it is simply hard to capture lightning and put it back into the bottle, and the commercial demands of the television industry incentivize showrunners to turn their works into shambling ghouls (another year of steady work for the crew, another year of solid residuals, another year of fun with coworkers with whom you've grown fond) rather than a surgical and cohesive narrative arc.

(All of these things are true of books series as well: take, for instance, the Berlin Game nonology which started with promise and ended with Deighton left with no other choice but to raze his characters [and goodwill] to the ground.)

Slow Horses bucks — or at least bucked — this trend. Its fourth season is a triumph: it rights every slight misstep of Slow Horses (Season 3) and Slow Horses (Season 2) (both solid seasons in their own right, but not nearly as great as the gemstone first season) and delivers the best six hours of thrilling, compelling spycraft in a long time.

Some miscellaneous notes:

• One of the keys to Slow Horses' success is the legitimacy of its stakes. There's no Game of Thrones-esque flippancy with which characters are killed off less out of narrative cohesion and more out of a desire to shock the user, but nobody (besides, presumably, Gary Oldman's Jackson Lamb) is safe from death: this is something we're reminded of every season, and it gives a sense of actual danger and urgency to every shootout or chase.
• Speaking of Jackson Lamb: this season probably had the least screentime from Oldman. This can be a dangerous gambit — it paid off here, and his absence (modulo the handful of convenient deus ex machina opportunities he's afforded) is felt but not painful.
• Spy stories must thread a very tricky needle when it comes to realism. It's really hard to be entirely grounded in craftwork and realpolitik unless you're le Carré or The Americans, and it's also hard to be bombastic and exciting in the traditional language of television action sequences without feeling like you've suddenly entered a different, sillier show. (This was, I'd argue, the main flaw in Season 3: we suddenly entered a Call of Duty commercial in the back third of the season.)
• Season-long arcs dedicated to flushing out a character's backstory can, similarly, be hit or miss: shows need to earn an audience's trust that we actually care and empathize with them enough to follow them down unexpected corridors, and actors need to be able to shift from being relatively one-note to full-fledged protagonism. Jack Lowden passed with flying colors.

A little over six months ago, I wrote Notes on Zed. My conclusion at that time was that Zed made a lot of great choices and felt really good to use, but lacked parity with VS Code's feature/ecosystem to its detriment.

Six months later, I spent a few days using it as my daily driver to see what had changed:

• It's still really, really fast — and in fact, the delta between VS Code and Zed has grown even larger (either objectively or subjectively) since I last used it.
• They've made a good amount of progress on Python support, but it's still not great:
  • Pyright is so bad as to almost entirely disqualify it from being a Python editor, and extensions are still so nascent so as to be an exercise in Linux-esque experimentation rather than outright pragmatism.
  • Similarly, the LSP architecture makes it difficult to employ both Ruff and Pyright.
  • I timeboxed an hour to figure out how to get Zed's task runner to recognize my pytest suite; I was unable to do so.
• Multi-buffer editing is great though it breaks my brain a little. It seems obviously like the correct interface change, not unlike so many other little lovely decisions Zed has made.
• There are a myriad of memory leaks. I diagnosed two during my time with Zed: one, due to it trying to index a SQLite database within the repository; another, due to inline assistance.
• I miss GitLens and ErrorLens much more than I would have expected.

This list probably sounds more negative than my experience warrants. Zed's core value proposition — performance and productivity — is still enticing and tangible, despite some of my fears about them pivoting towards AI and collaboration tools. The team is iterating quickly: I'm excited both to try Zed again in a few months.

What surprised me the most about this time was how much I missed Cursor, which I've been using ever since Using Cursor to port Django tests to pytest. Common wisdom is that the AI layer for all of these tools is essentially a commodity without much differentiation, but I found myself missing the aggressive autocomplete and file-level refactoring that Cursor is (in)famous for.

So: at least for Buttondown, back to Cursor. But the Zed team is doing really good work, and the things that they're investing in feel much more durable than the things their competitors are investing in. If I were to bet, this time next year Zed will be my full-time editor of choice.

We've had Lucy for two weeks, which qualifies us as experts, which means it is time to write about parenthood. (In all seriousness, consider the below descriptive and not prescriptive: mostly, it's a notepad filled with things that were remarkable or surprising or divergent from popular consensus.)

• American pop culture puts too much emphasis on the onus of diaper changing. Diapers are easy; swaddles are slightly more difficult (we've gone with what we refer to as "the three corners method"), and the true endgame foe is the onesie. Sleeves are difficult!
• We were nervous about Telemachus not adjusting well to having a baby sister in the house; instead, he is enamored with her, and already quite protective, and mostly just annoyed at my wife and me for getting up so often in the middle of the night.
• Common wisdom says that babies sleep for eighteen hours a day, the mechanics of which are hard to really internalize until you're amidst it. There is more downtime than you expect — it's just that the downtime comes in a random, somnambulant staccato.
• There is a surprisingly deep amount of variation amongst baby bottles. (We've gone with Dr. Brown; the form factor of the bottle itself is easy to wash, and the nipples are the right size.)
• You will spend so much time going up and down stairs; you will spend so much time washing (bottles, pumps, blankets, etc.)
• We have never been more vividly aware of our luck and fortune. DMs and texts and emails, hand-knit hats and blankets, embroidered sweaters, Doordashed lobster rolls and hand-delivered pots of 冬瓜海鲜羹, words of advice and encouragement: all of these things are glittering gems; our coffers overflow.

Chatted with Jess and Jeremy about a whole slew of things, from pricing strategy to terrifying and arcane differences between various Markdown parsers (including why I hate Slack.) They were particularly kind and gracious given that the amount of production-quality Rails I've written is ~zero; you'll enjoy this even if Ruby/Rails is not your cup of tea.

In descending order of importance:

• I had a daughter, Lucy, the light of my life.
• I launched a tiny side project, handbook.directory.
• I cautioned against equivocating profile with praxis and overreacting to vendor outages.
• I jotted down some notes on Bluesky, and post-Twitter social media writ large.
• I suggested always using an enum instead of a boolean and keeping your marketing site separate.
• I read and opined on Creation Lake, High Output Management, and Cable Cowboy.
• I jotted down some notes on RSS in Next 14, MD5-based constraints in Django, and Founder Mode

I hope you are well; I, for one, am bleary-eyed and blessed.

One of the many parts of Ocean's Eleven's terrific legacy is the mistaken notion that all you really need for a film to be smooth and pleasant is a sufficient number of A+ actors playing some effervescent version of themselves.

This is, of course, not the case: Ocean's Eleven has many imitators and few peers (Logan Lucky being one of the few), because dedicating a film to the pursuit of pure pleasure is more difficult than it looks, and made more so by hackneyed attempts to shoehorn in an Albanian mobster subplot.

Parts of this movie feel deeply successful. The first few montages of Old Clooney driving through sleepy wintertime New York is borderline pornographic; Austin Abrams' four minute monologue is a tour de force that manages to hit a Linklater-esque melange of earnestness and humor; Clooney and Pitt sparring with one another is, even in a not-great film, fun.

But the first act falls off a cliff and into the realm of Red Notice-tier "what the fuck are these setpieces, anyway?" quicker than even a cynical viewer might expect. We are treated to the world's most bizarrely choreographed chase scene; an ethnically ambiguous "Club Ice" with a pager-sized MacGuffin; a too-long shooting sequence involving three people walking straight ahead and shooting blindly as if they're redcoats on an FPS with the difficulty turned down to zero.

Your love (or, rather, your acceptance) of this movie will hinge on just how charmed you can be Pitt and Clooney, gamely smirking and chuckling at the increasingly-vacuuous goings-on while perfectly well-clad.

For me: they elevated the movie, but not enough to warrant rating it anything more than a not-unpleasant way to pass the time.

(One little needle to end on: something about this movie's relentless refusal to commit to a tone reminded me of the MCU and its awful determination to cycle viewership through irony, slapstick, CGI, and pathos all in the span of a single over-produced five minute scene, so I was satisfied to discover that the director and writer is most well-known for the most recent few Spiderman movies. At least he didn't try to cast Tom Holland, too!)

The majority of the best people I've worked with in industry — in either an individual contributor or management capacity — are not Meaningfully Online. Their Twitter profile (if they have one) is anodyne and focused on their city and personal interests; their LinkedIn is threadbare.

Building an online reputation from yourself is, in general, a good idea — one that requires a distinct skillset and yields a distinct portfolio of rewards. In particular, it's a great idea if you plan on persuing freelance work, educational / thought leadership work, or building your own business.

But if you're not interested in those career paths, becoming truly impactful in your field largely requires doing the kind of work that doesn't make for banal writing or pretty mockups:

• deeply understanding customers' problems and business goals;
• becoming an excellent communicator and mediator;
• having near-cursed levels of domain expertise.

These skills rarely go viral; the practitioners who have them never need to.

Lucy Frances Duke was born this week: on September 24, 2024. She came a few weeks early: a purported due date of October 13 was moved up a few weeks due to Haley having high blood pressure (asymptomatic, but slightly elevated nonetheless), but she came in at just over seven pounds and with the most perfect cheeks I have ever seen.

The first few days have been a bit of a glorious blur: I think fatherhood is one of those rare things uncapturable by prose. What I can say with confidence and poise is this: Lucy is adorable and healthy; Haley is strong beyond belief, and recovering well; I am the luckiest and happiest man alive, and the three of us are blessed to have not just each other but a family of uncles, aunts, cousins, grandparents, and friends far and wide.

Whether you enjoy Creation Lake or find it odious depends, I think, entirely on how credulous and interesting you find Sadie Smith, Kushner's unnamed and unmoored protagonist who glides through various left-wing subcultures and treatises with (excepting Bruno Lacombe, a sort of reclusive sage) equal parts indifference and disdain.

If you suspect Sadie to be mostly an authorial insert, whose chief role in the narrative is meant to voice Kushner's own (fatalistic, but with room for majesty and grandeur) commentary on the flaws and foibles of modern leftism, I think it is hard to enjoy this book. This was Brandon Taylor's perspective:

I mentioned to a friend that I was having a hard time with Sadie as a narrator because she seemed stupid and unaware that she was stupid, and my friend suggested that perhaps Kushner had done this on purpose, as a commentary on the ‘sharp woman’ archetype that has predominated in the fiction of the last decade. Perhaps. I replied that I couldn’t decide if the book was a smart person’s idea of a stupid book or a stupid person’s idea of a smart book. But I’ve come to think that the larger problem with Sadie is the difficulty presented by a character who reminds you on every page that nothing matters and nothing is real, and that the people she is scamming are phonies too, that everything is empty and hollow and that she’s smarter than everyone else because she knows the game is a game and is playing to win, but only for mercenary reasons. It brings me back to the question, why did you write this? What are you exploring here?

I think this is an uncharitable and uninteresting way to read the book. I do not think Kushner could be more explicit about Sadie's own shortcomings without fully sacrificing the first-person perspective: Sadie, in her own words, admits that she is an alcoholic, references only missions that she has failed to complete, and routinely fails to fulfill her plans. Indeed, it was hard not to read this book — to descend further into the psyche of Le Moulin, Sadie, and Bruno — and not be relentlessly reminded of Disco Elysium, a terrific and faceted work that ostensibly uses the trappings of genre to explore a very broken character through a leftist lens.

This is not to say that all of the book's flaws can be justified as authorial intent. Few characters besides Sadie and Bruno elevate themselves to more than one-line caricature (we've got the Houellebecq pastiche! we've got the Debord pastiche!); I enjoyed the long, didactic passages of Bruno's missives, but I think it's fair to criticize the novel less as something with concrete narrative propulsion and more as Kushner ambling slowly through a theme park of various topics (in a manner that is somewhat reminiscent of Cusk).

Still, this novel is rewarding. The prose is rich and sharp enough to warrant my renewed investment in The Flamethrowers; the tangents, even if they never quite cohere into something more elevated than "bourgeois cosplaying is silly, but the underlying beauty of leftist struggle remains", are compelling and crisp. I will be thinking about this book for some time to come.

Highlights

Pleasure augers survival.

“Sir, we hoe a row,” he told the police. “We plant potatoes. We don’t use pesticides. We nurture pollinators. But here is how the state does things: They have a deer population that’s getting out of control, so what do they do? They bring in lynx. When farmers get upset about the lynx, the government reintroduces wolves. The wolves kill livestock, so the state makes it legal to shoot them. Hunting accidents increase, so they build a new clinic, whose medical staff creates a housing shortage, necessitating new developments. The expanding population attracts rodents, and so they introduce snakes. And so far, no one knows what to do about the snakes.”

At risk of further succumbing to accusations of self-Flanderization, I'm compelled to draw comparisons between High Output Management and Final Fantasy VII: both works that are in some ways dated (not just by the technology of their time but by the extent to which they influenced the genre they propelled into the zeigeist) and yet stand up well in their own right, not just as part of the canon but as a useful and rewarding work in its own right.

I think many modern references to Grove evoke acidic, dusty feelings of OKRs and the distinct yesteryearishness of Intel; this is an unfair aspersion. Grove is down-to-earth and pragmatic here: much of the challenges he describes have been solved by the past decades of technology (and indeed it is hard not to come away from the book with a certain sense that our industrial organization system is too stymied by best practices that made sense in the 1980s more than they do today), but many things are evergreen: the importance of leverage, the necessity of judging a manager solely their team's performance, the criticality of information flow.

Most books about management do not actually improve the reader's ability as a manager; this is an exception.

It has been a while since I shipped a new thing; it felt good to get back in the saddle, even if the project was so small as to barely warrant the price of the (admittedly fun) domain. I present handbook.directory, a constellation of various once-internal PDFs for company culture. (This was inspired by the brouhaha around Mr. Beast's leaked manual.)

The entire contents is open source, though I'm using Blot to host it.

I was thinking about using this as an opportunity to kick the tires on Astro, but the sheer paucity of content and the fact that I really wanted to get this across the finish line in the course of a one-hour lunch break led me to Blot, which I had never used before but had roughly pigeonholed as "the easiest way to ship very simple static content".

That pigeonholing turned out to be accurate! There was a bit of onboarding jank choosing a theme and getting the DNS records set up (when is there not?) but I was very happy with how Blot performed. Moreover, the thing that really impressed me with Blot was just how fast deploys are. I hit git push blot and by the time I hop over to Chrome and refresh the tab the changes are there.

It is a good reminder of the strides that we've made on the Buttondown front with CI — down roughly from twenty minutes end-to-end to ten — might be impressive in a relative sense but we've got a long way to go in an absolute sense.

There's no real "endgame" with handbook.directory, but it got some solid traction on Twitter and ended up with a dozen GitHub stars and around three thousand pageviews. I'll take it!