fwlr

15 ms·

211.
by fwlr 2y ago
My favourite Gwern insight is “Bitcoin is Worse is Better”, where they summarize an extensive list of objections to Bitcoin and then ask if there’s a common thread: No! What’s wrong with Bitcoin is that it’s ugly. … It’s ugly to make
212.
by fwlr 2y ago
The article mentions both rockets and volcanoes as causes of the increase, but carefully elides any mention of the exact numbers each contributes. I couldn’t easily find any concrete numbers on how much water vapor is involved in a typical
213.
by fwlr 2y ago
I’m so used to fuzzy search at this point that it didn’t even occur to me, but yeah, you’re right. I guess tooling can still provide a solution - a snippet on “import”, where the first tab stop takes you to the module path string, and the n
214.
by fwlr 2y ago
In fairness - given a choice and some time to think, I would probably not choose to actively break commonjs. However, if I was tasked with increasing ESM adoption, I would go about it like this.
215.
by fwlr 2y ago
No no no, it’s 2024, we write everything in Rust now
216.
by fwlr 2y ago
I think that particular issue could be solved with tooling - if you can autocomplete from “package name” to a given exported identifier, you must already have a mapping from “package name” to “exported identifier” somewhere in your machine’
217.
by fwlr 2y ago
Bun put a lot of work into making both “import” and “require” always work regardless of whether it’s given a commonjs or an ESM target. I’d say that’s half of the right idea: make only import work with anything. Another angle that might b
218.
by fwlr 2y ago
Yes, that sounds a lot like the argument I remember hearing
219.
by fwlr 2y ago
Sure! In a professional capacity - our customers were dissatisfied with the length of time and flakiness of a particular workload job[0], and I was party to the discussions on how to solve this problem. Knowing Little's Law allowed me
220.
by fwlr 2y ago
In mathematical queueing theory, Little's law (also result, theorem, lemma, or formula[1][2]) is a theorem by John Little which states that the long-term average number L of customers in a stationary system is equal to the long-
221.
by fwlr 2y ago
Note that after several paragraphs, the author discloses that they were the Chairman of the US Nuclear Regulatory Commission from 2012-2014.
222.
by fwlr 2y ago
I think the A and the N nix this theory. It’s almost more like the Quake font designer saw this font, remembered the concept of hairlines reduced to nothing, but when they applied the concept they made different choices on which hairlin
223.
by fwlr 2y ago
From a brief skim of the whitepaper, it seems like the general idea is something like “GPUs are <big-multiple> faster than CPUs, because GPUs are built ground-up for parallel computation. The bottleneck now is parallel memory access,
224.
by fwlr 2y ago
The algorithm involved is actually very hefty: for each cell of a 9px by 15px grid over the image, compare each pixel of the cell to its equivalent pixel in each of the 95 ascii characters. To solve for optimal grid alignment, it repeats th
225.
by fwlr 2y ago
“Take a screenshot every few seconds, run OCR on it, and dump all the text into a SQLite database” honestly sounds like a machine learning intern wrote a keylogger in their coffee break. Hilarious that it’s a real Microsoft feature they pro
226.
by fwlr 2y ago
My gut instinct from reading about the phenomenon says that a “grokked” model of X parameters on Y tokens is not going to outperform an “ungrokked” model with 2X parameters on 2Y tokens - since “grokking” uses the same resources as paramete
227.
by fwlr 2y ago
Later on the author makes the point that everything changes some various register deep in the guts of the CPU so by definition everything is always effect-full, but since we invented programming languages to get away from writing assembly,
228.
by fwlr 2y ago
I’m worried that opponents of facial recognition are shooting themselves in the foot with the inaccuracy argument. Sure, maybe face rec ends up like Siri or self-driving cars. Even if it ends up like DallE or ChatGPT they probably still hav
229.
by fwlr 2y ago
> Should shoplifting once mean that a person is banned from all stores that use this service for life? I mean, it depends, right? Barred from buying any groceries for stealing a candy bar is obviously wrong… but if you walk out the door
230.
by fwlr 2y ago
Unsure on correct terminology but the comments in the implementation seem to imply that there are multiple CPUs per core (or perhaps multiple cores per CPU?), and the overall goal seems to be to ideally use just one of the members of each g
231.
by fwlr 2y ago
I partially agree - I would say more specifically “those situations are the easiest to write good tests for”, ie having a predefined spec will strongly guide you towards writing good and useful tests. “Testing that the code does what it doe
232.
by fwlr 2y ago
Tests that “calcify the designer’s whims” - great way to put it - can be quite useful if your job description happens to be “carrying out the whims of the designer” (and for many of us, it is!) With high coverage and dry-ish tests, changing
233.
by fwlr 2y ago
Community notes get displayed right underneath the tweet they’re correcting. Probably the most unique thing about them (relative to most other forms of correction/warning/note/reply/discussion on the internet) is they ha
234.
by fwlr 2y ago
Arguments against testing tend to fall prey to the von Neumann Objection: they insist there is something tests can’t catch, and then they tell you precisely what it is that tests can’t catch… so you can always imagine writing tests for that
235.
by fwlr 2y ago
“Y is being laughed at by X” is an odd linguistic construction to see in a headline - why not the clearer and shorter “X laughs at Y”? And yet by far the most efficient way to explain Perplexity is “a rival to Google search”, so writing it
236.
by fwlr 2y ago
There are some bugs/exploits for chatbots that are like “if I repeatedly ask for ‘extra rare’ when ordering my burger, the restaurant will serve me a raw ground beef patty”. These are dangerous. There are other bugs/exploits that
237.
by fwlr 2y ago
It's governments vs corporations; the 500 million people aren't actually being represented in either camp. Being a subject is little better than being a consumer, and in some ways much worse.
238.
by fwlr 2y ago
I was quite surprised by this, I was certain it wasn’t true, but I checked and it is!
239.
by fwlr 2y ago
No financial incentive or relationships to disclose, just a satisfied user: I found that SuperMaven was a better “coding copilot”. If you happen to use VSCode I’d check that one out this afternoon.
240.
by fwlr 2y ago
Mismatched brackets seems like a lower-level job - better done by a linter, not an LLM. Likewise, I wouldn’t rely on an LLM to tell me what properties are on the response object - that job is better done by e.g. TypeScript. Analogously, “co

More ›