🇺🇦 Ingvar Stepanyan
@rreverser.com
2.8K followers 880 following 3K posts
Sr. Principal Engineer at Cloudflare by day, WebAssembly consultant by night. You might also know me from my work on OSS tools and libraries (JS / Wasm / Rust) or Wasm DevRel at Google Chrome 📝 https://rreverser.com/ 📷 https://instagram.com/rreverser
Posts Media Videos Starter Packs
Pinned
> walking through marrakech at night
> some guy comes up and asks if I want hash
> I ask if it's a cryptographic hash
> he doesn't understand
> I show him the NIST definition
> he laughs and says "it's a good hash"
[...]
> it's not a cryptographic hash
Sometimes "referenced by" section of a foundational paper is pure gold.

Tomato quality assessment, star maps, battery defects, ... - you'd think all of these topics can't possibly have anything in common, and yet.
No, only doing astrophotography as a hobby.
Impostor syndrome really be like: "hey, don't you feel bad that some of this professional astronomer's *raw* images look better than the ones you've spent hours postprocessing?"

Why, yes, now I do.
Since it's Node.js, maybe rps? 😜
Reposted by 🇺🇦 Ingvar Stepanyan
One of the swimmers had used ChatGPT to consult the tide times: “I made the mistake of using ChatGPT for research to see when the low tide was."

uk.news.yahoo.com/restaurant-b...
Overwhelmingly positive, right? Right?
Wait till you hear about the books situation.
Typical: you're just trying to use a tool, but then you run into an annoying issue it has and instead of ignoring, you just _have_ to go on a side quest to fix it.

At least this was fairly quick (left: release version, right: with my fix).
Lol looks like this astrophotography catalog has some unexpected superstar images.
You do need to fiddle with settings a bit to get the right degree of similarity avoiding false positives, as different hashes are better for different kinds of images, but, once you got them right for your catalogue, it's blazing fast.
I keep being amazed at just how fast Czkawka is (a duplicate finder written in Rust).

Comparing over 30,000 images by actual image similarity hashes in like... 20 seconds?! github.com/qarmin/czkawka
Huh I guess some apps just crash when you present them with over 32768 images in one go.

Perfectly reasonable usecase, c'mon.
Ah from reading around, apparently they are just raising required level from B1 to B2 🤷‍♂️
I'm confused, from the description in follow up posts it sounds like the same IELTS English test level I had to pass to get work visa 10 years ago. What's new?
Oh, testing my Rust alternative is the very reason I'm collecting those samples.
Ah no it processed the 6th file and updated the estimate to 20 days 😬
Managed to convert them to JPEG, filter out duplicates etc, and now just need to platesolve them all using established tooling as a reference for my tests.

At least that will take only... 4 DAYS?
Turns out, using DevTools + the new File System Access API to download shitload of files from Drive can be a lot easier, more flexible, and faster than official apps and APIs.

Got enough files to get me going now.