![doodlestein's tweet photo. I made another tool out of my own desperation because my agents kept filling up disks (including special locations that are memory-backed) with Rust build artifacts and making the machines unresponsive.
It’s called storage_ballast_helper (“sbh”) and you can get it here:
https://t.co/QfTOmGwF5z
It’s based on the concept of ballast used on ships. The basic idea is simple: create a bunch of sacrificial files filled with random data (e.g., 1GB per file). When you start to experience disk-space pressure, you can buy yourself some time by deleting these ballast files one at a time.
What does that get you, and isn’t that just delaying the inevitable? Or does it make the problem even worse because now you’re “wasting” space on the ballast files?
It gets you time to intelligently analyze the situation and decide what substantive actions to take.
sbh implements a sophisticated multi-factor scoring model that draws up a “hit list” of the best candidates for deletion: stuff that’s clearly not being used, that takes up a lot of space, that’s obviously not something useful like code or .md files, etc. Based on that score, it decides what to delete to optimize the system. (It does a lot more, too.)
Rather than describe everything myself, I asked Claude Code to finish this post based on a careful analysis of the code. Here’s what it told me (everything above this line is mine):
Here are some additional angles worth mentioning that would resonate on X, based on what’s actually in the codebase:
The control theory angle (this tends to get engagement from technical audiences):
It uses a real PID controller (like cruise control) with EWMA rate estimation to predict disk exhaustion 30 minutes ahead and start reacting before you hit critical. Not cron-job-every-5-minutes stuff. 1-second polling with adaptive smoothing that tracks bursty consumption.
The safety story (people love “what could go wrong” engineering):
Six independent safety layers, any one of which can veto a deletion. It checks for open file handles via /proc/fd so it won’t nuke a build directory mid-compilation. It detects .git directories as a hard veto. It uses a circuit breaker that halts after 3 consecutive failures. And the scoring uses a Bayesian decision-theoretic framework where wrongly deleting something useful is weighted costlier (50) than failing to clean up something stale (30).
The zero-write emergency mode (great hook):
When your disk is at 99%, most cleanup tools fail because they need to write temp files. sbh has a zero-write emergency mode: pure in-memory scoring, direct unlink calls. It can recover a system that nothing else can touch.
The progressive deployment model (borrowed from feature-flag practice):
It has observe/canary/enforce modes like a feature flag rollout. You can shadow-run it to see what it would delete without deleting anything, then graduate to capped real deletions, then full enforcement. If its own predictions drift from reality, it auto-demotes to safe mode.
The “why not just…” preemption:
Why not cron + rm? No pressure awareness; can’t tell a 2-hour-old build artifact from a 2-hour-old source file. Why not tmpwatch? Only knows about /tmp. Why not quotas? The problem is aggregate consumption across legitimate workloads, not one bad actor.
The Rust angle:
#![forbid(unsafe_code)], no async runtime, just OS threads with crossbeam channels. ~25K lines of Rust. Full TUI dashboard with 7 screens, incident playbooks, and a command palette.
A punchy closer:
It even monitors /dev/shm and tmpfs separately because RAM-backed filesystems filling up is a different kind of catastrophe than disk filling up.](https://pbs.twimg.com/media/HBVdl43XwAE4Iyl.jpg)
![doodlestein's tweet photo. I made another tool out of my own desperation because my agents kept filling up disks (including special locations that are memory-backed) with Rust build artifacts and making the machines unresponsive.
It’s called storage_ballast_helper (“sbh”) and you can get it here:
https://t.co/QfTOmGwF5z
It’s based on the concept of ballast used on ships. The basic idea is simple: create a bunch of sacrificial files filled with random data (e.g., 1GB per file). When you start to experience disk-space pressure, you can buy yourself some time by deleting these ballast files one at a time.
What does that get you, and isn’t that just delaying the inevitable? Or does it make the problem even worse because now you’re “wasting” space on the ballast files?
It gets you time to intelligently analyze the situation and decide what substantive actions to take.
sbh implements a sophisticated multi-factor scoring model that draws up a “hit list” of the best candidates for deletion: stuff that’s clearly not being used, that takes up a lot of space, that’s obviously not something useful like code or .md files, etc. Based on that score, it decides what to delete to optimize the system. (It does a lot more, too.)
Rather than describe everything myself, I asked Claude Code to finish this post based on a careful analysis of the code. Here’s what it told me (everything above this line is mine):
Here are some additional angles worth mentioning that would resonate on X, based on what’s actually in the codebase:
The control theory angle (this tends to get engagement from technical audiences):
It uses a real PID controller (like cruise control) with EWMA rate estimation to predict disk exhaustion 30 minutes ahead and start reacting before you hit critical. Not cron-job-every-5-minutes stuff. 1-second polling with adaptive smoothing that tracks bursty consumption.
The safety story (people love “what could go wrong” engineering):
Six independent safety layers, any one of which can veto a deletion. It checks for open file handles via /proc/fd so it won’t nuke a build directory mid-compilation. It detects .git directories as a hard veto. It uses a circuit breaker that halts after 3 consecutive failures. And the scoring uses a Bayesian decision-theoretic framework where wrongly deleting something useful is weighted costlier (50) than failing to clean up something stale (30).
The zero-write emergency mode (great hook):
When your disk is at 99%, most cleanup tools fail because they need to write temp files. sbh has a zero-write emergency mode: pure in-memory scoring, direct unlink calls. It can recover a system that nothing else can touch.
The progressive deployment model (borrowed from feature-flag practice):
It has observe/canary/enforce modes like a feature flag rollout. You can shadow-run it to see what it would delete without deleting anything, then graduate to capped real deletions, then full enforcement. If its own predictions drift from reality, it auto-demotes to safe mode.
The “why not just…” preemption:
Why not cron + rm? No pressure awareness; can’t tell a 2-hour-old build artifact from a 2-hour-old source file. Why not tmpwatch? Only knows about /tmp. Why not quotas? The problem is aggregate consumption across legitimate workloads, not one bad actor.
The Rust angle:
#![forbid(unsafe_code)], no async runtime, just OS threads with crossbeam channels. ~25K lines of Rust. Full TUI dashboard with 7 screens, incident playbooks, and a command palette.
A punchy closer:
It even monitors /dev/shm and tmpfs separately because RAM-backed filesystems filling up is a different kind of catastrophe than disk filling up.](https://pbs.twimg.com/media/HBVdii5bkAAwKlU.jpg)
![doodlestein's tweet photo. I made another tool out of my own desperation because my agents kept filling up disks (including special locations that are memory-backed) with Rust build artifacts and making the machines unresponsive.
It’s called storage_ballast_helper (“sbh”) and you can get it here:
https://t.co/QfTOmGwF5z
It’s based on the concept of ballast used on ships. The basic idea is simple: create a bunch of sacrificial files filled with random data (e.g., 1GB per file). When you start to experience disk-space pressure, you can buy yourself some time by deleting these ballast files one at a time.
What does that get you, and isn’t that just delaying the inevitable? Or does it make the problem even worse because now you’re “wasting” space on the ballast files?
It gets you time to intelligently analyze the situation and decide what substantive actions to take.
sbh implements a sophisticated multi-factor scoring model that draws up a “hit list” of the best candidates for deletion: stuff that’s clearly not being used, that takes up a lot of space, that’s obviously not something useful like code or .md files, etc. Based on that score, it decides what to delete to optimize the system. (It does a lot more, too.)
Rather than describe everything myself, I asked Claude Code to finish this post based on a careful analysis of the code. Here’s what it told me (everything above this line is mine):
Here are some additional angles worth mentioning that would resonate on X, based on what’s actually in the codebase:
The control theory angle (this tends to get engagement from technical audiences):
It uses a real PID controller (like cruise control) with EWMA rate estimation to predict disk exhaustion 30 minutes ahead and start reacting before you hit critical. Not cron-job-every-5-minutes stuff. 1-second polling with adaptive smoothing that tracks bursty consumption.
The safety story (people love “what could go wrong” engineering):
Six independent safety layers, any one of which can veto a deletion. It checks for open file handles via /proc/fd so it won’t nuke a build directory mid-compilation. It detects .git directories as a hard veto. It uses a circuit breaker that halts after 3 consecutive failures. And the scoring uses a Bayesian decision-theoretic framework where wrongly deleting something useful is weighted costlier (50) than failing to clean up something stale (30).
The zero-write emergency mode (great hook):
When your disk is at 99%, most cleanup tools fail because they need to write temp files. sbh has a zero-write emergency mode: pure in-memory scoring, direct unlink calls. It can recover a system that nothing else can touch.
The progressive deployment model (borrowed from feature-flag practice):
It has observe/canary/enforce modes like a feature flag rollout. You can shadow-run it to see what it would delete without deleting anything, then graduate to capped real deletions, then full enforcement. If its own predictions drift from reality, it auto-demotes to safe mode.
The “why not just…” preemption:
Why not cron + rm? No pressure awareness; can’t tell a 2-hour-old build artifact from a 2-hour-old source file. Why not tmpwatch? Only knows about /tmp. Why not quotas? The problem is aggregate consumption across legitimate workloads, not one bad actor.
The Rust angle:
#![forbid(unsafe_code)], no async runtime, just OS threads with crossbeam channels. ~25K lines of Rust. Full TUI dashboard with 7 screens, incident playbooks, and a command palette.
A punchy closer:
It even monitors /dev/shm and tmpfs separately because RAM-backed filesystems filling up is a different kind of catastrophe than disk filling up.](https://pbs.twimg.com/media/HBVdZ9iWcAEuwb0.jpg)
Olivia
Online
Bryan Archilla
@BryanArchilla
E-mail: [email protected]
Joined May 2012
1.5K Following
425 Followers
14.3K Posts
Pinned Tweet
I really liked #HaloInfinite, so I made a video about it.
A little celebration of the nostalgia bug that it hit for me and why I think Halo is such an important franchise for me.
https://t.co/yOiwmbjXPQ
Here's a link to my master content database template, and a page that explains a little how I do things.
https://t.co/BZhrwFLPXM
This is just the template for the content database where I craft all my long and short form content, not for the document I use sharing with brands or anything with n8n. I moved my brand-sharing to my website and don't use Notion for this anymore, and I don't have any n8n templates that are share ready.
@Naaackers Your lighting is exceptional, I've been interested in finding out how the s9 would work for me. I think it would make a great companion camera for corporate shoots and a smaller way to do normal content.
@Naaackers awe man, I hope the little guy pulls through okay. It's so stressful, take all the time that you need brother.
Who to follow
Ryan Sprague
@RyanSprague51
Podcast Host @SomewhereSkies | Author
Grandiose_Nature
@KingCarrtoon
Character is the most important thing
Iona Morrison
@ionacrv
#bookexcellenceaward Bestselling author of The Blue Cove Mystery Series, published by The Wild Rose Press. Writing paranormal, mystery, and romance.
@Naaackers I know your screen was filthy after that LMFAO
You know those edits of incels making female characters "hotter". DLSS5 now does it for you. LMFAOOO.
This is why you cant afford hard drive or PC parts. Hellscape.
Shout to @Naaackers for letting us know about the control panel, sound management thing in Windows 11,
The knowledge in that section of the video https://t.co/wO3FSJ8MsE just finally let me minimize the echo of me and my wife playing games in the same room.
<3
@Vexian This isn't about reiterating. Its literally stealing work word for word. Making money off the backs of people who actually did the research and spent the time investing themselves into the work.
Those people need economic support to keep doing what they are doing.
@Vexian If you watch the video. You'll hopefully gleam thst this isnt about yoinking and twisting and making similar content. This is about how content farms and individuals can economically hurt journalistic work and the people doing it in a way that is harmful for general truth.
@saradietschy Hey sara! I fit that criteria pretty well! I even have a cloud workflow set up with resolve. Ill try to dm or hunt down your email through your website as well. Leaving the comment just in case its your preferred method of contact
It has come to our attention that a cloned version of our
vertical plugin has been repackaged and distributed by @Restreamio
This has caused issues for streamers seeking support from us in our discord.
Below is our statement addressing this issue.
@Naaackers I cant help but picture him with a little wrench in his mouth trying to help
So you can aim in a third person perspective?
why can't video game characters be centre frame anymore
I understand that Ai is here to stay. But you cant tell me youre connecting me to a support agent and then connect me to fucking Mary the AI assistant.
The amount of loopholes to talk to a person are mental
@FrostyCoolSlug Hey man. We dont know each other. But offering any support I can offer here.
I struggle with depression and am currently medicated.
Would love to chat and listen to ya
I made another tool out of my own desperation because my agents kept filling up disks (including special locations that are memory-backed) with Rust build artifacts and making the machines unresponsive.
It’s called storage_ballast_helper (“sbh”) and you can get it here:
https://t.co/QfTOmGwF5z
It’s based on the concept of ballast used on ships. The basic idea is simple: create a bunch of sacrificial files filled with random data (e.g., 1GB per file). When you start to experience disk-space pressure, you can buy yourself some time by deleting these ballast files one at a time.
What does that get you, and isn’t that just delaying the inevitable? Or does it make the problem even worse because now you’re “wasting” space on the ballast files?
It gets you time to intelligently analyze the situation and decide what substantive actions to take.
sbh implements a sophisticated multi-factor scoring model that draws up a “hit list” of the best candidates for deletion: stuff that’s clearly not being used, that takes up a lot of space, that’s obviously not something useful like code or .md files, etc. Based on that score, it decides what to delete to optimize the system. (It does a lot more, too.)
Rather than describe everything myself, I asked Claude Code to finish this post based on a careful analysis of the code. Here’s what it told me (everything above this line is mine):
Here are some additional angles worth mentioning that would resonate on X, based on what’s actually in the codebase:
The control theory angle (this tends to get engagement from technical audiences):
It uses a real PID controller (like cruise control) with EWMA rate estimation to predict disk exhaustion 30 minutes ahead and start reacting before you hit critical. Not cron-job-every-5-minutes stuff. 1-second polling with adaptive smoothing that tracks bursty consumption.
The safety story (people love “what could go wrong” engineering):
Six independent safety layers, any one of which can veto a deletion. It checks for open file handles via /proc/fd so it won’t nuke a build directory mid-compilation. It detects .git directories as a hard veto. It uses a circuit breaker that halts after 3 consecutive failures. And the scoring uses a Bayesian decision-theoretic framework where wrongly deleting something useful is weighted costlier (50) than failing to clean up something stale (30).
The zero-write emergency mode (great hook):
When your disk is at 99%, most cleanup tools fail because they need to write temp files. sbh has a zero-write emergency mode: pure in-memory scoring, direct unlink calls. It can recover a system that nothing else can touch.
The progressive deployment model (borrowed from feature-flag practice):
It has observe/canary/enforce modes like a feature flag rollout. You can shadow-run it to see what it would delete without deleting anything, then graduate to capped real deletions, then full enforcement. If its own predictions drift from reality, it auto-demotes to safe mode.
The “why not just…” preemption:
Why not cron + rm? No pressure awareness; can’t tell a 2-hour-old build artifact from a 2-hour-old source file. Why not tmpwatch? Only knows about /tmp. Why not quotas? The problem is aggregate consumption across legitimate workloads, not one bad actor.
The Rust angle:
#![forbid(unsafe_code)], no async runtime, just OS threads with crossbeam channels. ~25K lines of Rust. Full TUI dashboard with 7 screens, incident playbooks, and a command palette.
A punchy closer:
It even monitors /dev/shm and tmpfs separately because RAM-backed filesystems filling up is a different kind of catastrophe than disk filling up.
![doodlestein's tweet photo. I made another tool out of my own desperation because my agents kept filling up disks (including special locations that are memory-backed) with Rust build artifacts and making the machines unresponsive.
It’s called storage_ballast_helper (“sbh”) and you can get it here:
https://t.co/QfTOmGwF5z
It’s based on the concept of ballast used on ships. The basic idea is simple: create a bunch of sacrificial files filled with random data (e.g., 1GB per file). When you start to experience disk-space pressure, you can buy yourself some time by deleting these ballast files one at a time.
What does that get you, and isn’t that just delaying the inevitable? Or does it make the problem even worse because now you’re “wasting” space on the ballast files?
It gets you time to intelligently analyze the situation and decide what substantive actions to take.
sbh implements a sophisticated multi-factor scoring model that draws up a “hit list” of the best candidates for deletion: stuff that’s clearly not being used, that takes up a lot of space, that’s obviously not something useful like code or .md files, etc. Based on that score, it decides what to delete to optimize the system. (It does a lot more, too.)
Rather than describe everything myself, I asked Claude Code to finish this post based on a careful analysis of the code. Here’s what it told me (everything above this line is mine):
Here are some additional angles worth mentioning that would resonate on X, based on what’s actually in the codebase:
The control theory angle (this tends to get engagement from technical audiences):
It uses a real PID controller (like cruise control) with EWMA rate estimation to predict disk exhaustion 30 minutes ahead and start reacting before you hit critical. Not cron-job-every-5-minutes stuff. 1-second polling with adaptive smoothing that tracks bursty consumption.
The safety story (people love “what could go wrong” engineering):
Six independent safety layers, any one of which can veto a deletion. It checks for open file handles via /proc/fd so it won’t nuke a build directory mid-compilation. It detects .git directories as a hard veto. It uses a circuit breaker that halts after 3 consecutive failures. And the scoring uses a Bayesian decision-theoretic framework where wrongly deleting something useful is weighted costlier (50) than failing to clean up something stale (30).
The zero-write emergency mode (great hook):
When your disk is at 99%, most cleanup tools fail because they need to write temp files. sbh has a zero-write emergency mode: pure in-memory scoring, direct unlink calls. It can recover a system that nothing else can touch.
The progressive deployment model (borrowed from feature-flag practice):
It has observe/canary/enforce modes like a feature flag rollout. You can shadow-run it to see what it would delete without deleting anything, then graduate to capped real deletions, then full enforcement. If its own predictions drift from reality, it auto-demotes to safe mode.
The “why not just…” preemption:
Why not cron + rm? No pressure awareness; can’t tell a 2-hour-old build artifact from a 2-hour-old source file. Why not tmpwatch? Only knows about /tmp. Why not quotas? The problem is aggregate consumption across legitimate workloads, not one bad actor.
The Rust angle:
#![forbid(unsafe_code)], no async runtime, just OS threads with crossbeam channels. ~25K lines of Rust. Full TUI dashboard with 7 screens, incident playbooks, and a command palette.
A punchy closer:
It even monitors /dev/shm and tmpfs separately because RAM-backed filesystems filling up is a different kind of catastrophe than disk filling up.](https://pbs.twimg.com/media/HBVdrlOW8AAgTLz.jpg)
@StevenDoud440 Youre an insane human being.
Perfect example of bully gets challenged and they run with his tail between his legs.
@iSageSen @Naaackers Hey! Just so you know. I would absolutely buy this if it became available. That goes so hard.
Fun fact: Snake is the only voice I ever try an impression of.
One of these days ill Crack it.
Hey @DavidBHayter years ago, me and my buddies sent you a script for an animated cartoon parody of the spy genre.
I found the script again and remember the kind words and encouragement you gave some young writers.
I carry that memory with me everytime I write something.
Thank you for that!
When im feeling bad about my creative output or cant write. I keep trying, partly because you took the time to read a silly idea and treat it seriously.
Cheers <3
Last Seen Users on Sotwe
PUTRI DEWI LESTARI
Seen from United States
♥️sultan birol♥️
Seen from Turkey
Homura_ra
Seen from United States
Đạt đủng đỉnh
Seen from Vietnam
GayAmateursProd 
Seen from Italy
tiểu thái
Seen from Vietnam
Syafiera Anis
Seen from United States
semra.akk41
Seen from Turkey
Nacho +130k
Seen from Argentina
Maryam Fatima 🎀
Seen from Germany
Most Popular Users
Elon Musk
@elonmusk
240.2M followers
Barack Obama
@barackobama
119.3M followers
Donald J. Trump
@realdonaldtrump
111.6M followers
Cristiano Ronaldo
@cristiano
109.4M followers
Narendra Modi
@narendramodi
106.9M followers
Rihanna
@rihanna
97.4M followers
NASA
@nasa
92.1M followers
Justin Bieber
@justinbieber
90.7M followers
KATY PERRY
@katyperry
87M followers
Taylor Swift
@taylorswift13
80.9M followers
Lady Gaga
@ladygaga
72.4M followers
Kim Kardashian
@kimkardashian
69.5M followers
Virat Kohli
@imvkohli
68.9M followers
YouTube
@youtube
68.6M followers
Bill Gates
@billgates
63.5M followers
The Ellen Show
@theellenshow
62.5M followers
CNN
@cnn
61.9M followers
Neymar Jr
@neymarjr
61.5M followers
X
@x
60.9M followers
Selena Gomez
@selenagomez
60.1M followers