joshmoody24

Or maybe ants and humans are qualitatively different. Maybe there's a critical mass of intelligence where you can pretty much understand anything, and maybe humans are past that threshold. I don't know that for sure, but I don't think we're anywhere close to hitting fundamental limits to our ability to understand the universe.

An LLM could also in theory fake alignment and create technology that seems really useful but secretly harms humanity in the long run. If humans don't understand what it does, that possibility becomes more likely.

You're responding to a comment about weather forecasts, not current weather. Pointing to the sun is not a forecast.

And clicking outside the modal

My team uses a Claude Code hook that blocks any comment more than 2 lines long, and when tripped it encourages the agent to rewrite the comment more concisely and focus only on the "why" not the "what" of the code. I've found this extremely useful for code reviews.

> Want it to go away, almost like magic?

True, almost like magic and magic are not the same thing, but I'd hesitate to call this a strawman.

Would it be useful if the models were actually good at it?

I second the Victorian Internet recommendation!

The book makes a good argument that if you plucked someone from the late 1800s and brought them into the modern day, the internet might be the invention they are least impressed by, since they basically already had one.

The similarities between the telegraph network and the internet go way deeper than I expected. Both at the technology level, like domain names, encryption, data compression, routing tables, etc., and also at the social level like hackers, online dating, insane levels of hype like predicting world peace, and government regulation unable to keep up with the technology. Sometimes I forget how insanely smart people have always been. Awesome book. More timeless than the average book about technology

Obviously it's impossible to understand the whole stack. My rule of thumb is to understand at least two layers of abstraction above and below the thing I'm actively working on.

The old parser has p95 of like 450ms, that seems weirdly slow to me even for a parser generator. Is my intuition wrong? Maybe it's parsing some truly enormous SQL queries?

I've been on a crusade against sloppy use of AI recently. A conversation between my brother and I spawned the idea of "emoji reaction dog whistles" and I couldn't stop thinking about the idea until I wrote it down. Not meant to be taken very seriously, but not entirely UNseriously either

I've seen the pendulum swing between duplication and abstraction a few times in my career, and I'm currently on team "it's usually not that hard to find a good abstraction up front."

IMO it's easier to inline a bad abstraction than it is to consolidate a bunch of subtly different things that should have been abstracted from the beginning.

But I expect people's opinions on this differ wildly based on their personal experiences. Just my anecdotal take.

As someone who has listened to this If Books Could Kill episode and read the actual book, I think the podcast is mostly looking for problems where there aren't any. Despite the book's sleazy title ("influence people") it's mostly just basic, innocent social advice.

So you don't care about things that indirectly affect your net worth? Credit score? Your overall health? How many friends you have?

In my experience, lack of style customizability is the biggest obstacle to using native HTML features

We don't need to give up streetlights to make huge progress. 20 to 50 percent of outdoor lighting is wasted. Just shielding streetlights would go a long way.

https://www.nps.gov/subjects/nightskies/sources.htm

What do you mean by strictly more powerful?

Antkeeping, rockhounding, and flint knapping are things I've gotten into recently. Antkeeping and rockhounding in particular have a ton of room for citizen science. Lots of undiscovered stuff out there.

It's not like the Shopify CEO intentionally tried to make something low quality. Was he out of his depth? Sure, but I don't fault him for trying. I think the people hyping up this event have caused more harm than the event itself.

Lambda calculus is a model of computation. A pretty lightweight model that can still do everything that other programming languages can. Combinators are an even simpler model of computation that is still equally powerful. That simplicity / power ratio is what makes them cool.

A while back I built all the way up to FizzBuzz from just S and K combinators. Took days of doing all the math by hand, lol.

Here's my write up of doing that. I did it in JavaScript because most combinator articles online were prohibitively academic for my layman mind. https://joshmoody.org/blog/programming-with-less-than-nothin...

What do you mean by "too heavy"? File size?

This is my experience as well. MJML is the older, more reliable, better documented technology. And when it comes to debugging email rendering, you really, REALLY want as much documentation as possible.

I used to think this, but lately I'm getting a lot of plain text marketing emails that are clearly LLMs. Now I dislike plain text emails just as much as HTML ones.

Writing HTML for emails is a lot harder than you're making it sound. But MJML does a good job of simplifying it for most use cases.

This surprises me too. I've experimented with using LLMs to convert lambda calculus expressions into combinatory logic. There is a simple deterministic way to do this, and LLMs claim to know it, and then they confidently fail.

I made a queuing theory calculator for this: https://joshmoody.org/blog/number-of-agents/#number-of-agent...

Although you'll have to mentally replace the word "agent" with "PR" for it to make sense in this context. The math is the same. It all boils down to how much those context switches costs you. If it's a large cost, then you can get a huge productivity boost by increasing review speed.

In the "show calculations" section, the amount of wasted time caused by context switching is the delta between the numbers in the phrase "T_r adjusted from 30.0 to 35 minutes". That number is increases as context switching cost and "average agent time" (AKA "average PR review time") goes up.

In my experience the places worth working for care about cornerstone principles far more than the hot buzzwords on a resume.

Half the tests aren't passing and the code is a lot less readable, but the core idea is cool

I've been using NixOS for the last year and have loved it. Rebuilding doesn't fail for me nearly as often as it does for the author. Running nix garbage collection frequently isn't a big deal and can easily be automated. The network usage is a fair point. But IMO a small price to pay for keeping all of my devices perfectly in sync and running into weird "works on my machine" issues way less frequently.

My ring finger is always slightly sore due to typing ( so often. Not just in LISP but any programming language. I've been experimenting with mapping l; to () when the caps lock key is pressed. So far so good, but I haven't used it long enough to develop muscle memory.