I had to take a brief detour to triage a surprising new class of SemVer breakage. It's always *a time* when the net result is "8 issues across 5 repos" 😅 The net result is the same: even more lints! github.com/obi1kenobi/c...

Great talk by @anarazel.de on profiling postgres youtu.be/8Y6GfTY5eFE

After spending all of the time with DeepSeek and GLM, I start debugging something with Claude and get this.

The time crate has unintentionally released breaking changes in two recent releases. While this may have been avoided, please let's stop piling on with GitHub comments like "it broke for me too", "it should have been a 0.4". Learn how to fucking use `cargo update` and shut up!

Do you use #rustlang hyper/h2 at scale? Would you be able to easily try out a performance run using a PR that is meant to significantly reduce contention with multi-threaded HTTP/2? It would be a big help! github.com/hyperium/h2/...

I'm having so much fun using DeepSeek V4 Flash on OpenRouter, it's highly disappointing that Scaleway closed the feature request for DeepSeek V4 Flash/Pro the other day in favor of GLM 5.2. Yes GLM 5.2 is better, but it's also 10x the cost on input and 20x the cost on output (on OpenRouter)

GLM 5.2, Q2_K routed experts (effectively ~2.6 bits) running with SSD streaming on an M5 Max 128GB computer.

🦀 first basic prototype for the diff viewer on crates.io seems to work as planned 😍 and the regular code viewer is now publicly released :) #rustlang

They added GLM 5.2 but it costs more than the average price on OpenRouter, and the lack of caching makes it even more expensive. Let's see if we can get DeepSeek V4 Flash added. Please vote. feature-request.scaleway.com/posts/1383/d...

Scaleway just added GLM 5.2

I decided to finally start playing around with native GUI programming via the eframe crate, with assistance from LLMs. Impressed by the results for now.

I've been playing with DeepSeek V4 in the past few days. If you too want a European provider host this amazing model, could you 🔺 this request? I wish I didn't need to buy more expensive hardware or have to deal with non-EU providers. feature-request.scaleway.com/posts/1350/d...

In 2020, OpenSSL had a vuln related to the signature_algorithms_cert ext. openssl-library.org/news/secadv/... Palo Alto apparently "solved" this in their IPS by blocking connections with "unknown" algs in sigalg_cert. Six years later, we can't add ML-DSA to sigalg_cert in Go. sigalg_cert is dead.

Yesterday I released rusty-s3 v0.10 with the ability to use aws-lc or graviola instead of RustCrypto crates (except for md5). This release also replaces quick-xml with instant-xml. github.com/paolobarboli...

I released Graviola 0.4.0 yesterday, with a new and quite quick ML-KEM-768 implementation. More perf work to do on this, but it is already quicker than libcrux-ml-kem, and competitive with mlkem-native on ARM64. And still builds mega fast! Consider supporting this work on github.com/sponsors/ctz

This new graviola release is amazing! aws-lc-rs v1.17.0: 18 dependencies - 16s release build rustls-graviola v0.3.4: 45 dependencies - 9s release build rustls-graviola v0.4.0: 19 dependencies - 5.5s release build Tested on a 24 vCPU 5950X VM. github.com/ctz/graviola...

Trying gpui...

We're excited to share that Firefox now uses zlib-rs for gzip (de)compression. This has both performance and safety advantages, but it took a while to get zlib-rs into production. Read why: trifectatech.org/blog/zlib-rs... Thanks to @gabrielesvelto.mas.to.ap.brid.gy and @mozilla.org #rustlang

The `int_format_into` feature is now stable! It provides a very fast integer to string conversions. It's a built-in `itoa` replacement (the inspiration of this feature): ``` use core::fmt::NumBuffer; let n = 0u32; let mut buf = NumBuffer::new(); let n_s: &str = n.format_into(&mut buf); ``` 1/1

Typst 0.15 is out now. It contains multitudes, including some of the most anticipated features and improvements that will smoothen your work. Learn what’s new in the thread below.

Serper?

Step 1 (getting a working account) done

🦀 I've now extracted the dataset to github.com/rust-lang/st... This should make it easier to other tools to make use of the data. The combined dataset is available as rust-lang.github.io/std-replacem... :) Contributions welcome! ✨ #rustlang

I've just signed up for staan.ai. Once they'll approve me I'll probably make a pi.dev extension

When I struggle to structure my thoughts about what's happening I turn to writing. Today about the recent US Anthropic ban news, what it says about power and dependency, and what it should mean for Europeans and citizens of the world. It's a long one. lucumr.pocoo.org/2026/6/13/am...

Marketing or not. Do not use tech from the USA

Fable + ultracode = 65 agents spawned, Max 5x session consumed in 16 minutes

Some speculation on my end: npm: adds bot protection to package pages so that they can avoid costly SSR. crates.io: 🖕 we'll return an empty shell. go call the API. we don't need to deal with the cost and security implications of running node on our servers. I like what crates.io did.

A bunch of fixes in rust-postgres brought to you by Fable and from the security team at the University of Sydney. github.com/rust-postgre...

This software is a smoking pile of trash. I waited until the last moment to upgrade Gitlab Registry and it still broke.

The recording of our talk at @canonicalltd.bsky.social's #UbuntuSummit is available now: www.youtube.com/watch?v=qKmR... Happy to answer questions here!

Reasoning around futures is improved, and it found multiple cases of lost wakers in massping.

I just reviewed rusty-s3...

Fable is already finding small issues that I didn't find in previous runs with Opus and Qwen. I'll try finding the time to scan all of the crates I maintain in in the following two weeks

my shai hulud series may be at an end: github.blog/changelog/20...

A few weeks ago I realized our websites accidentally disabled post-quantum key exchange because our haproxy had an old version of the modern ssl-config.mozilla.org config. Imagine how long it will take for everyone to realize this and fix their configs. This is not a scalable way of doing things...

This doesn't feel very solid

Ma... www.ilpost.it/2026/06/03/c...

Is this a way to flex that you're migrating from a shitty piece of software to another shitty piece of software?

I've updated Gitlab. The projects page started taking 3s, until enough refreshes got the cache to populate and reduced the response times to 200ms. How do you make something so bad. What are they doing???

Why did I subscribe to this conversation?

I wrote an article on epoll and io_uring basics through the lens of a simple HTTP file server written in C. Paywall has expired, give it a read! theconsensus.dev/p/2026/05/18...

More open source projects providing hosting should have a clear policy like this one. You can't expect CDN providers to provide infinite money. operations.osmfoundation.org/policies/til...

Hell yeah

📣 Big news for #RustLab2026! This year the conference is moving to a brand new location: the beautiful city of Bologna 🇮🇹 ❤️ A new city, a new venue, and the same amazing community spirit we love every year.

Absurd results with unsloth/Qwen3.6-35B-A3B-MTP-GGUF Q4_K_M

Small office server upgrade

Qwen3.6 now runs 2x faster with MTP GGUFs! Run locally on just 18GB RAM. ⚡️ MTP enables Qwen3.6 to generate ~1.4–2.2× faster with no accuracy change. Qwen3.6-27B MTP runs at 160 tokens/s. 35B-A3B reaches 240 t/s. GGUFs: huggingface.co/unsloth/Qwen... Guide: unsloth.ai/docs/models/...

Speedrun xkcd.com/3246/

Yesterday