marmada

Will Cloudflare eventually open-source Pingora? I know there was a mention of open-sourcing it, but I haven't seen the code, which is a shame :/

Can someone confirm what intensity is high intensity?

E.g, I have a rowing machine that leaves mildly out of breath and makes my heart beat quickly, is that HIIT? Or is HIIT more like stair climbers, which cause me to almost collapse after doing them for 12+ minutes.

I've used Prisma before, and I will note that there pace of feature development seemed mind numbingly slow for a company as large as they were. Entirely possible I'm missing a lot of things, but it was just a big disappointment overall.

The good thing about all of this is that the law doesn't really matter. People will complain about copyright, but the tech is here and the law will follow.

Does anyone really thing this anti-Copilot case has a chance of winning when my guess is that Copilot adoption is exponential over time.

Strongest counter argument is the Supreme Court struck down abortion even though abortion was rising in popularity. So that's a bit worrisome.

I'm curious, what the is the bend of nix for your dev env vs using a docker container with everything pre-installed

What is the reason structured logging is bad? I'm curious, as I felt like it made my life a lot easier.

I wonder if with Next.JS server components, we can just get rid of tRPC + trpc-sveltekit.

I was investigating doing a typescript compiler speedup. I think the time might have finally arrived where we can use ML to generate a faster compiler in a language like Zig.

Probably not for all portions of the compiler. But I think with some reinforcement learning, it ( ... might ...) be possible to automatically generate a robust Typescript parser in Zig.

I have a theory that in another decade ML models will do this better than any optimizing compiler -- similar to a hand written chess engine vs alpha go.

I'm a little more scared, I don't think a glorified Markov chain can -- for example -- explain monads in uwu speech.

It understands things to a degree for sure.

Ask it "why is this wrong" or try some of the "let's think this through step by step" prefixes

All these people whining about copyright need to consider: is the issue Copilot, or is the issue copyright.

Humans are trained on petabytes of data. From birth, we ingest sights, sounds, smells etc. Imagine a movie of every second of your life. And an audio track of every second of your life. Etc. Etc.

Humans get a lot of data.

Wasmtime / WASI does this extremely well.

This doesn't make any sense in this context and I think you know it bud.

I was making the point that stable diffusion != JPEG compression

By that case all art should be copyrighted since our brain stores a highly compressed version of everything we've seen.

I wager 100 TB => 4GB is different from JPEG compression and more similar to what happens in our brains. "Neural compression" so to speak

How does this work with things like WASI--back when I did WASM things I recalled that WASM relies on imported functions for things like Io, etc. so what functions is the Docker runtime providing here?

My hope (for codex, stable diffusion, etc.) is that the models become so popular that it will be impossible to legislate them for issues like copyright. I think there might be a limited window before legal repercussions start happening -- so hopefully the models are in extremely wide spread use by then

One thing I'm curious about is, fundamentally it looks like Zig is going to be the more popular language. I don't know why, but that's just my gut feel. It looks like Zig will get the traction that D never did.

So when I see comments like this that say "D also has X feature", I'm curious about why D didn't take off but Zig (likely) will.

My guess is the true reason is that HN has many people who are scared of threats to their identity.

Take for example all the comments about how AI won't be good enough. My best guess at what is going on is HN people feel threatened by AI (they realize doing standard programming is nothing special), so they are forced to downplay it.

It's like the famous quote: "it is hard to convince a man of something when his job depends on not understanding it", or whatever the quote actually is.

Another example is when people bash (even technical) crypto posts with very generic crypto bad claims. I roughly expect someone who doesn't know anything about, e.g zkSNARKs, to comment generic crypto-bad claims. They don't have enough knowledge to actually engage with the article so they post unrelated criticism.

The last thing is Dunning-Kruger. The Dropbox story is a solid example of Dunning-Kruger at work.

I think Copilot will make cheating easier. A benefit of this is that employers may look at side projects much more, which are harder to cheat

This is the most "just use rsync instead of Dropbox" HN-style argument I've ever heard.

It's sufficiently comical to be satire

This is too funny. I predicted that when I entered this post the first comment would be HN people bemoaning complexity. And indeed it was.

The commenter's proposed solution was WASM. Wasm doesn't solve the issues in this post. What does this post talk about:

- serving content from the edge (wasm doesn't do anything there) - asset optimization (wasm doesn't help)

- pre-rendering complex pages to static HTML+css (wasm doesn't help)

More broadly, minimizing network trips, pre-fetching data, etc. etc. these are all things where Wasm won't help any more or less than JS.

This seems good. Yes, good. No one complains when Tesla gives free software upgrades to car owners. So to be consistent, Tesla should have the ability to also enforce software limitations.

I think it is strongly likely that Copilot will be 1 million times better by the end of the decade. Increasing compute, new techniques, etc

The top comment on this post is complaining about a bug in Copilot generated code. 1 million times better Copilot won't do that.

I do not understand why this leak portrays Uber badly.

Is there anything wrong with having a kill switch? Is there anything wrong with breaking laws that entrench local monopolies.

Laws are not always good. I won't say this is the next civil rights movement, but sometimes extrajudicial measures are needed for reform.

Hey. If you're hear to post about how the web is always changing, and it's so complicated, no one cares. It's been posted 20 times by now and there are good reasons (we build much more complicated webapps these days).

I like Vite. A few things I like:

- Super fast

- Works out of the box (like parcel)

- Even production builds are super fast -- this is a bit of an underrated feature because sometimes things only break in prod builds, so having fast prod builds = more frequent prod tests = better

People use idealism to justify decisions that would only be useful in an ideal world, but are terrifying in the real world. This comment thread is a good case study.

Sure, maybe, in an ideal world we want congress to pass laws in the place of every single regulatory body. (This in and of itself is totally unclear to me).

However, the reality is that by getting rid of regulatory agencies we prevent important limits from being enforced.

I mean, read this part of the dissent: > Again, Section 111(d) tells EPA that when a pollutant—like carbon dioxide—is not regulated through other programs, EPA must undertake a further regulatory effort to control that substance’s emission from existing stationary sources.

There's a backstop in place to allow the EPA to prevent pollutants from fucking us up. Ideally it would be regulated through a formal program. It's not. Does that mean we should just get rid of all backstops & regress to the stone age?

The 0.1% figure comes from OpenAI: https://twitter.com/eevee/status/1410037309848752128 (person disagrees w/ me, the image is what's relevant).

I talk about Lodash one-liners because it makes it pretty obvious that Copilot is not just copy-pasting code (which would be copyright infringement). It's quite unlikely (even if we consider all variables to have the same name), that Copilot is copy-pasting an exact copy of some other snippet, given that the snippet is very specific to my problems. (I'm not asking it to write a generic math function, I'm asking it to use lodash on data structures in my codebase to accomplish a very specific outcome). By quite unlikely I mean (1 / (2 ^ 32)), if we consider the one-liner to be composed of 32 different AST nodes.

Is a Markov chain "Creative"? That's essentially what you're describing here.

Have you read the paper that (eventually) inspired Copilot? "Attention is All You Need". It's not a Markov chain. There's many, many, many different steps / layers. I think if you know & understand the fundamental building block that is responsible for it working (the "transformer"), then a lot of the worries around plagiarism go away.

Also. What is coding if not a search problem over a very large space? I'm searching for the next N lines to write over the space of all possible lines. To guide my search I use things like my prior knowledge.

In my head, this knowledge is encoded using neurons. In Copilot, the knowledge is encoded in parameters. Abstract concepts in my head are encoded in neurons. In Copilot's "head", abstract concepts are encoded using "embedding vectors" of 512 bytes (maybe more / less, not sure). Yeah, maybe I use more than 512 bytes to encode a concept, but still, I don't see a huge difference. If I'm not plagiarizing, then I can't imagine Copilot is (except for in the 0.1% of cases where it's over-fitting)

Is io_uring the reason this is faster? I'm curious because redis is in memory right? And io_uring is mostly for disk ops, I assume?