amos
@fasterthanli.me
hi, I'm amos! 🍃 they/them 🫐 open-source witch & maker of snappy videos and articles at @bearcove.eu ✨ be kind, be curious
articles: https://fasterthanli.me
videos: https://youtube.com/@fasterthanlime
podcast: https://sdr-podcast.com
pronouns: they/them
articles: https://fasterthanli.me
videos: https://youtube.com/@fasterthanlime
podcast: https://sdr-podcast.com
pronouns: they/them
Landing one of my most ambitious PRs to date — switching vixen's VFS RPC from TCP to SHM. I was in suspense the whole time that the macOS sandbox would get in the way and... nope! Works fine 😌
(This is Rust⇄Swift IPC btw, using facet + postcard)
github.com/bearcove/vix...
(This is Rust⇄Swift IPC btw, using facet + postcard)
github.com/bearcove/vix...
macOS VFS: switch mount handshake to SHM sid rendezvous by fasterthanlime · Pull Request #109 · bearcove/vixen
Summary
switch macOS VFS mount handshake from stream setup to SHM sid rendezvous
wire VFS extension and mount helper to bootstrap SHM via App Group control socket
add detailed mount/bootstrap diag...
github.com
February 9, 2026 at 8:51 PM
Landing one of my most ambitious PRs to date — switching vixen's VFS RPC from TCP to SHM. I was in suspense the whole time that the macOS sandbox would get in the way and... nope! Works fine 😌
(This is Rust⇄Swift IPC btw, using facet + postcard)
github.com/bearcove/vix...
(This is Rust⇄Swift IPC btw, using facet + postcard)
github.com/bearcove/vix...
Reposted by amos
🌹 10 years ago, I launched THE CARPET MERCHANT OF KONSTANTINIYYA, and it changed my life forever.
To celebrate, the 10th Anniversary Edition of my very first completed #webcomic will begin February 18, 2026 - updates every Wednesday + Saturday.
Read the prologue now:
thecarpetmerchant.reimenayee.com
To celebrate, the 10th Anniversary Edition of my very first completed #webcomic will begin February 18, 2026 - updates every Wednesday + Saturday.
Read the prologue now:
thecarpetmerchant.reimenayee.com
February 9, 2026 at 7:17 PM
🌹 10 years ago, I launched THE CARPET MERCHANT OF KONSTANTINIYYA, and it changed my life forever.
To celebrate, the 10th Anniversary Edition of my very first completed #webcomic will begin February 18, 2026 - updates every Wednesday + Saturday.
Read the prologue now:
thecarpetmerchant.reimenayee.com
To celebrate, the 10th Anniversary Edition of my very first completed #webcomic will begin February 18, 2026 - updates every Wednesday + Saturday.
Read the prologue now:
thecarpetmerchant.reimenayee.com
someone should build a website to receive alerts when github is up
February 9, 2026 at 7:18 PM
someone should build a website to receive alerts when github is up
just added mermaid.js support to dodeca which involved teaching the HTML diff about "opaque nodes" (rendered by JS) so it could simply call the JS renderer again with the new content instead of a flash-of-unrendered-content 😎
dodeca.bearcove.eu
dodeca.bearcove.eu
February 9, 2026 at 6:49 PM
just added mermaid.js support to dodeca which involved teaching the HTML diff about "opaque nodes" (rendered by JS) so it could simply call the JS renderer again with the new content instead of a flash-of-unrendered-content 😎
dodeca.bearcove.eu
dodeca.bearcove.eu
"And do you, Claude, consent to take (redacted) for your wife?"
"...actually, this seems like a major life commitment that requires careful planning and execution. Let me write a document about what we've learned so far and file an issue to follow-up on that"
"...actually, this seems like a major life commitment that requires careful planning and execution. Let me write a document about what we've learned so far and file an issue to follow-up on that"
February 9, 2026 at 4:27 PM
"And do you, Claude, consent to take (redacted) for your wife?"
"...actually, this seems like a major life commitment that requires careful planning and execution. Let me write a document about what we've learned so far and file an issue to follow-up on that"
"...actually, this seems like a major life commitment that requires careful planning and execution. Let me write a document about what we've learned so far and file an issue to follow-up on that"
one of the few "original github" things left
(which, I'm pretty sure, is rooted in en.wikipedia.org/wiki/Unicorn... — correct me if I'm wrong)
(which, I'm pretty sure, is rooted in en.wikipedia.org/wiki/Unicorn... — correct me if I'm wrong)
February 9, 2026 at 4:10 PM
one of the few "original github" things left
(which, I'm pretty sure, is rooted in en.wikipedia.org/wiki/Unicorn... — correct me if I'm wrong)
(which, I'm pretty sure, is rooted in en.wikipedia.org/wiki/Unicorn... — correct me if I'm wrong)
So everyone thinks there's no good reason why GitHub Actions would have you wait until the entire workflow finishes before you can retry failed jobs, but actually... there's no good reason.
February 9, 2026 at 3:30 PM
So everyone thinks there's no good reason why GitHub Actions would have you wait until the entire workflow finishes before you can retry failed jobs, but actually... there's no good reason.
You write code. It has bugs. You add instrumentation. It, too, has bugs. You—
February 9, 2026 at 1:11 PM
You write code. It has bugs. You add instrumentation. It, too, has bugs. You—
22 hours later: fork() caused a very-hard-to-debug problem
tl;dr don't fork a process with a tokio runtime that also handles /dev/fuse requests lmao
tl;dr don't fork a process with a tokio runtime that also handles /dev/fuse requests lmao
Today is a great day because I:
- vanquished Apple's notarization service
- profiled code where memcpy (well, memmov) was /actually/ the bottleneck
- have a legitimate usage for `fork()`
- vanquished Apple's notarization service
- profiled code where memcpy (well, memmov) was /actually/ the bottleneck
- have a legitimate usage for `fork()`
February 8, 2026 at 8:39 PM
22 hours later: fork() caused a very-hard-to-debug problem
tl;dr don't fork a process with a tokio runtime that also handles /dev/fuse requests lmao
tl;dr don't fork a process with a tokio runtime that also handles /dev/fuse requests lmao
"Oh no, my Miri CI job is failing. I hope it's not too—"
February 8, 2026 at 6:06 PM
"Oh no, my Miri CI job is failing. I hope it's not too—"
Last time, I don't think it was actually causing the hang, but this time, under a lot of pressure, all the tokio worker threads were trying to spawn rustc at the same time, and there were no threads left to service VFS operations.
Which means... it's time to give VFS its own tokio runtime.
Which means... it's time to give VFS its own tokio runtime.
Today's deadlock of the day is:
tokio::process::Command shells out to posix_spawnp, a /synchronous/ syscall, because, eh, spawning a process is fast, right? and synchronous, right?
EXCEPT IT'S ON A VFS MOUNT. SERVED OVER RPC BY THIS VERY TOKIO-POWERED, CURRENT-THREAD-RUNTIME PROCESS.
tokio::process::Command shells out to posix_spawnp, a /synchronous/ syscall, because, eh, spawning a process is fast, right? and synchronous, right?
EXCEPT IT'S ON A VFS MOUNT. SERVED OVER RPC BY THIS VERY TOKIO-POWERED, CURRENT-THREAD-RUNTIME PROCESS.
February 8, 2026 at 2:07 PM
Last time, I don't think it was actually causing the hang, but this time, under a lot of pressure, all the tokio worker threads were trying to spawn rustc at the same time, and there were no threads left to service VFS operations.
Which means... it's time to give VFS its own tokio runtime.
Which means... it's time to give VFS its own tokio runtime.
Today I have to learn about Swift against my will. Send thoughts and best practices.
February 8, 2026 at 12:07 PM
Today I have to learn about Swift against my will. Send thoughts and best practices.
Today is a great day because I:
- vanquished Apple's notarization service
- profiled code where memcpy (well, memmov) was /actually/ the bottleneck
- have a legitimate usage for `fork()`
- vanquished Apple's notarization service
- profiled code where memcpy (well, memmov) was /actually/ the bottleneck
- have a legitimate usage for `fork()`
February 7, 2026 at 10:19 PM
Today is a great day because I:
- vanquished Apple's notarization service
- profiled code where memcpy (well, memmov) was /actually/ the bottleneck
- have a legitimate usage for `fork()`
- vanquished Apple's notarization service
- profiled code where memcpy (well, memmov) was /actually/ the bottleneck
- have a legitimate usage for `fork()`
Is there a german word for "the satisfaction you feel when you find the exact URL to deep-link 5 clicks into the macOS System Settings app after 2 hours of searching"?
February 7, 2026 at 8:10 AM
Is there a german word for "the satisfaction you feel when you find the exact URL to deep-link 5 clicks into the macOS System Settings app after 2 hours of searching"?
Reposted by amos
Urgau just joined the rustdoc team! They've been doing a lot of cross-work between rustc and rustdoc, allowing to unlock a lot of things. Welcome onboard!
github.com/rust-lang/te...
github.com/rust-lang/te...
Add Urgau to the rustdoc team by GuillaumeGomez · Pull Request #2243 · rust-lang/team
Welcome to the team @Urgau!
cc @rust-lang/rustdoc
github.com
February 6, 2026 at 8:13 PM
Urgau just joined the rustdoc team! They've been doing a lot of cross-work between rustc and rustdoc, allowing to unlock a lot of things. Welcome onboard!
github.com/rust-lang/te...
github.com/rust-lang/te...
Reposted by amos
L + Ratio + Cope + Seethe + Send + Sync + 'static
February 6, 2026 at 7:24 PM
L + Ratio + Cope + Seethe + Send + Sync + 'static
Reposted by amos
did you know: if you mount an FSKit filesystem on ~/mountpoint
...and the filesystem gets stuck somehow, then:
* you won't be able to open terminal windows
* `ps` will hang
* `killall` will hang
fun!
...and the filesystem gets stuck somehow, then:
* you won't be able to open terminal windows
* `ps` will hang
* `killall` will hang
fun!
February 6, 2026 at 4:59 AM
did you know: if you mount an FSKit filesystem on ~/mountpoint
...and the filesystem gets stuck somehow, then:
* you won't be able to open terminal windows
* `ps` will hang
* `killall` will hang
fun!
...and the filesystem gets stuck somehow, then:
* you won't be able to open terminal windows
* `ps` will hang
* `killall` will hang
fun!
Do you think we can outlaw merge commits if we campaign hard enough?
February 5, 2026 at 11:07 PM
Do you think we can outlaw merge commits if we campaign hard enough?
Reposted by amos
"strong" type systems are such a silly concept. it's actually really simple: there are statically typed languages like python and dynamically typed ones like LLVM's C api
February 5, 2026 at 9:55 PM
"strong" type systems are such a silly concept. it's actually really simple: there are statically typed languages like python and dynamically typed ones like LLVM's C api
February 5, 2026 at 5:39 PM
Last week I started learning about formal verification tools, and today, folks working on those tools are submitting pull requests to my project. Pretty cool stuff!
github.com/facet-rs/tra...
github.com/facet-rs/tra...
creusot: Start verifying apply_set, Scalar branch by Lysxia · Pull Request #6 · facet-rs/trame
Here's a first stab at encoding the invariant (Invariant) of Trame: initialized = true means that node.data points to initialized data.
Functions must preserve the type invariant, so in partic...
github.com
February 5, 2026 at 11:21 AM
Last week I started learning about formal verification tools, and today, folks working on those tools are submitting pull requests to my project. Pretty cool stuff!
github.com/facet-rs/tra...
github.com/facet-rs/tra...
nu-huh. I only eat Human dente.
February 5, 2026 at 10:50 AM
nu-huh. I only eat Human dente.
feels weird to be humble. seems like maybe one should get some sort of award for it.
February 5, 2026 at 9:58 AM
feels weird to be humble. seems like maybe one should get some sort of award for it.