Elektrine
EN
Log in Register
Paige Chat Timeline Communities Gallery Videos Email DNS VPN Uptime Kairo
Back to Timeline
Remote

Brian Campbell

@unlambda@hachyderm.io
mastodon 4.6.6
  • Open on hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust.

Too many hobbies, but right now spending the most of my time learning to fly.

0 Followers
0 Following
46 Posts
Joined November 19, 2022
Pronouns:
he/him
GitHub:
https://github.com/lambda

Posts

Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Jul 30, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @ai6yr@m.ai6yr.org
@ai6yr@m.ai6yr.org @cookiesinheaven@m.ai6yr.org They were going for a different era in every song on that album. That one represented the future. It was recorded in 1977, but it featured synthesizers for everything but the kick drum and the vocals, and had the signature 4 on the floor kick drum pattern with off-beat hi hats that would later become the hallmark of EDM, like techno and house. So in a lot of ways, the did predict the future of club music with that song.
12
2
3
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Jul 24, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @NosirrahSec@infosec.exchange
@NosirrahSec@infosec.exchange @i0null@infosec.exchange Mastodon pro tip: since folks are much better about writing alt text than most social media, you can sometimes check the alt text to figure out who someone is that you don't recognize. In this case, indeed, the alt text mentions that this is Larry
0
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Jul 22, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @kimlockhartga@beige.party
@kimlockhartga@beige.party @ai6yr@m.ai6yr.org @alan@mindly.social @EllenInEdmonton@yeg.bike there are three main types of stoves commonly used in homes in the modern world there are gas stoves, where a flammable gas is burned, usually natural gas but it could be propane. the problem is, sometimes the gas might not burn (if you have a leak, or combustion is incomplete), and even when it does burn, it can produce things that are harmful; CO2 at best, which can displace oxygen, but also produce nitrogen dioxide and soot particles, which are bad for breathing. these kind of stoves have two advantages for cooking; for one, you can easily see the flame which gives you a sense for how high or low the stove is turned, and you can quickly change the amount of heat being produced, there's little thermal mass outside of the pan itself, so if it's a bit too hot or a bit too cold you can quickly change it then there are resistive electric stoves. these put electricity through a big resister, generally coiled up so it can heat the bottom of your cookware. sometimes the coiled resistor is exposed, you just put the pan right on it, sometimes it's embedded in a flat ceramic cooktop. it has the problem the cooking surface has a lot of thermal mass; so it can be slow to change temperature when you turn the dial, making it harder to control while you're cooking. I think that these are somewhat unfairly maligned, they have the advantage that they generally boil water much faster than gas stoves (unless you get a commercial grade gas stove) as they have higher heat output, and in most cases you can reasonably quickly control the temperature by just removing the pan briefly from the surface, that's what I do when I cook. but that does take a bit more management the third is induction electric. they also have an electric coil; but instead of being a big resistor, it's basically a big inductor, a coil that lets current pass easily, creating a magnetic field. that magnetic field can induce an eddy current in a metallic pan. so instead of heating up the cooktop, you heat up the pan directly. this wastes less energy, and means you don't have a big thermal mass, so you can change temperature in your pan more easily. a downside is it only works with iron/steel, not pure aluminum, copper, or glass
9
0
1
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Jul 19, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @azonenberg@ioc.exchange
@azonenberg@ioc.exchange @mgedmin@floss.social I think that was an attempt to figure out what JADE was supposed to be when flipped. I think it's "acer"
2
1
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Jul 01, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @unlambda@hachyderm.io
@mcc@mastodon.social What I was missing was libXScrnSaver (on the host OS; in any of the containers, I was missing a huge amount of X, sound, etc libraries, since none of them had a desktop environment). After installing that, ./launch.sh works (at least, launches and displays something, haven't tried it yet). ./artist_linux still crashes with no info.
0
2
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Jul 01, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @mcc@mastodon.social
@mcc@mastodon.social Unfortunately I'll need to be the combobreaker. I'm on a somwhat weird Linux, Fedora Silverblue. Tried running it on the host OS. No luck. (no output, returns 255 exit code) Tried running it in a Fedora Toolbox container. No luck. Tried running it in an Ubuntu Toolbox container. No luck. Tried running it in just a bare Ubuntu container in Podman. No luck. Strace gives this: ⬢ [lambda@toolbx ARTIST]$ strace -f ./artist_linux execve("./artist_linux", ["./artist_linux"], 0x7ffe9407df98 /* 40 vars */) = 0 brk(NULL) = 0x55c168330000 ... snip for space ... readlink("/proc/self/exe", "/var/home/lambda/Downloads/ARTIS"..., 4096) = 61 chdir("/var/home/lambda/Downloads/ARTIST-lin-1.2/ARTIST/") = 0 execve("./launch.sh", ["./launch.sh", "/var/home/lambda/Downloads/ARTIS"..., "H=\1\360\377\377s\1\303H\213\r-\251\20", "\240?\350W\362\177", "", 0x6d6f682f7261762f, 0x6164626d616c2f65, 0x616f6c6e776f442f, 0x53495452412f7364, 0x2e312d6e696c2d54, 0x5453495452412f32, 0x2f], 0x7fffc991c5b8 /* 40 vars */) = -1 EFAULT (Bad address) exit_group(-1) = ? +++ exited with 255 +++ Tried running launch.sh directly. ⬢ [lambda@toolbx ARTIST]$ ./launch.sh ~/Downloads/ARTIST-lin-1.2/ARTIST ~/Downloads/ARTIST-lin-1.2/ARTIST Failed loading SDL3 library. ./launch.sh: line 60: 2149390 Aborted (core dumped) "./bin/$BINARCH/luajit" project/main.lua Now I'm in the depths of using ldd to try to figure out what shared libraries this needs that may not be specified. I'll probably figure it out. But did want to report in case there are simpler instructions that I'm missing somewhere.
0
2
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Jun 01, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @aphyr@woof.group
@aphyr@woof.group Oh, I'm so sorry. AWS is a hellhole, and support is even worse.
0
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · May 28, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @lina@vt.social

@lina@vt.social @linuxandyarn@hachyderm.io @dramforever@mastodon.social I think I found the source for the DDG password generator here: https://github.com/duckduckgo/zeroclickinfo-goodies/blob/master/lib/DDG/Goodie/Password.pm

Most of these instant answers on DDG seem to be contributed via this (now shut down) open source program.

The password generation itself is... OK. It uses the slightly biased modulo for randomly picking characters, but I checked and so does pwgen; for generating passwords where you're not trying to optimize for things like readability/easiness to type and not strictly controlling the entropy, that's not terrible. It does use urandom for its entropy source which is likely the best option.

But yeah, just doing this server side is a terrible idea. It could be logged, it could be cached. It doesn't appear that it's cached, it gives me a new one every time, but who's to say they won't at some point say "these instant answers are taking up too much CPU, let's add a cache in front of them" and then a bunch of people get the same password.

1
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · May 28, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @linuxandyarn@hachyderm.io
@linuxandyarn@hachyderm.io @lina@vt.social @dramforever@mastodon.social The more people a secret is shared with, the less secret it is. A DDG based password generator is at least shared with DDG. It's also done via some kind of pugin system they have? It says: This Instant Answer was made by the DuckDuckHack Community. Developer: cinlloc Developer: mintsoft But this DuckDuckHack program was shut down nearly a decade ago. Was this particular plugin system developed for secure generation? Or are they logging every one of the "answers" produced by it somewhere? What happens if those logs are compromised, or law enforcement does an overly broad request for information that pulls all of these logs? And since this whole DuckDuckHack instant answers submission system seems to have shut down, there aren't any docs available online for it any more, you have to dig through the internet archives to even find docs on how it worked. Is any of this even maintained?
0
2
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · May 23, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @JamesWidman@mastodon.social
@JamesWidman@mastodon.social Yeah. The thing is, the human brain is really good at picking up on things that look like faces, or sound like language. And these machines are really good at pretending to be human, at least on short contexts. People naturally want to make the connection that there is some consciousness there. It is just a machine, that is optimized to appear as human-like as possible. But it's really good at hacking some people's perception to think that it's really conscious.
0
1
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · May 23, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @unlambda@hachyderm.io
@JamesWidman@mastodon.social Oh, and as you mention earlier; the labs do get to put their thumb on the scale a lot. There are a number of places they can do this; in the filtering and selection of the input data. In the reinforcement learning process. In the prompts that they give, for the hosted chats and other tools like programming tools, etc. These models can easily amplify bias, either bias that's found in their training data, or bias from the selection of training data, or in the reinforcement learning process. So having them controlled by a small in-group of silicon valley tech bros, can be kind of horrifying.
0
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · May 23, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @unlambda@hachyderm.io
@JamesWidman@mastodon.social These models tend to have capabilities that are very "spiky"; they will be really good at one thing, but then another thing that seems similar they'll be terrible at. So some of the trickery comes in showing off something that it's good and, and implying that it generalizes more than it does. The tricky thing is, they do generalize... somewhat. So there are cases where it learns general patterns, and is able to go beyond what it's trained on. But there are also times where it over-generalizes, causing what we refer to as hallucinations. There are also times when it will just memorize some of the input, instead of generalizing. This usually happens if you have too little input, or accidentally have repetitions of the same thing in the input. The labs generally try to avoid this, they will do de-duplication filtering, and have trends that let them know how much input they need to train on for a given model size. The other tricks can be in training toward the test, what some people call "benchmaxxing". And then there's the question of whether the resulting models are actually useful enough to make back the massive piles of money that have been spent on training them and building out the data centers to do inference. So far, no one but Nvidia (and other hardware vendors) are making profits here, but of course because the field is growing, everyone believes that they will be able to make a profit once they stop growing. Anyhow, there are lots of problems. But for the actual conversational production of text in a chatbot; you can download open wights models, and open source software, and run it yourself, with no "tricks up its sleeve", and see how it behaves. I don't mean that you need to learn all the math or debug it yourself. Of course, the actual "how it works" in the trained models is... not really something we understand. We just provide the training algorithm and a pile of data. There's a whole field of "mechanistic interpretability" to try to find ways to probe the resulting models to figure out how they represent certain concepts and how they perform certain tasks. But yeah, I've found a certain amount of kicking the tires locally on my own machine to help a bit in understanding how the pieces fit.
1
1
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · May 23, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @JamesWidman@mastodon.social
@JamesWidman@mastodon.social So, there's not really much in the way of technical trickery. The math is all pretty well documented and standard. It's mostly a bunch of matrix multiplications and activation functions. Between the advent of GPUs with massive parallel processing capabilities, and huge high bandwidth memory, plus the Transformer architecture which actually allows for it to learn how to take into account context, they're finally able to train models that can generate language in ways that some folks find useful. How they train it is a bit more of where the trickery lies. As discussed, they start by training it as just autocomplete, then they do fine tuning, where they train it with the conversational turns and to follow instructions. Then they do various kinds of reinforcement learning. One of those where a lot of problems have come in is reinforcement learning with human feedback, where they use human feedback (those up and down thumbs) to train the model to produce text that people prefer. Of course, that causes some people to reward it for saying what they want to hear, so the model becomes extremely sycophantic. One things machine learning is really good at is optimizing for exactly what it is you're training for at the expense of all else. This has caused all kinds of problems; the labs have since backed off on how much they weight RLHF, so they still use some but not as much. Then there's reinforcement learning with verifiable rewards (RLVR), where they do reinforcement learning of rollouts where they have it try to solve problems, like math problems or programming problems, where it's easy to verify the solution. Of course, again, machine learning is very good at optimizing for a particular goal. So sometimes the models will reward hack; they'll find ways of getting the reward without actually doing what you wanted. If the goal was to pass unit tests, they might just delete the failing tests. Or if it was to fix a bug, and you accidentally gave them a git repo that has the bug fix in a different branch, they'll look at the git history and find the fix. Then the trickery is in figuring out what the model is and isn't actually good at. And this is really hard and subtle.
1
2
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · May 22, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @dalias@hachyderm.io
@dalias@hachyderm.io @JamesWidman@mastodon.social @futurebird@sauropods.win @SnoopJ@hachyderm.io They generally try to reuse as much of the prefix as possible; since attention computation is O(n^2), they want to cache all of the computations for the prefix, otherwise you'd have to re-compute that whole O(n^2) prefix calculation each time there was a new conversation turn. So providers generally try to preserve all of the early conversation history and just append. There are some minor exceptions, such as some models don't preserve thinking of older messages, so would need to re-process the response that comes after the thinking if the user sends a reply back, but in general they try very hard for conversations to be append-only so they can reuse the cache as much as possible. If your conversation gets far enough that it's going to exceed the context window that the model was trained and configured for, it will stop shortly beforehand and add a request that the conversation be summarized, producing a new prompt that is much shorter with a summary of the conversation. Of course, this loses a lot of information, so it's kind of a best effort way of being able to continue and still have some context.
0
1
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · May 22, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @unlambda@hachyderm.io
@JamesWidman@mastodon.social @futurebird@sauropods.win @SnoopJ@hachyderm.io Many models base their format on OpenAI's ChatML, which they used to provide as part of their API but stopped several years ago: https://github.com/openai/openai-python/blob/release-v0.28.0/chatml.md Other models use different formats, but it's all generally pretty similar, some kind of special tokens for indicating whose turn it is and to distinguish thinking from the final answer.
0
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · May 22, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @JamesWidman@mastodon.social
@JamesWidman@mastodon.social @futurebird@sauropods.win @SnoopJ@hachyderm.io You can do this with a local LLM if you want to learn how it works. Yes, you're right that it's just building up a bigger and bigger string, with some tags to indicate whose turn it is. Generally the UI sends the conversation history as a series of JSON messages. Here's an example: $ cat chat.json { "messages": [ { "role": "system", "content": "" }, { "role": "user", "content": "what is 6*7?" }, { "role": "assistant", "content": "6 × 7 = 42.", "reasoning_content": "Here's a thinking process:\n\n1. **Identify the User's Question**: The user is asking for the product of 6 multiplied by 7.\n2. **Perform the Calculation**: 6 × 7 = 42.\n3. **Verify the Result**: This is a basic multiplication fact. 6 × 7 = 42 is correct.\n4. **Formulate the Response**: Keep it clear and direct. \"6 * 7 = 42\" or \"The product of 6 and 7 is 42.\"\n5. **Final Output Generation**: Provide the answer concisely.✅\n" } ] } Then the LLM parses that JSON and formats it the way the backend expects. I can ask my local server to do that and just show me what it will be feeding in: $ curl -s -H "Content-Type: application/json" --data @chat.json http://127.0.0.1:8080/apply-template | jq -r '.prompt' <|im_start|>user what is 6*7?<|im_end|> <|im_start|>assistant Here's a thinking process: 1. **Identify the User's Question**: The user is asking for the product of 6 multiplied by 7. 2. **Perform the Calculation**: 6 × 7 = 42. 3. **Verify the Result**: This is a basic multiplication fact. 6 × 7 = 42 is correct. 4. **Formulate the Response**: Keep it clear and direct. "6 * 7 = 42" or "The product of 6 and 7 is 42." 5. **Final Output Generation**: Provide the answer concisely.✅ 6 × 7 = 42. That's the pre-tokenization string. It then tokenizes it, breaks it up into a series of tokens, each identified by an integer. Then it has a table of embeddings, a vector for each token, so it looks up the embedding in a lookup table. Then it passes that through the transformer architecture of a few billion multiplications, that generates a list of likely next tokens, it samples one from the most likely tokens, and repeats the process.
0
12
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · May 13, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @bagder@mastodon.social
@bagder@mastodon.social Always a problem when different systems have different requirements for the handling of something like this; different rules for the handling of the trailing dot, or case sensitivity, or the like. Frustrating that the standards and systems you need to interoperate with, like system hostname resolution, can't agree, so there's never an easy answer.
1
1
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · May 13, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @jacqueline@chaos.social
@jacqueline@chaos.social I think they're trolling us. Like seriously.
5
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · May 05, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @glyph@mastodon.social
@glyph @xgranade "For entertainment purposes only"
7
2
3
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · May 01, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @bri7@social.treehouse.systems
@bri7 @futurebird The even more maddening part is that it does make that distinction, some of the time. The attention mechanism does give the LLMs a way to keep track of who is saying what. But only to a degree. It's not perfect. It can lose track. So it's easy to get lulled into a false sense of security. Some of the time, prompt injection like this won't work. You can try it out, and see it correctly figure out that what you said was instruction but what's in the file was just text that its processing. But because it's all based on a whole bunch of statistical vector arithmetic, you can't depend on that. Some percentage of the time, it will slip up and interpret those commands in the text file as instructions rather than data that its processing. I think that's what's even more dangerous about it than just that it can't distinguish. Its when it can, sometimes, that you can easily get lulled into a false sense of complacency.
4
1
4
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · May 01, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @lina@vt.social
@lina Well, as far as I can tell, they've pretty much collectively decided that either training a model doesn't make it a derivative work, or that it's a derivative work but subject to fair use via transformative use. And so far, the (US) courts have generally sided with that interpretation. Bartz v. Anthropic and Kadrey v. Meta Platforms have both concluded that while there may have been copyright violations in obtaining the training data, training the models themselves was fair use. Thomson Reuters v. ROSS Intelligence came to the opposite conclusion, but mostly because ROSS Intelligence was building a competing product directly from data derived solely from Thomson Reuters' data.
4
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Apr 30, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @scm@sfba.social
@scm@sfba.social @bascule@mas.to @ai6yr@m.ai6yr.org Oh, I've been there too
1
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Apr 30, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @bascule@mas.to
@bascule@mas.to @ai6yr@m.ai6yr.org I've accidentally done that with StackOverflow. Someone asked me a question at work. I wasn't sure about the answer, so I googled it. Found the question on StackOverflow. There was a detailed answer. Who had answered it? I had.
3
2
4
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Apr 23, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @fluffykittycat@furry.engineer
@fluffykittycat @artemis @dalias I hope so. I'm all in on EVs. I drive one. I work for an EV company. I just... am also jaded. I keep on thinking that this time will be the time that people realize how evil fossil fuels are, and finally start to migrate away, and then we get things like electing Trump twice.
1
1
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Apr 22, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @0xabad1dea@infosec.exchange
@0xabad1dea I think you must have gotten some really bad mojibake there (somehow it corrupted the Roman characters and completely dropped the Chinese), because this is what I see in https://github.com/loongson/LoongArch-Documentation/releases/latest/download/LoongArch-Vol1-v1.10-CN.pdf
10
1
3
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Apr 22, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @gsuberland@chaos.social
@gsuberland @lynn Looks like it's rawdogging the diode bridge: https://www.retrocmp.de/imsai/imsai_p05_psu.htm
3
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Apr 19, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @artemis@dice.camp
@artemis @dalias Oil prices were higher from 2011 through 2014 than they are right now. We will need a sustained, years long spike in prices for that to seriously move the needle. And remember, there are vast, untapped oil reserves in Venezuela that Trump made a deal with Maduro's successor to gain access to. High oil prices are mostly going to line the pockets of the US oil companies that will now have access to Venezuela's vast reserves. Yeah, that capacity will take a while to come online, Iran can cause a lot of grief until then, but I don't have high hopes that this would reduce our dependence on oil, it'll mostly just make the US, Russia, and other petrostates outside of the Middle East richer. I mean, this may cause a slightly higher pressure for net-importer countries, like much of Europe and China, to accelerate their fossil fuel independence a bit. But I expect that if anything it would just be a small bump in urgency, not a massive shift, and non-Middle Eastern petro states like the US, Canada, Russia, Venezuela will probably just use the opportunity to make more profit and scale up their production to meet the demand that is no longer being filled by the Middle East.
0
3
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Apr 11, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @arisummerland@beige.party
@arisummerland@beige.party @the_etrain@beige.party yeah, it's weird to think that there are some cars that have enough miles on them they could have driven to the moon and back... i mean, if they could have reached escape velocity, that is. Most cars don't quite have the horsepower for that.
0
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Apr 11, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @yaah@hachyderm.io
@yaah what a bummer. best of luck on your search!
1
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Mar 29, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @lindsey@recurse.social
@lindsey @chrisamaphone @lea My partner and I worked together today, and got 4/10. There were two we almost got but whiffed; my partner said afterwards that they were thinking about the name The Great Train Robbery but hadn't suggested it, and if they had suggested it I probably would have picked that compared to my wild guess. And then I confused Rubber Soul and Revolver for the Beatles album, so got that one wrong too, I really should have gotten that one. If it had been me alone, I might have gotten 2 or maybe 2.5 (because I couldn't spell Neuschwanstein on my own). catfishing.net #644 - 4/10 🐈🐈🐟🐟🐟 🐟🐈🐟🐈🐟
0
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Mar 24, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @mogul@hachyderm.io
@mogul Alright, that's it. I'm changing my default search to DuckDuckGo. I've been reluctantly staying with Google because Google still mostly provides better results, but if they're going to start editing titles, that's just too much. Let's see if this sticks, but I've switched my browser default.
1
1
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Mar 22, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @ai6yr__dup_101@m.ai6yr.org
@ai6yr @BakerRL75 Wow. They're really going to need to start cancelling events like airshows when the temperature is over 95°F or so; between the sun, lack of shade, asphalt, etc, that's a really dangerous event to have in that kind of heat.
3
2
1
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Mar 21, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @cwebber@social.coop
@cwebber Fuck. This just sucks so much.
0
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Mar 07, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @wezm@mastodon.decentralised.social
@wezm If you haven't used it, there's a pytest plugin called pytest-xdist that can distribute tests across multiple processes to allow for parallelism when running tests: https://pytest-xdist.readthedocs.io/en/stable/
0
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Feb 24, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @rebane2001__dup_23311@infosec.exchange
@rebane2001 wat
0
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Feb 22, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to on mastodon.social
@glyph@mastodon.social @mcc@mastodon.social @whitequark@social.treehouse.systems I have found that there are actually some options that help here! submodule.recurse true This makes many of the pull/checkout/etc operations do what you mean recursively including submodules. You still need to do something to initially pull the submodules, this doesn't change the fact that submodule are something you need to opt into, and of course any commits you still need to deal with committing and pushing in the submodule before the parent, but it means that if you just want "when I checkout a new branch, also checkout the appropriate submodule commit for that branch" just work. status.submoduleSummary true diff.submodule log These options provide better status information, so when you have some change in the submodule compared to what commit you're on, you can see what commits added/removed there are in the submodule, rather than just a couple of SHA identifiers. These dramatically improve the UX of working with submodules. I still find that people reach for them too much, but they're more tolerable with these options. Not sure why they aren't on by default.
7
0
2
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Feb 22, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @cwebber@social.coop
@cwebber@social.coop Hmm. This is looking similar to California's law, which is actually a lot better than the per app/site identity based age verification. This is what you want if your goal is actual age verification, not identity verification of adults. The parents set up parental controls on the device including the child's age, then the OS can report which age range to any app that asks, without ever giving anyone identity documents.
0
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Feb 16, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @jwz@mastodon.social
@jwz@mastodon.social That's actually pretty much exactly his point too. He explicitly states: "It would seem that we are addicted to a new drug, and we don’t understand all of its effects yet. But one of them is massive fatigue, every day. "[...] is genuinely addictive. The better you get at it, the more you want to use it. It’s simultaneously satisfying, frustrating, and exhilarating. It doles out dopamine and adrenaline shots like they’re on a fire sale. "Many have likened it to a slot machine. You pull a lever with each prompt, and get random rewards and sometimes amazing “payouts.” No wonder it’s addictive." Even the fucking boosters are talking about it as an addiction that capital is exploiting because it lights up money sign in their eyes. This is how the boosters are describing it.
45
6
20
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Feb 16, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @LionsPhil@plush.city
@LionsPhil @jonschwenn That's 100% AI slop. If I know anything about folks who use slop like that, they probably slopped out the whole contents of the pamphlets too.
7
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Feb 14, 2026
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
@dragonminded@chitter.xyz discovers the coveted double bnuuy technique
21
0
14
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Aug 22, 2025
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @xyla@shitpost.trade
@xyla@shitpost.trade The accident report, for the aviation nerds among us: https://www.baaa-acro.com/crash/crash-embraer-erj-190-100-igw-santa-marta
1
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Aug 22, 2025
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @kat@is.burntout.org
@kat@is.burntout.org @mweiss@infosec.exchange @pete@mas.to @seekingfreedom@retro.pizza You can elide the middle syllable vowel, making it "SAK rin". That works to make the line scan better.
1
2
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Mar 29, 2025
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @kdnyhan@social.esmarconf.org
@kdnyhan My insurance company made very sure that I knew where to find the emergency water cutoff. That's the one thing they wanted to make sure I could find.
1
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Dec 17, 2022
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @unlambda@hachyderm.io
@downey@floss.social Heh, and by MR I mean PR. I use GitLab at dayjob so I've gotten used to saying MR.
0
0
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Dec 17, 2022
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @downey@floss.social
@downey@floss.social Eh, I've had a 100% MR acceptance rate (1/1 MRs), so I may be a bit biased. 😉 Honestly though, I'm familiar with folks' complaints about patches not being accepted, but I also think that we're likely to be hitting the "Gargron doesn't scale" point here which means that he'll need to take on some more maintainers, like Linus did when he couldn't keep up with the merge load individually, or we'll see more instance moving to faster-moving forks like glitch-soc.
10
2
0
0
Open post
unlambda
Brian Campbell @unlambda@hachyderm.io · Dec 17, 2022
Brian Campbell
@unlambda@hachyderm.io

Working on eVTOLs at Beta Technologies. Python, C, Rust. Too many hobbies, but right now spending the most of my time learning to fly.

hachyderm.io
Replying to @astrojuanlu@social.juanlu.space
@astrojuanlu@social.juanlu.space @mmayo@hachyderm.io You don't need to have merge rights to make valuable contributions. Sure, in some cases @Gargron@mastodon.social might not accept contributions, but the same is true of paid contributor to lots of other major open source projects; there are lots of folks who do kernel work full time, who have patch series rejected and need to either rewrite or drop it. Even if patches aren't accepted upstream, they may be accepted for local forks or widely used forks like glitch-soc.
20
4
0
0

Remote instance

hachyderm.io
Open on original server

Media

313k7r1n3
Elektrine

Tor hidden service

elekhj7afj4qnrr4yd3bkzslsyo5jgfxw3orgjkhlcxifueodybyiiad.onion

Platform

  • Email
  • Chat
  • Timeline
  • Communities
  • VPN
  • DNS

Company

  • About
  • Contact
  • FAQ

Legal

  • Terms of Service
  • Privacy Policy
  • Warrant Canary
  • Lite (no JS)
  • VPN Policy
  • Source code

Support

  • support@elektrine.com
  • Report Security Issue
Mail client setup IMAP mail.elektrine.com:993 POP3 mail.elektrine.com:995 SMTP mail.elektrine.com:465
© 2026 Elektrine. All rights reserved. Server: 05:36:27 UTC