is justinjay.wang/methods-for-... the right one?
Methods for random gradients
An overview of techniques I’ve used to generate random gradient images.
justinjay.wang
August 27, 2025 at 11:14 PM
is justinjay.wang/methods-for-... the right one?
while I can't speak for everyone's experience, I don't feel threatened at all by these tools. there's a perpetual shortage of good talent and never enough time to get everything done.
if execs want to replace their employees with LLMs I only ask that they stay long enough to see how that plays out.
if execs want to replace their employees with LLMs I only ask that they stay long enough to see how that plays out.
August 7, 2025 at 2:39 PM
while I can't speak for everyone's experience, I don't feel threatened at all by these tools. there's a perpetual shortage of good talent and never enough time to get everything done.
if execs want to replace their employees with LLMs I only ask that they stay long enough to see how that plays out.
if execs want to replace their employees with LLMs I only ask that they stay long enough to see how that plays out.
note that this graph is usage of OpenAI's models via a broker called OpenRouter, so this skews commercial and doesn't include ChatGPT usage.
The drop was likely because a big customer moved to paying OpenAI directly, so this really represents an on-paper thing rather than any real change in usage.
The drop was likely because a big customer moved to paying OpenAI directly, so this really represents an on-paper thing rather than any real change in usage.
August 7, 2025 at 4:55 AM
note that this graph is usage of OpenAI's models via a broker called OpenRouter, so this skews commercial and doesn't include ChatGPT usage.
The drop was likely because a big customer moved to paying OpenAI directly, so this really represents an on-paper thing rather than any real change in usage.
The drop was likely because a big customer moved to paying OpenAI directly, so this really represents an on-paper thing rather than any real change in usage.
(i ended up using wasm-ld's --no-merge-data-segments with a classic `wasm2wat | grep memory.init` then memcpying the sections where i wanted them at startup, clearly a much less cursed option)
July 24, 2025 at 2:11 PM
(i ended up using wasm-ld's --no-merge-data-segments with a classic `wasm2wat | grep memory.init` then memcpying the sections where i wanted them at startup, clearly a much less cursed option)
ok i'm slightly hyperbolizing but i was once so annoyed patching the linker script dependency out of some software that i genuinely contemplated writing my own wasm linker with linker script support.
July 24, 2025 at 2:08 PM
ok i'm slightly hyperbolizing but i was once so annoyed patching the linker script dependency out of some software that i genuinely contemplated writing my own wasm linker with linker script support.
my life would be significantly better if wasm-ld supported linker scripts
July 24, 2025 at 3:12 AM
my life would be significantly better if wasm-ld supported linker scripts
GPT/3.5 (compatible) Gemini-Pro/1.5 (PaLM-2; Bard) Claude/3-sonnet LLaMA/3.2-3B-int4 (like LLaMA/3-8B) Gemma-7B (SearchGrounded/2.0; like Flash)
July 1, 2025 at 10:26 AM
GPT/3.5 (compatible) Gemini-Pro/1.5 (PaLM-2; Bard) Claude/3-sonnet LLaMA/3.2-3B-int4 (like LLaMA/3-8B) Gemma-7B (SearchGrounded/2.0; like Flash)
I'm pretty sure this was only said in reference to React's class components, not classes as a language feature.
Looking at the small amount of the new Remix that's already published (github.com/mjackson/rem...), they seem perfectly happy to use classes where they're appropriate.
Looking at the small amount of the new Remix that's already published (github.com/mjackson/rem...), they seem perfectly happy to use classes where they're appropriate.
May 31, 2025 at 6:50 AM
I'm pretty sure this was only said in reference to React's class components, not classes as a language feature.
Looking at the small amount of the new Remix that's already published (github.com/mjackson/rem...), they seem perfectly happy to use classes where they're appropriate.
Looking at the small amount of the new Remix that's already published (github.com/mjackson/rem...), they seem perfectly happy to use classes where they're appropriate.
the design isn't close to finalized, but they've actually been doing the async and rpc stuff on public branches if you know where to look: github.com/sveltejs/kit...
kit/playgrounds/basic/src/routes/+page.svelte at rpc · sveltejs/kit
web development, streamlined. Contribute to sveltejs/kit development by creating an account on GitHub.
github.com
April 29, 2025 at 8:47 AM
the design isn't close to finalized, but they've actually been doing the async and rpc stuff on public branches if you know where to look: github.com/sveltejs/kit...
the dilemma of jj:
you can use it and have nobody on your team even tell the difference, and yet you will feel compelled to evangelize it to anyone that will listen.
you can use it and have nobody on your team even tell the difference, and yet you will feel compelled to evangelize it to anyone that will listen.
March 15, 2025 at 4:34 AM
the dilemma of jj:
you can use it and have nobody on your team even tell the difference, and yet you will feel compelled to evangelize it to anyone that will listen.
you can use it and have nobody on your team even tell the difference, and yet you will feel compelled to evangelize it to anyone that will listen.
I'm really hoping that with memory64 shipping now, more of a focus is put on memory control/virtual memory. I think it's the last wasm feature needed for true native parity.
January 8, 2025 at 9:59 AM
I'm really hoping that with memory64 shipping now, more of a focus is put on memory control/virtual memory. I think it's the last wasm feature needed for true native parity.
ok but what's more impactful: breadth of shaming, or depth of shaming?
January 4, 2025 at 8:07 AM
ok but what's more impactful: breadth of shaming, or depth of shaming?
arch is immutable if you mount / as readonly
December 30, 2024 at 12:45 PM
arch is immutable if you mount / as readonly
you might've already seen this, but browsers have a pretty well defined (but complicated) algorithm for when they decide to show the encoded or decoded form:
chromium.googlesource.com/chromium/src...
chromium.googlesource.com/chromium/src...
chromium.googlesource.com
November 9, 2024 at 5:08 AM
you might've already seen this, but browsers have a pretty well defined (but complicated) algorithm for when they decide to show the encoded or decoded form:
chromium.googlesource.com/chromium/src...
chromium.googlesource.com/chromium/src...
like maybe I'm overindexing on that lightbulb moment, where you realize the elegance of the handful of base primitives jj gives you, and how you can really build any composite workflow on top of them, but first impressions are important!
October 29, 2024 at 2:45 PM
like maybe I'm overindexing on that lightbulb moment, where you realize the elegance of the handful of base primitives jj gives you, and how you can really build any composite workflow on top of them, but first impressions are important!
I might be coming at this with my perspective of really enjoying your first tutorial, rather than as a newcomer who only knows git, but I feel like starting by demonstrating jj's unique aspects might be more valuable than presenting it as a different interface to the same git workflows.
October 29, 2024 at 2:45 PM
I might be coming at this with my perspective of really enjoying your first tutorial, rather than as a newcomer who only knows git, but I feel like starting by demonstrating jj's unique aspects might be more valuable than presenting it as a different interface to the same git workflows.
Any reason you're introing `jj commit` rather than `jj desc && jj new`?
It might be the right choice because it maps closer to git, but OTOH I think it would be valuable to open with "we have a familiar but different mental model"
It might be the right choice because it maps closer to git, but OTOH I think it would be valuable to open with "we have a familiar but different mental model"
October 29, 2024 at 2:02 PM
Any reason you're introing `jj commit` rather than `jj desc && jj new`?
It might be the right choice because it maps closer to git, but OTOH I think it would be valuable to open with "we have a familiar but different mental model"
It might be the right choice because it maps closer to git, but OTOH I think it would be valuable to open with "we have a familiar but different mental model"
Then, to map from icon name to svg file, take the hash of the name mod the number of buckets you want.
October 24, 2024 at 2:16 PM
Then, to map from icon name to svg file, take the hash of the name mod the number of buckets you want.
Assuming we're talking svg icons, they'll compress extremely well when chunked together.
I would do a pretty small number of spritemaps, then you're only dealing with a few svg files so you can just inline the map of hashes.
I would do a pretty small number of spritemaps, then you're only dealing with a few svg files so you can just inline the map of hashes.
October 24, 2024 at 2:16 PM
Assuming we're talking svg icons, they'll compress extremely well when chunked together.
I would do a pretty small number of spritemaps, then you're only dealing with a few svg files so you can just inline the map of hashes.
I would do a pretty small number of spritemaps, then you're only dealing with a few svg files so you can just inline the map of hashes.
I feel like this is a tough find. A library that does more than just string templating, but hasn't had enough scope creep that it wants to own the entire client architecture.
The closest thing I can think of is lit-html (and clones), but that really depends on your definition of templating library.
The closest thing I can think of is lit-html (and clones), but that really depends on your definition of templating library.
October 22, 2024 at 2:09 PM
I feel like this is a tough find. A library that does more than just string templating, but hasn't had enough scope creep that it wants to own the entire client architecture.
The closest thing I can think of is lit-html (and clones), but that really depends on your definition of templating library.
The closest thing I can think of is lit-html (and clones), but that really depends on your definition of templating library.
Hmm the question is do you ever actually want an isomorphic templating library?
Surely you'd want a string templating library on the server, but a dom templating library on the client.
(or is that what you're looking for?)
Surely you'd want a string templating library on the server, but a dom templating library on the client.
(or is that what you're looking for?)
October 22, 2024 at 12:51 PM
Hmm the question is do you ever actually want an isomorphic templating library?
Surely you'd want a string templating library on the server, but a dom templating library on the client.
(or is that what you're looking for?)
Surely you'd want a string templating library on the server, but a dom templating library on the client.
(or is that what you're looking for?)