archevel's recent activity
-
Comment on Timasomo 2026: Week 3 Updates in ~creative.timasomo
-
Comment on What can be done about Linux being slopified? in ~tech
archevel Link ParentGood points! While all structured formats aren't programming languages I get your point. This part I don't necessarily think is true. A lot of compiled languages ship bloated binaries which can be...Structured format for defining tasks is a programming language.
Structured format for defining expectations is programming and tests.Good points! While all structured formats aren't programming languages I get your point.
If the compiled assembly is an incoherent mess, the software will run like an incoherent mess, and we should definitely care, and we should poke at the compiler until we understand why it's doing what it is doing.
This part I don't necessarily think is true. A lot of compiled languages ship bloated binaries which can be really hard to unwind. Sometimes these run poorly too, but that's not a given. It's great if we can understand the underlying assembly, but that's not a requirement for functioning and working software. I'd even argue that as an industry pre llms few pieces of software had someone who fully understood it. If you go into legacy systems even the purpose of the system could be lost a decade ago... Now that doesn't mean we shouldn't care about software quality, just that the situation hasn't really changed much beside the pace at which (often por quality) software is produced.
-
Comment on What can be done about Linux being slopified? in ~tech
archevel Link ParentLLMs, as far as I understand the underlying implementations, are deterministic. Given the same input (text, seed, temperature etc) it will give you the same output. They behave non-deterministic...LLMs, as far as I understand the underlying implementations, are deterministic. Given the same input (text, seed, temperature etc) it will give you the same output. They behave non-deterministic just because we usually currently run them that way. Committing the promts, models versions and other input into VCS would let you recreate the output deterministically. I think there is some nuances that might impact this in practice with how the service providers run the models (batching and hardware differences might impact the output, but I'm not sure as to what extent). Local models should mitigate that to a large extent of needed.
Way back were the compilers really deterministic? Are they that way even now? If I compile something on my laptop will I get the exact same binary as you would? I don't think so. Most likely there will be differences due to (in no particular order):
- compilers versions
- compilers flags
- underlying hardware
- operating system
There are probably other confounding issues there too. Some people can probably understand those differences in detail, most devs probably shrug and don't really care. If it compiles and pass the tests we consider them equivalent to another build.
I agree that English is a terrible programming language. So maybe we should look at ways of optimizing the input so we get better output. Maybe some semi structured format for defining tasks and expectations? Point is that today most devs don't care if the compiled assembly is "an incoherent mess", potentially we shouldn't care if the source code we compile is either.
I haven't really landed in what I think. I kind of mourn the loss of the craft. I like programming as a discipline and I don't enjoy poking an agent to do the right thing. It does however look like that's the direction the industry is barreling towards...
-
Comment on What can be done about Linux being slopified? in ~tech
archevel Link ParentI am a bit unsure about this, but could we not liken LLMs to the introduction of compilers? Before those became common place people wrote assembly and even as far as the 90s people were better at...I am a bit unsure about this, but could we not liken LLMs to the introduction of compilers? Before those became common place people wrote assembly and even as far as the 90s people were better at writing assembly code than compilers could reasonably produce. This might still be the case today, but I doubt it. It is definitely not true for most developers! Compilers now take a high level descriptions of the operations we want to perform and boils it down to some bare essentials. Now with LLMs we take a sequence of fairly vaguely worded instructions and generate a sequence of operations that then compilers will boil down further. Now, an argument can be made that LLMs are not the same because agent's are not deterministic... but that is only kind of true. If we discount hardware differences LLMs are still "just" an operation on input data and given the same seed you will get the same output. The same analogy could be made for compilers; if you give them slightly different input flags you can get wildly different results! So maybe we should take the last few decades of experience with working with software and see what we should keep around. Maybe an approach where we actually version control our prompts and other model parameters would be advisable? The point of building software has never been to create the software itself, it has always been to create something that solves a problem (or that entertains in the case of games, which can be viewed as the problem of boredom). So we never really care about the actual machine code anymore, maybe it is time we stopped caring about the source code too?
-
Comment on What programming/technical projects have you been working on? in ~comp
archevel Link ParentSeems like a highly worthwhile goal! Any plans to deal with malicious actors? Speed limit might reduce the immediate impact, but it doesn't prevent bots from spreading if someone introduces "well...Seems like a highly worthwhile goal! Any plans to deal with malicious actors? Speed limit might reduce the immediate impact, but it doesn't prevent bots from spreading if someone introduces "well behaved bots" that ultimately invites bad ones...
-
Comment on Timasomo 2026: Week 2 Updates in ~creative.timasomo
archevel LinkMade some progress on the painting of three minis. Need to paint some areas with metallics and still need a lot of highlighting. Then some basing... I'll probably just work on the three in the...Made some progress on the painting of three minis. Need to paint some areas with metallics and still need a lot of highlighting. Then some basing... I'll probably just work on the three in the linked image to get a feel for how the rest should be done. Lot's of fun, but slow going!
-
Comment on Software developers are not okay in ~tech
archevel Link ParentPreferably to something that isn't next in line to be gobbled up by agents+harnesses. Hard also to find a field that has the same benefits (salary levels + other benefits). Something that allows...Preferably to something that isn't next in line to be gobbled up by agents+harnesses. Hard also to find a field that has the same benefits (salary levels + other benefits). Something that allows for focus at the same level is kind of rare I think. Let me know if you find it!
-
Comment on Software developers are not okay in ~tech
archevel LinkJupp. There is definitely skill atrophy going on. I also think he is hinting at the core reason why this shift in the industry is basically inevitable. It's the economics of it. Let's assume that...Jupp. There is definitely skill atrophy going on.
I also think he is hinting at the core reason why this shift in the industry is basically inevitable. It's the economics of it. Let's assume that today's models can't deliver software of the same quality as a skilled developer. That I think is fairly uncontroversial. The economics of getting software at 80% the quality level for a fraction of the cost just makes this the sane path economically. If it bugs out 20% of the time, you can likely afford to pay someone to put a bandaid on it for a fraction of the cost. The question is if the software is good enough, not if the software is good. In many cases the answer will be that it is good enough even if it is a bugridden mess that is impossible to maintain.
As for myself... I agree that there is a reckoning happening in the industry and I have felt a bit of an identity crisis looking at what these models are capable of. Well, I'll stick around for a while longer. I enjoy the people I work with and have fairly large degrees of freedom, but I'm not sure if I will stick around till the end of my career.
-
Comment on Why is there no word for please in the Nordic languages – are their speakers simply more direct, or is politeness expressed in ways outsiders may not immediately recognise in ~humanities.languages
archevel LinkI think the Danish word for this is "123456" IIRC.I think the Danish word for this is "123456" IIRC.
-
Comment on Timasomo 2026: Week 1 Updates in ~creative.timasomo
archevel (edited )LinkI got the kit for the Hearthkyn Salvagers end of last week and have been assembling them. I found some stls of orc heads, but they lacked necks so I imported them.into blender and started fiddling...I got the kit for the Hearthkyn Salvagers end of last week and have been assembling them. I found some stls of orc heads, but they lacked necks so I imported them.into blender and started fiddling around with the sculpt tool. I have no idea how to use them, but managed to extrude the back of the heads into neck-ish shapes. Printed them out and have glued some of the figures together. The heads aren't as detailed as the rest of the models, but they're good enough for me. This weekend I'm going to put on a base coat and start painting!
-
Comment on What are you reading these days? in ~books
archevel LinkJust finished Ashes of Man by Christopher Ruocchio. It's a space opera and this is the fourth book in the series. I liked this one better than the previous one which was a bit too dark and...Just finished Ashes of Man by Christopher Ruocchio. It's a space opera and this is the fourth book in the series. I liked this one better than the previous one which was a bit too dark and gruesome for my tastes. I'll likely pick up the next book in the series from the library, but this one ends kind of bleak for the protagonist...
I'm gonna read Ringworld by ??? niven? next in any case.
-
Comment on Timasomo 2026: Roll Call in ~creative.timasomo
archevel Link ParentThanks! I've now settled on starting out with a kill team of Hearthkyn Salvagers as the base. I haven't played 40k since late 90s so it is fun to get into the hobby again! Still have a based but...Thanks! I've now settled on starting out with a kill team of Hearthkyn Salvagers as the base. I haven't played 40k since late 90s so it is fun to get into the hobby again! Still have a based but unpainted metal dreadnought on a shelf so if I get bitten by the paint bug that's probably what I'll go for next (even if it's not a proper kin like mine definitely absolutely will be).
-
Comment on What programming/technical projects have you been working on? in ~comp
archevel LinkThe state of building secure software is rife with pitfalls. It's super easy to forget some flag or config, unintentionally leave some door open or not act quickly enough on the latest CVEs. You...The state of building secure software is rife with pitfalls. It's super easy to forget some flag or config, unintentionally leave some door open or not act quickly enough on the latest CVEs. You can even do most things right, but due to the tech stack you build on having vulnerabilities attackers can exfiltrate data in unexpected ways or gain a foothold to execute arbitrary commands in your infrastructure.
Over a few months I've been building a stack for secure software. It strictly limits what you can do and locks down the execution environment. At the core sits an orchestrator that listens for https connections. It sniffs the connection for the domain to pass the connection too and starts a VM in tens of milliseconds that's associated with that domain. The VM processes the request and sticks around for a few seconds (since requests are often "bursty") after which it gets stopped.
The app VM is very limited; no file system and no network stack (the request/response is mediated over vsock). Each app runs on boot without any other binaries installed so you'd have to pull off a hypervisor escape from within a unikernel in order to gain the a permanent foothold. The VM boots with certificates it uses for the https-connection so the orchestrator only ever sees the encrypted traffic. I ended up needing to make calls to external services so for that I mediated the calls through the orchestrator. The orchestrator checks an app specific allow-list to see if that ip/host is explicitly trusted before setting up a TCP connection. So the app is prevented to talk to anyone which you haven't explicitly allowed.
The above approach works great for stateless servers! But, most apps need a database. You could set this up via allowing connections to an external db, but I figured it would be nice with a similarly locked down approach to persiatence. So I ended up creating a small Linux application that starts on boot and launches postgres. This VM has access to an encrypted disk, but again no network. Instead it just runs on the same host as the VM app. The app-db communication is mediated over vsock (needed a small patch to postgres for this), which incidentally removes all network latency! I ended up setting up an encrypted export of the WAL and feed that asynchronously into a standby postgres on another node. That way data loss is unlikely, but I'm not sure if I should keep this part. The redundancy is nice, but it's quite a bit of code necessary to mediate the transfer and also a bunch to handle failovers etc... I might end up slimming the whole project down a bit.
-
Comment on Timasomo 2026: Roll Call in ~creative.timasomo
archevel LinkMy goal is to build and paint at least one miniature for a Warhammer 40k army and/or Kill team. My idea is to do an ork version of the Leagues of Votann called "Legumes of Waaaghtann". They will...My goal is to build and paint at least one miniature for a Warhammer 40k army and/or Kill team. My idea is to do an ork version of the Leagues of Votann called "Legumes of Waaaghtann". They will be highly sneaky and thus painted in pink. Lore-wise I imagine them to be a band of Orks that took over a fallen squat stronghold and managed to convince the semi-senile Votann core that they were legitimate. To start I want to take a box of votann and print some ork heads to swap out. I might go with a kill team box if I like the models better.
-
Comment on What programming/technical projects have you been working on? in ~comp
archevel Link ParentCool! I hope my move input didn't break it :) fun to see it act. Didn't dare to touch the "flip board"-button, bit it was tempting!Cool! I hope my move input didn't break it :) fun to see it act. Didn't dare to touch the "flip board"-button, bit it was tempting!
-
Comment on Announcing the eighth annual Tildes' Make Something Month (Timasomo)! in ~tildes
archevel LinkI will try to make something this time around... Have to pick which idea to pursue though. Choices!I will try to make something this time around... Have to pick which idea to pursue though. Choices!
-
Comment on What programming/technical projects have you been working on? in ~comp
archevel LinkMentioned in another post, but might be interesting for others to take a peek at https://archevel.github.io/laya/ It is a condensed model that runs in the browser similar to typesafe.ai's Jev model.Mentioned in another post, but might be interesting for others to take a peek at https://archevel.github.io/laya/
It is a condensed model that runs in the browser similar to typesafe.ai's Jev model.
-
Comment on Typesafe AI: System One Models and Jev in ~tech
archevel LinkSo just a few days ago Laya was released which is an open weight variant on the same concept IIUC. I had claude slap together a version that runs locally in the browser (even works on chrome on...So just a few days ago Laya was released which is an open weight variant on the same concept IIUC. I had claude slap together a version that runs locally in the browser (even works on chrome on Android for me). Anyway, I can see it could be useful to run something like this, but it feels like you'd be putting a magic 8-ball into your system. From my example questions, if I switched the order of Arnold and Conan the answer flipped! It is also not consistent cross devices, i.e. if I ran it on my laptop I got one result and another on my phone... Doesn't inspire confidence for me at least.
-
Comment on Independents/freelancers: how do you find new business? in ~finance
archevel LinkHad a chat with a sales person recently that work at my current company and who is doing a great job. I mentioned that I can't do sales; I don't like calling people at all and I don't think I'd...Had a chat with a sales person recently that work at my current company and who is doing a great job. I mentioned that I can't do sales; I don't like calling people at all and I don't think I'd enjoy it. Anyway, she told me her "process" was to call 5 people she knows and then call 5 people she don't. So she is basically cold 5 calling potential leads (after a bit of research) each day. To me that sounds exhausting. I mean even of I'd just call 5 friends that I know and like I'd be exhausted. But, I guess that doing this consistently teaches you what works and what doesn't after a while. I also guess that the pace she has isn't strictly necessary, maybe it would be enough to do 5 and 5 over the course of a week.
Anyway, I am clearly not a sales person so take it with a big pinch of salt!
-
Comment on Where do we go from here? (Regarding AI) in ~tech
archevel Link ParentFair enough. I don't think they are just sitting tight though. I think they are genuinely worried about AI and, like most people, not sure exactly what the right move is. That doesn't mean we...Fair enough. I don't think they are just sitting tight though. I think they are genuinely worried about AI and, like most people, not sure exactly what the right move is. That doesn't mean we should do nothing, just that it takes some time for parties etc to formulate their plans, agree on those internally, star to communicate those goals externally. Here where I live a party is divided into regional districts that are further divided into local districts. A discussion of any topic takes time to percolate through that machinery. That is partly by design, people need to get to wrestle with ideas a bit. Committing to an agenda shouldn't be up to individual politicians (though individuals can certainly champion some ideas), it ought to be done as s mandate of the party's members. Getting that mandate tend to be slow by the very nature of the organizations. This might be completely different elsewhere of course.
Still puttering along. Now the entire kill team minis have some paint on them. I ended up doing the basic painting on all of them instead of focusing exclusively on a few. Still need to sort out the bases, use some washes, maybe some light from some weapons etc and possibly edge highlighting. There's really not a definite "now they are done" point I'm starting to realize!