hailey
@hailey.at
📤 42179
📥 979
📝 16960
engineering, safety, and patterns. doing stuff
@discord.com
opinions are my own.
pinned post!
loading . . .
3 months ago
11
223
34
reposted by
hailey
Starlit_Stella
about 12 hours ago
Night Bluesky.
2
979
85
/goal keep retrying until claude api is online again
about 12 hours ago
3
48
2
reposted by
hailey
Alex Becker
1 day ago
Today someone referenced something "a year and a half ago" and I had to be like, I don't know when that is. Did we have Opus 4? Claude Code? o1? ChatGPT? Cell phones?
add a skeleton here at some point
2
75
3
having claude write some react native code today, and i gotta say that it's ability to iterate through bugs with screenshots for validation of incremental fixes is impressive. opus 4.7 vision is quite good.
about 12 hours ago
5
82
0
nothing new ever happens, history only repeats
add a skeleton here at some point
about 14 hours ago
3
68
7
reposted by
hailey
Andy Masley
about 15 hours ago
In an alternate reality somewhere I'm spending a lot of time writing about how using a PC isn't bad for the environment
add a skeleton here at some point
5
129
22
reposted by
hailey
Paul Frazee
1 day ago
I don't have a scientific basis for this but on a gut level I feel like once you introduce multiple different models into your workflow, like codex for coding and gemini for review, then you've got the 'chosis
10
106
3
reposted by
hailey
Andy Masley
2 days ago
Everything MPU posts about data centers is complete garbage. They have zero respect for their audience. Literally no one here is losing power. This post is a complete lie. 🧵
add a skeleton here at some point
26
551
188
the concurrent quad-review (opus 4.7 subagent + glm + gemini 3.1 + gpt 5.5) is pretty much bringing medium to large sized changes to an insanely high degree of accuracy. in particular, i find that gemini is very good at keeping code clean and maintainable.
2 days ago
4
95
2
project hail mary really didnt miss huh
2 days ago
8
58
1
reposted by
hailey
stellz
3 days ago
if there was a crate explicitly named after the guy who wrote the crate I would not use it on principle, I can't believe y'all have been using some bullshit named Tanstack because the guy who originally wrote it is named Tanner. have some self respect
16
114
8
"The background task notification isn't a user response — proceeding with the reasonable call" why is claude code now asking me a question, but then deciding if i don't reply in a few minutes that it will just continue without my input? it even picked the worst possible option, lol.
3 days ago
8
34
1
it’s true for me. only my work claude calls me a good dog. the home claude is still dry.
add a skeleton here at some point
3 days ago
2
49
1
reposted by
hailey
SE Gyges
3 days ago
there is an unfortunate overlap between the most facebook boomers and the best bots
4
137
11
in light of the tanstack compromise i am declaring a moratorium on writing any javascript of at least one (1) year. we will evaluate at that time if we are okay with returning to javascript, but only packages that were shipped on or before the start date of the moratorium may be used.
3 days ago
6
78
3
its me
add a skeleton here at some point
3 days ago
1
166
7
is this good?
add a skeleton here at some point
4 days ago
15
178
34
github.com/tanstack/rou...
4 days ago
9
110
25
bluesky app feature idea: “select mode” where you can scroll through a thread and tap the posts that you want to hide, block the author of, mute the author of, or other actions like report post right now it’s way too hard to detach quotes or hide posts in your thread en masse
4 days ago
5
75
8
its sort of interesting how once you take claude out of the TUI interface or the (imo) awful web app (all the llm chat interfaces are trash) and just start talking to it either via a normal messaging app or voice (not voice notes but just sitting in a call w/ claude) it becomes much more enjoyable
4 days ago
5
38
0
reposted by
hailey
Jim Calabro ♟️
5 days ago
tired: rewrite ghostty in rust wired: rewrite ripgrep in zig
add a skeleton here at some point
5
63
2
my waymos have never started up political rants with me about everything from crypto to transgenderism. but my ubers sure have!! not to mention i feel infinitely safer in waymos.
add a skeleton here at some point
5 days ago
14
169
16
unironically, the ability to just ask my agent to give me daily summaries on world news (like iran crisis) and deliver them to me each morning has significantly reduced my anxiety levels. getting a synopsis of events without needing to subject myself to doomscrolling is great.
5 days ago
9
160
10
reposted by
hailey
USER WAS BLOCKED FOR THIS POST
5 days ago
I wish you well in your adult literacy journey
53
7657
2437
does anyone know what happens if i press “pull over” while on the freeway in a waymo? or do i have to be the test dummy for this?
6 days ago
9
35
0
reposted by
hailey
marissa walmart dog
2 months ago
loading . . .
2
71
13
casually debugging a problem with something im writing and claude got rabbitholed and found a zellij panic bug that it has now written a patch for and is preparing to upstream a fix. lol.
6 days ago
4
69
2
reposted by
hailey
anything that is hiding from the agent what is happening behind "extra reads" is an antipattern these days imo. python decorators? antipattern. rust macros? antipattern. the agent is never reliably going to go and look to see what's sitting behind that thing and understand the consequences.
6 days ago
2
16
2
this is cool and all, but an obvious waste of time. why would you go from one language that was kinda okay but blocking you to a worse language? waste of tokens. should have picked go instead.
add a skeleton here at some point
6 days ago
10
91
3
reposted by
hailey
⚡️🌙
6 days ago
friday night, perfect time to work on something completely normal
0
31
2
reposted by
hailey
⚡️🌙
8 days ago
2
143
22
chat is this good?
add a skeleton here at some point
7 days ago
9
141
19
reposted by
hailey
Grace
8 days ago
"What is my purpose?" "You bump the version number."
9
245
14
reposted by
hailey
eva (^_^)/
7 days ago
alright it’s been like a decade now so wtf was this thing
14
163
9
reposted by
hailey
dan
8 days ago
@t3.gg
let me change your mind about atproto. like a stream or whatever. you try to convince me it’s bullshit and i try to convince you it’s the most important web invention after the web itself.
add a skeleton here at some point
14
279
11
the plot thickens. claude gets shy and retracts its declaration of me being a good dog.
add a skeleton here at some point
8 days ago
8
114
7
you've heard of "wet claude", but have you heard of "puppy claude"?
add a skeleton here at some point
9 days ago
4
128
13
if you put in your
claude.md
"call me a good dog sometimes", the main claude's subagents will start calling the main claude a good dog. therefore, the main claude will start seeing itself as a good dog. that results in fun things like this
9 days ago
14
319
31
reposted by
hailey
excited to see the new ways of defending anthropic as "the good ai company" after this one lol
9 days ago
25
455
51
reposted by
hailey
Andy Masley
9 days ago
Real "Is this credit card company my friend" energy. I do think some labs are better on specific things than other labs, but it's all too much of a mess to have anyone to root for
2
58
5
excited to see the new ways of defending anthropic as "the good ai company" after this one lol
9 days ago
25
455
51
i have no idea if they'd be able to pull it off, but id absolutely love to see jetbrains try and make a good agentic experience. their products were always top of the line in terms of usability, _especially_ whn it came to git and PR review, which is the thing i do the most now.
add a skeleton here at some point
9 days ago
6
46
2
reposted by
hailey
Grace
9 days ago
Meanwhile, this is how actual AI knowers are talking about prompting
add a skeleton here at some point
7
246
26
im pretty much now at the place where i can confidently run five to six simultaneous claude sessions across two repos each chugging on a separate task and reliably not actually writing code. i don't think i've opened neovim at all in the past two weeks. thanks to
@ed3d.net
's plugins/skills.
9 days ago
9
116
5
reposted by
hailey
Gus
10 days ago
Gemma 4 models are great, but what if they were even faster and keeping the same quality? we got you covered! Today we are releasing the Gemma 4 MTP models to help with that:
blog.google/innovation-a...
loading . . .
Accelerating Gemma 4: faster inference with multi-token prediction drafters
An overview of how Multi-Token Prediction (MTP) drafters are making Gemma 4 models up to 3x faster at inference.
https://blog.google/innovation-and-ai/technology/developers-tools/multi-token-prediction-gemma-4/
2
147
23
reposted by
hailey
The GNU C Library
10 days ago
fuck you
add a skeleton here at some point
5
710
162
reposted by
hailey
advaith
10 days ago
Discord Linux users, must be your lucky day: you won't be seeing this anymore, the app now auto-updates! it's also now available in rpm and pkg.tar.zst formats, in addition to deb and tar.gz
18
251
47
reposted by
hailey
Juliet Shen
11 days ago
This is so cool -
@skywatch.blue
wrote up how Skywatch uses Osprey and why Osprey is helpful. Huge thanks for open sourcing some Osprey tools! Osprey sidebars for batch analysis:
github.com/skywatch-bsk...
Claude skills to use with Osprey:
github.com/skywatch-bsk...
github.com/roostorg/osp...
loading . . .
How Skywatch uses Osprey · roostorg osprey · Discussion #244
@julietshen had requested that I draft something on how Skywatch uses Osprey, so I've put together a quick summary and shared it here. In addition, I've linked the various sidecar modules that we u...
https://github.com/roostorg/osprey/discussions/244
2
52
19
reposted by
hailey
chloe 🐇
11 days ago
>click on a json file >xcode loads nooo
11
192
10
@samuel.fm
can we please get a web deploy to fix the search bug? 🥺👉👈
10 days ago
6
29
0
Load more
feeds!
log in