frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Do transformers need three projections? Systematic study of QKV variants

https://arxiv.org/abs/2606.04032
51•Anon84•58m ago•5 comments

Anthropic's open-source framework for AI-powered vulnerability discovery

https://github.com/anthropics/defending-code-reference-harness
224•binyu•3h ago•74 comments

VoidZero Is Joining Cloudflare

https://blog.cloudflare.com/voidzero-joins-cloudflare/
550•coloneltcb•11h ago•246 comments

IPv6 zones in URLs are a mistake

https://xeiaso.net/notes/2026/ipv6-zones-go-url/
60•xena•2h ago•48 comments

When AI Builds Itself: Our progress toward recursive self-improvement

https://www.anthropic.com/institute/recursive-self-improvement
292•meetpateltech•7h ago•378 comments

Branchless Quicksort faster than std:sort and pdqsort with C and C++ API

https://tiki.li/blog/blqsort
57•birdculture•2d ago•7 comments

Queen bees emerge from special wax chambers

https://cen.acs.org/materials/biobased-materials/queen-bees-special-wax/104/web/2026/06
28•gmays•2h ago•0 comments

Ian's Secure Shoelace Knot

https://www.fieggen.com/shoelace/secureknot.htm
483•mooreds•12h ago•184 comments

Retro-Tech Parenting

https://havenweb.org/2026/05/28/retro-tech.html
230•mawise•8h ago•159 comments

KVarN: Native vLLM backend for KV-cache quantization by Huawei

https://github.com/huawei-csl/KVarN
111•theanonymousone•8h ago•11 comments

Show HN: FFmpeg WebCLI – Full FFmpeg in Browser, Offline PWA, No Uploads(WASM)

https://github.com/tejaswigowda/ffmpeg-webCLI
63•tejaswigowda•3h ago•22 comments

Castor: CERN Advanced STORage Manager

https://castor.web.cern.ch/content/home.html
40•naves•4h ago•15 comments

Latent Agents: A Post-Training Procedure for Internalized Multi-Agent Debate

https://arxiv.org/abs/2604.24881
5•PaulHoule•1h ago•0 comments

Samurai City

https://worksinprogress.co/issue/samurai-city/
87•zdw•2d ago•13 comments

External Clock Generation on RTX 50 Series

https://www.xtremesystems.us/post/external-clock-generation-on-rtx-50-series
5•mfro•1d ago•0 comments

JLink JTAG Access on the Pinecil

https://danielmangum.com/posts/jlink-jtag-pinecil/
34•hasheddan•2d ago•5 comments

Making Debian or Fedora persistent live images

https://sigwait.org/~alex/blog/2026/05/28/smdBC8.html
56•henry_flower•3d ago•6 comments

Show HN: Hitoku Draft – Context aware local assistant

https://hitoku.me/draft/
9•lostathome•6h ago•1 comments

Meta's ships facial recognition on smart glasses

https://www.buchodi.com/meta-glasses-facial-recognition/
203•buchodi•4h ago•187 comments

Zettascale (YC S24) Is Hiring Founding FPGA Engineers

https://www.ycombinator.com/companies/zettascale/jobs/O9S1vqO-founding-engineer-fpga-rtl-asic-arc...
1•el_al•7h ago

Mornings and nights no longer exist at 47C: A day in the hottest place in India

https://www.bbc.co.uk/news/articles/crmp0krp98ro
97•mellosouls•2d ago•63 comments

Show HN: Mercek – A Desktop IDE for AWS ECS

https://www.mercek.dev/
14•utibeumanah•2h ago•2 comments

Sum-product, unit distances, and number fields

https://www.erdosproblems.com/forum/thread/blog:6
52•robinhouston•3d ago•12 comments

Show HN: Uruky (EU-based Kagi alternative) now has Image Search and URL Rewrites

https://uruky.com/?il=en
204•BrunoBernardino•15h ago•193 comments

WSL 2 is getting faster Windows file system access

https://www.boxofcables.dev/wsl2-per-device-swiotlb-pools-for-virtiofs-and-virtioproxy/
22•haydenbarnes•4h ago•8 comments

Gaussian Point Splatting

https://momentsingraphics.de/Siggraph2026.html
173•ibobev•13h ago•65 comments

AI, Ashby Engineering, and the future

https://www.ashbyhq.com/blog/engineering/ai-ashby-engineering-and-the-future
49•fredley•9h ago•31 comments

U.S. Army Corps of Engineers Bay Model

https://en.wikipedia.org/wiki/U.S._Army_Corps_of_Engineers_Bay_Model
206•tosh•2d ago•53 comments

White House will dump $700M of public funds into costly, unreliable coal again

https://electrek.co/2026/06/04/white-house-will-dump-700m-of-public-funds-into-costly-unreliable-...
11•Bender•1h ago•5 comments

3D-printed book turns its own G-code into raised lettering

https://www.designboom.com/design/3d-printed-book-manual-darius-ou-benson-chong/
72•surprisetalk•2d ago•27 comments
Open in hackernews

A Rust API Inspired by Python, Powered by Serde

https://ohadravid.github.io/posts/2025-05-serde-reflect/
58•lukastyrychtr•1y ago

Comments

ohr•1y ago
(Author here) I needed to do a bit of "reflection" in a Rust crate but didn't want to implement a procedural macro, so I used Serde (which is a (de)serialization crate) instead.

This is also a deep dive into Serde internals - hope you'll like it!

snthpy•1y ago
Yes, great post! Thank you.
dundarious•1y ago
Pardon me, but I prefer the original by 1 million miles.

  let res = raw_api::query("SELECT * FROM Win32_Fan");
  for obj in res {
    if obj.get_attr("ActiveCooling") == Value::Bool(true) {
        if let Value::String(name) = obj.get_attr("Name") {
            if let Value::UI8(speed) = obj.get_attr("DesiredSpeed") {
                println!("Fan `{name}` is running at {speed} RPM");
            }
        }
    }
  }
If actually concerned about the need to know UI8, then create a typedef DesiredSpeedT or similar. This is equivalent to the struct Fan.

Edit: I understand the post is probably more of a playful exercise than anything else, but I really think the original is far far better (smaller, simpler, etc.) and hope that is not lost on people.

ohr•1y ago
That's understandable, but I think it depends on how many different structs like this you have and how many fields you need to work with (for our usecase, we had tens of structs with tens of fields each).

There's also an Alternatives section in the article about other approaches that can achieve similar results, but of course 'do nothing' is also a valid option.

Edit: > If actually concerned about the need to know UI8 ..

Just a small note: even if you don't care about the fact that it's a UI8, you still have to use the correct type. For example, if the field happens to be returned as UI4, this code won't work!

dundarious•1y ago
Right, but isn't the struct definition equivalent in line count and effort compared to some typedefs and perhaps a handful of trivial-to-inspect oneline helper functions?

Regarding the UI8, don't you have to get your version's struct data member type correct to the exact same degree as a typedef in my suggestion?

ohr•1y ago
> don't you have to get your version's struct data member type correct

No, since Serde will happyly fill a `u64` field with any other `u{8,16,32}` value, and even with signed types (as long as the actual value is non-negative) - this is sort of what happens when you deserialize a JSON `[1, 2, 3]` into `[u64]`.

dundarious•1y ago
Yes, but an equivalent to `impl<'de> Deserializer<'de> for ValueDeserializer` handles that. That could be a useful helper.
LtWorf•1y ago
dgacmu•1y ago
This kind of sells the reason not to wrap things behind an object interface, doesn't it?

    for fan in c.query("SELECT * FROM Win32_Fan"):
        if fan.wmi_property("ActiveCooling").value is True:
            print(f"Fan `{fan.wmi_property('Name').value}` is running at {fan.wmi_property('DesiredSpeed').value} RPM")
vs "SELECT Name, DesiredSpeed from Win32_Fan where ActiveCooling"

Obviously, this doesn't matter when you have 5 fans, but in general, you want to push your restrictions as deeply into the query as possible from an optimization standpoint.

ohr•1y ago
In WMI, the fields are lazy loaded when you do a `*` query, but the real crate [does use the same Serde reflection tricks](https://github.com/ohadravid/wmi-rs/blob/main/src/query.rs#L...) to create the correct field list when you query a struct which improves perf a lot!
vlovich123•1y ago
> Obviously, this doesn't matter when you have 5 fans, but in general, you want to push your restrictions as deeply into the query as possible from an optimization standpoint.

Depends where the database lives. If it's an in-process SQLite DB instance, there's no difference & doing this in code is easier to understand than more complicated SQL queries (of course not necessarily in this case but in general). But in all other cases you are correct about efficiency in general (although again other effects can dominate & make it irrelevant).

lnyng•1y ago
Interesting post. We wrote this “below” utility [1] that monitor system metrics similar to atop. We want the ability to collect all metrics into a single object, pass it around and visualize it elsewhere. Naturally we need some way to query into fields or even nested-struct fields. For example, to get the file cache usage of a particular process, we need to go through sample->processes->pid->memory->file cache. To do it ergonomically and also type-safely, we end up using proc macro to generate enums that represent field paths of the structs and then use them to query values of non-struct (leaf) fields. I always wonder if there are simpler ways or existing proc macro derives to safe us the efforts. Maybe I do need to look into serde internals for some inspirations.

[1] https://github.com/facebookincubator/below/blob/main/below/b...

lovasoa•1y ago
In my opinion, the clean way to implement this is with methods instead of attributes for name, desired_speed, etc...
xpe•1y ago
The title is vague in my opinion. What kind of API? What problem does it hope to solve? The article uses querying system data as examples, but after skimming it, I’m not sure why I would care. My comment is also a criticism of the article, since I couldn’t skim in quickly to figure out if I should spend more time on it.
ohr•1y ago
(Author here) Thanks! That's useful feedback.

I also agree - the final article isn't skim-friendly enough, which drives away some readers.

xpe•1y ago
Glad you are open to feedback. My top question is: What kind of people do you want to read this and why?
vlovich123•1y ago
I don't really understand what this offers above diesel.rs which AFAIK is a similar reflection interface except with much more flushed out ORM capabilities (much more complex filtering, joining etc) & support for an assortment of SQL dialects.
VWWHFSfQ•1y ago
> let res: Vec<Fan> = query();

It might feel more natural, and less magical if this used a turbofish instead

    let res = query::<Fan>();

Very neat
throw_a_grenade•1y ago
That's wdat minijinja does internally. As an argument to Template::render() you can give it any struct that implements Deserialize. That's how you can get varying (sic) variables to a single function and even attributes on "objects" processed by tte template.
I wrote typedload in python. Once they show you an API with hundreds of types you appreciate not having to do like that all the time.
dundarious•1y ago
I don't see the issue with just using an equivalent to `impl<'de> Deserializer<'de> for ValueDeserializer` then.
LtWorf•1y ago
There's unions, there's stuff that uses reserved words in the language as field names... You are obviously not familiar with this task.
dundarious•1y ago
The "generic macro over struct definitions" approach is the one that has at least some trouble dealing with these situations, not the "getter function with string argument" approach. I've had to do plenty of json and protocol buffers wrangling (meaningfully different, I know), and versioned packed struct memcpy/reinterpret_cast "parsing" as well.

My point is I don't want to do "this task" at all if "this task" means layering funky auto-generated machinery on top of what is already a perfectly reasonable API.

olalonde•1y ago
Why? It's much more verbose and error prone (e.g. "stringly typed"). Do you never deserialize JSON?
dundarious•1y ago
What's the difference between mistyping in the string here and mistyping in the struct definition? And yes I have.
olalonde•1y ago
You only need to get it right once, and from then on the compiler will catch any mistakes if you use it incorrectly. In contrast, every time you write obj.get_attr("DesiredSpeed"), there's a chance you'll make a typo and the compiler won't warn you about it.