That just reminded ne of something from here I had completely forgotten about. Someone had posted making sort of dubious claim of guitar power chords being Turing complete. I made some small improvements (as I saw it, chiefly around how they had implemented their clock) to the code they posted
https://news.ycombinator.com/item?id=42294766
which lead me down a rabbit hole of guitar tab as opcodes, which eventually lead to something useful after throwing away the silly guitar virtual machine, since I had now done the work of implementing far more than the simple tablature parsing code I intended. That project that resulted from the silly toying around with tab notation as machine code lead to a midi-to-tab generator, which lead to audio to midi to tab, and a bunch more.
That project, for anyone interested in such things, is today https://github.com/scottvr/gtrsnipe
There are useful demonstrations in the repo's wiki.
I was surprised by the timing of this post, after recently shipping a similar tool quite similarly named (vizbin as opposed to binviz). A little more surprised to see that so many folks have also developed similar tooling.
I don't know the name of the tool that did entropy visualization you're remembering, but this was my cue to mention that vizbin does that using the colorizer of your choosing.
Another comment mentions using such a tool on source code; the binary strings example shows just that, with the source embedded in a tarball. Text is rendered as an 8x8 glyph font so you can immediately eyeball both.
Rendering known pure text as 8x8 bitmap fonts via the tool might be a bit ludicrous, but you can force a different byteclass and see only the colors and not the actual ASCII characters, which I think might be what they are getting atwanting to do. https://github.com/scottvr/vizbin/tree/main#profile-structur...
Another useful things falls out from combining these concepts, which is that you could run `vizbin` on an entire repo. The example from the README (which assumes binaries) is given:
# which files stand out? cluster by their region composition
vizbin profile corpus/\*.bin --json | \
jq -r '[(.regions|map(.kind)|unique|join("+")), .source] | @tsv'
The github repo has been linked a handful of times above, but you can also just install it from PyPi with `pip install vizbin`
apologies I answered the three of you all in this one reply. Hope it's of enough interest and easy enough to discern if something is not what you asked.*
OK, really? Y'all dig it; that's really cool. It is my first foray into creating something rap-adjacent (if you dont count the AI version mtlsnk discovered, and that was two years ago. I wrote the lyrics longer ago than that, slowly adding to them as new rhymes would come into my head. I really wanted it to exist but I don't have a typical voice you'd associate with rap (as you can hear even in this track's vocals.. Even with all the digitala help making it sound like I know what Im doing, the tone and tibre of my voice remain very much not a rap voice.) Oh and thanks for calling out a line that had you rolling; that's one o f my favorites. The bit about Hans Reiser might be a little risky, but very few people are going to have all the pieces to know what that;s (morbidly) funny as a diss.
I have basely zero natural flow really, and I was too self-conscious for many many takes to have anything worth keeping. That "incredible flow" you heard, zahlman, is the result of tedious manual slicing and moving and chopping of syllables in an old copy of Audobe Audition and to a lesser extent Audacity. And VocAlign is your friend. (I've seriously come to doubt that any recordewd rapper is as tight and precise as it sounds on the album and suspcet VocALign or something like it is the Auto-Tune (well, the other auto-tune, since they do use auto-tune too) of Hip Hop production.) I do most instrumental sequencing in an even older copy of Sony ACID Pro, but I'll have to put a little time and effort into listing the VSTs used if you're interested in that.
I'm actually working on documenting more of that production stuff that because the handful of people in the real world who have heard it aren't really familliar with a lot of the terms, but have told me that can tell that someone is geetting insulted and likely outclassed and they just really liked the track regardless. Amazing.
I'm curious, was I right that I could find enough folks who understand at least the surface "battle rap" style punch lines, or are you guys just saying you like the song? This is very, very different than anything I've made before.
Yeah I'm stil working on the subtitles. Subtitles are a solved problem, right? Well, not when you want to display something like /etc/passwd (and many such things) it becomes a pain.. The funny thing is that the main purpose in making a video was to maks e Lyric videom but that ended up the last thing I did insofar as editing the video, and it doesn't work right. I was tired from being up for two days trying to finish it.
I'm finishing up the lyric aligner I made for this video (which is why it's missing a section and starts to lag. But I was soo tired of editing and I wanted people to hear it so I took a break and upploaded it anyway I digress.) It let's you have a spoken lyric sheet for the aligner (whisper and stable_ts) to listen for, where you can put words in their phonetic spellings like you would for TTS, but the another lyrid sheet for presentation spellings that will be used in the subtitles. This was necessary due to all of the technical terms and acronyms that may not be pronounced in an expected way by the aligner tools. It zips (interleaves) the two together and creates a yaml file for hand editing tweaks suchas subtitle effects, then it generates the proper .ass file for that. I'll push the lyric aligner up to the repo when it's working 100%.
Thank you for the kind words. As I said, this is very different from what I've done before. A buddy and I put out a post-punk/new wave album in 2007 under the nema The Detached. It's free on the same site. Mostly we were doing pastiche of TubeWay Army, Devo, etc. You can find all of those tracks at
https://killsignal.net/thedetached/
There's a couple of tracks the other half of the detached did not want to do, so they're solo efforts, like https://killsignal.net/thedetached/3.scottvr-every_day_I%20d...
Im on vocals, keys and occasional guitar. Dave Griggs played bass and guitar. Drums were all on my laptop.
Oh and I guess I released a video for an unfinished (sensing a theme?) cover of Black Sabbath's "Changes" you might enjoy, if you're info horror and such. https://www.youtube.com/watch?v=6btgKG0tWxg This is a different side of my limited vocal range and stylings. :-)
Rap is new to me; to get it to sound good takes surgery in a DAW slicing, moving, shifting, quantizing, etc. Even my Voice Live Rack can't make me sound like a rapper. Which is why the song maybe sounds more in the direction of 90's industrical club music (Front 242, Ogre from SkinnyPuppy, a lot of those guys were essentially "spoken word to rhythm, which one might call "rapping". Even if "El User" succeeds as "rap", I think we can all agree it is not so much "hip hop".
I'm so grateful you guys listened to it and enjoyed it. I just wanted a vehicle for the lyrics (which might be deeper or more layered than the nerd humor allows people to see. I don't know, which is why I wanted out in the world. I will get back to work on the multiple aligner tool and get the video perfect soon, and I'll let y'all know.. Oh also discoveref that the very last line, "#!bash" accidentlaly was elided from the final mix, so Ill have to tack that back on and make it not oend so abruptly.
Again thanks a lot for the feedback.
--Scott "UID Zer0" (I don't realy use that name; I just made up the DJ Defrag and UI Zero duo because I don't of course have a "rap name", but needed to be able to use some rap cliches like "first name x, last name y"
After finding the AI version I assumed you had generated this as well. It's been a hot minute since I've done anything creatively with audio/music, but would have been interesting to work with AI generated or AI-assisted vocals like that. Off-topic-ish: I have the beginnings of a modular/eurorack synth boxed up after a move, still. Also very much a noob with modular, but I've been playing with the idea to hook up the cv to real-time network monitoring to give my network a voice. It would take some time/tuning to get somewhere musically, if it would evolve from pure noise at all, but I'll take a packet powered pseudo-RNG noise box. Putting the ARP in arpeggio.
Having only heard the final edit of course, but I'd say your voice works well for rap/nerdcore. As with all things: if you put in the effort/time you'll improve. You may find yourself not having to correct the flow as much. And if you're not looking to perform live, why would you care when you get the flow/outcome you like through slicing/editing?
The lyrics to me seem like you put too much thought into referencing tools/tech in every line, which makes me fail to stay with the story/subtext. The deeper meaning is lost on me. That may mean I'm not the intended audience, and therefore be an N=1 answer to the question you were floating here on HN.
Tool/tech ref to word/line ratio is lower for artists like Ohm-I, Dual Core, ytcracker, making it easier to follow the "plot". More palatable with a hook as well, but if that makes it too clubby/hip hop/not you then don't, obviously.
Well, I'd disagree with the "too much thought" suggestion. Too many words? Oh probably so, but not much thought. Wait, that sounds dismissive or like faux-humility, but that isn't how I meant it.
Working as I have for over three decades continually in IT, spanning industries, skills, tools, languages, responsibilities, seeing the coming and going of many "next" generations, etc.. I suppose the pool of words to choose from and the range of things to use as metaphors is just pretty large and much deprecated, outside of the contrived use case of tech nerd battle rap.
Nerd flexing boast/diss tracks until now, I didn't know was already a thing. I kinda thought I was innovating. Lol, maybe I was from your description of it being too loaded with technical jargon, but the jargon, the word play, was the point. I was having fun with it. Some are shallow puns, others, like the ReiserFS lines were amusing to me because of the way they just wrote themselves. Hmm. There may be a distinction between real connections and patterns and apophenia, and there may be a way that it is missing in my head or blurred, in which case it may also be the case that the symbolism I find obvious and amusing just seems inane or pointless to everyone else. That would explain a lot of things now that I think about it. Heh.
I knew "nerdy white guy raps" was a thing from a show called "Dave" (I think) a year or so ago with a (I think) real rapper called Lil Dicky. (Again, "I think".) The show was good but I wasn't connected enough to modern pop culture to really know how much was fictionalized or otherwise. (I did recently see that Selena Gomez got married and was a little too excited, announcing to nobody "that's... that's! From the show!" I just looked it up and his name is Benny Blanco so that part of the show was at least accurate about him existing in real life and being a producer or somehow connected to hip hop. That said, your message is the second time in a handful of days someone has mentioned ytcracker but the first time had less context, and the context I did have (it was related to a video) caused me to assume it was a mangled or auto-corrected term and that ytcracker referred to some tool for downloading videos (like yt-dl/yt-dlp) and so I didn't look further. I guess this is signal that maybe I should.
I don't actually plan on making another track of this sort (though it has occurred to me a few times even though it wasn't on the table prior to putting it out there; I've never had so many people tell me they've had it on repeat - or maybe I've never had anyone ever tell me that about any music I've put out until this one, for which the response has been way more positive than expected.
On the audio output from network traffic (oh congrats on the developing eurorack habit btw) I did something enough in that zone that it causes me to want to recommend a rabbit hole to you (if it is not one you've already been down.) That rabbit hole is Sonification. My entrance was specifically parameter-mapped sonification and a search will find you many interesting papers. I might save you some time, or possibly send you straight into an ocean of time-suckage by suggesting you start with "The Sonification Handbook". On the plus side it will reference and was edited by one of the authors of one of the main papers I was about to have to look up until. i remembered the larger work so there's two birds to get started, on the down-side it references loads of stuff that are rabbit holes of specialization that could prolong your foray underground. I hope that is useful to you.
> apologies I answered the three of you all in this one reply. Hope it's of enough interest and easy enough to discern if something is not what you asked.
As a heads-up, you're lucky this time — it seems to be pretty rare on HN that people keep track of threads past their initial encounter, unless they get a direct reply (and even then it's far from guaranteed). It's not a great place to have a conversation (but then, where is?).
> Yeah I'm stil working on the subtitles. Subtitles are a solved problem, right? Well, not when you want to display something like /etc/passwd (and many such things) it becomes a pain.. The funny thing is that the main purpose in making a video was to maks e Lyric videom but that ended up the last thing I did insofar as editing the video, and it doesn't work right. I was tired from being up for two days trying to finish it.
> I'm finishing up the lyric aligner I made for this video (which is why it's missing a section and starts to lag. But I was soo tired of editing and I wanted people to hear it so I took a break and upploaded it anyway I digress.) It let's you have a spoken lyric sheet for the aligner (whisper and stable_ts) to listen for, where you can put words in their phonetic spellings like you would for TTS, but the another lyrid sheet for presentation spellings that will be used in the subtitles. This was necessary due to all of the technical terms and acronyms that may not be pronounced in an expected way by the aligner tools. It zips (interleaves) the two together and creates a yaml file for hand editing tweaks suchas subtitle effects, then it generates the proper .ass file for that. I'll push the lyric aligner up to the repo when it's working 100%.
I probably would have just fixed everything manually instead of trying to improve the system...
Yes, but maybe you know how it is... similar to that threshold where "I've done this three times before, if I had automated it, I'd maybe have spent the same amount of time writing the automation, but the fourth time on would be free" and doing the sort of quick on-the-fly gut-math cost/benefit analysis that either results in an eventual time savings to be celebrated, or a "dammit, I should have just done this by hand again" unique type of regret.
So for this particular set of words and timing, I found I really disliked eyeballing the lines and doing quick time arithmetic and sanity checks because there were no obvious errors the two or three passes I made. So I found myself making a quick script to return the differences, to try to see what odd combination might be triggering a bug, etc. plus although what you see now looks like just standard subtitles, the original burn of them used some of the ass format's unique rendering capabilities and, as with the one-off script type automation cost/benefit.. I think it's actually cool enough and potentially useful for other cases (generative TTS/ai vocals, etc) that even if I never make another "rap"-style track and video, someone else is likely to encounter the same problem and find this solution useful.
But probably mostly it's a combination of sunk cost fallacy and stubbornness, with some laziness and fatigue thrown on top that has now a video for a song I wanted to share with whoever it might reach with broken subtitles and an unfinished tool to correct them.
Oh, I guess I should have linked the reason I mention the limited target audience that will get the "punchlines", especially for those who don't feel like watching a video or listening to a 50+ year old white nerd's attempt at producing a "rap" diss track. Here's the lyrics I've been sitting on for a couoke of years trying to make this.
Hey! Neat to see some more folks that have a story to tell about the oerhaps unintended side effects of the way the dot was implemented in gmail addresses.
Reading these dot-related comments reminded me that I once made one here about how I was able to leverage this possibly asynchronous dot-handling at gmail. Because the story is a bittersweet sentimental one for me I had to go back and read it now, and though I am sitting here with bad allergies streaming down my face, it also felt really nice to remember that.
I know this is a bit of a sideways comment, and I apologize for that but it isn't often I am overcome with emotion on HN, or on the topic of email addressing, and well, I appreciated the opportunity to re-read it now. Thanks.
You might find my gmail dots story interesting too. It's a little long, but the dot-character portion begins about a third of the way down in my comment:
I am honored to have my recent paper, "The Grand Unified Model of DevOps/SRE Dynamics" (at times referred to simply as "GUM"; the direct link to my paper only is at https://github.com/scottvr/GUM_of_Devops/blob/main/The_Grand...), appear in the proceedings of SIGBOVIK 2026. (Finally, the tull proceedings are released! This is not a dupliate of older submitions, best I could tell, and in any case the focus is on my paper, now tha the full proceedings are out there. https://sigbovik.org/2026/)
The venue and publication are a good fit for the paper and serve as useful signals for the temperament of the paper and the treatment throughout the development of the model. It also says something about the reviewers acuity and elite selection criteria, which are to be celebrated for what they are.
As my paper's abstract makes clear, the model is not offered as a predictive instrument in the strict scientific sense. It is instead a formalized account of a familiar practitioner truth: software delivery is not shaped only by pipelines, tooling, deployment frequency, or architectural complexity; it is also shaped by technical debt, morale, urgency campaigns, competence mismatch, and executive volatility.
The ethos of GUM does not stem from a belief that DevOps metrics are useless. Rather, they are useful enough to make omissions conspicuous. If we can assign symbols to deployment frequency and change failure rate, we may eventually have to admit that organizations themselves also perturb the system. Recent literature has done much of the work of formalizing the example proxies given in GUM 1.0, which allows us to construct a new model that may satisfy the critics who claimed GUM 1.0 required "measuring the immeasurable."
While researching for GUM 2.0, we were surprised by how rapidly the recent literature appears to be moving into territory adjacent to that of the GUM. One paper formalizes delivery speed as a function of automation and CI/CD maturity; another models developer-experience variables such as cognitive load and technical frustration as causal contributors to release-cycle duration, which looks quite a lot like the GUM term M (Developer Morale Multiplier). A third attempts to quantify technical debt as a compound-interest problem with remediation ROI. It is, of course, an honor to see how much impact the GUM has had, even if it has not yet been cited in any papers. A more thorough survey of these papers from recent literature can be found at the GUM's primary site at https://github.com/scottvr/GUM_of_Devops/blob/main/sigbovik2...
We are currently working to address these developments in GUM v2.0. As stated in the original "Grand Unified Model of DevOps", when the real world begins to collide with a model, *it is time to introduce more formalism.*
I can't think of a single case of any AI content, be it prose or code, where I thought "I wish I had written that". With AI code, it's more like I wish I hadn't let the AI write that.
We’re using Copilot at work to build reporting and automation tools. Nothing ground breaking, but very useful and tailored to our needs.
Frankly without AI assistance many of these tools just wouldn’t exist at all. We can build stuff in 6 weeks part time as a side project that would have taken at least 3 months full time, and therefore would not have been feasible. Then we can iterate on it at least 2-4 times faster than with hand coding.
So I’d love to have an extra few developers to just work on that stuff full time, but I don’t.
Whether that means our organisation spend on AI overall is a positive, I really can’t say. Quite possibly not, but my team are getting real benefits.
I’m building reporting for my company and what you said mirrors my experience nearly 100%.
I’m a backend developer so I know what it takes to build a half decent reporting system. Writing all those queries, slice and dice charts and what not takes real time and effort. All that has been outsourced to Claude Code. I now focus on ensuring that the system is sound architecturally and that useful reports are being surfaced.
An engineer doesn't care about how fast something is made (at least, not as a primary metric engineering). A salesman cares about how fast they can push to market.
It's clear HN is a bastion of salesmen who happen to have "engineer" in their work title. But the mentality towards actual engineering makes it clear they are primarily salesmen.
> An engineer doesn't care about how fast something is made
That is absurd, these are tools only my own team use. Why would I not care whether I had them in a month or two, or fur many of these tools quite possibly never because we don’t have the spare capacity for how long it would take without AI?
>Why would I not care whether I had them in a month or two,
Because you're thinking like a salesman. What difference does a month make for a supportive tool without financial incentive? Why can't you justify a month of development without the idea of corporate breathing down you neck?
What's interesting to me is that while it was obvious to all of us who came to think in the Unix Way, that insofar as composability, usage discoverability, and gobs of documentation in posts and man pages that are hugely represented in training corpora for LLMs, that the CLI is a great fit for LLM tool use, it seems only a recent trend to acknowledge this (and also the next hype wave, perhaps.)
Also interesting that while the big vendors are following this trend and are now trying to take a lead in it, they still suggest things like "but use a JSON schema" (the linked article does a bit of the same - acknowledging that incremental learning via `--help` is useful AND can be token-conserving (exception being that if they already "know" the correct pattern, they wouldn't need to use tokens to learn it, so there is a potential trade-off), they are also suggesting that LLMs would prefer to receive argument knowledge in json rather than in plain language, even though the entire point of an LLM is for understand and create plain language. Seemed dubious to me, and a part of me wondered if that advice may be nonsense motivated by desire to sell more token use. I'm only partially kidding and I'm still dubious of the efficacy.
* Here's a TL;DR for anyone who wants to skip the rest of this long message: I ran an LLM CLI eval in the form of a constructed CTF. Results and methodology are in the two links in the section linked:
https://github.com/scottvr/jelp?tab=readme-ov-file#what-else
Anyhow... I had been experimenting with the idea of having --help output json when used by a machine, and came up with a simple module that exposes `--help` content as json, simply by adding a `--jelp` argument to any tool that already uses argparse.
In the process, I started testing, to see if all this extra machine-readable content actually improved performance, what it did to token use, etc. While I was building out test, trying to settle on legitimate and fair ways to come to valid conclusions, I learned of the OpenCLI schema draft, so I altered my `jelp` output to fit that schema, and set about documenting the things I found lacking from the schema draft, meanwhile settling to include these arg-related items as metadata in the output.
I'll get to the point. I just finished cleaning the output up enough to put it in a public repo, because my intent is to share my findings with the OpemCLI folks, in hopes that they'll consider the gaps in their schema compared to what's commonly in use, but at the same time, what came as a secondary thought in service of this little tool I called "jelp", is a benchmarking harness (and the first publishable results from it), the to me, are quite interesting and I would be happy if others found it to be and added to the existing test results with additional runs, models, or ideas for the harness, or criticism about the validity of the method, etc.
The evaluation harness uses constructed CLI fixtures arranged as little CLI CTF's, where the LLMs demonstrate their ability to use an unknown CLI be capturing a "flag" that they'll need to discover by using the usage help, and a trail of learned arguments.
My findings at first confirmed my intuitions, which was disappointing but unsurprising. When testing with GPT-4.1-mini, no manner of forcing them to receive info about the CLI via json was more effective than just letting them use the human-friendly plain English output of --help, and in all cases the JSON versions burned more tokens. I was able to elicit better performance by some measurements from 5.1-mini, but again the tradeoff was higher token burn.
I'll link straight to the part of the README that shows one table of results, and contains links to the LLM CLI CTF part of the repo, as well as the generated report after the phase-1 runs; all the code to reproduce or run your own variation is there (as well as the code for the jelp module, if there is any interest, but it's the CLI CTF eval that I expect is more interesting to most.)
That project, for anyone interested in such things, is today https://github.com/scottvr/gtrsnipe There are useful demonstrations in the repo's wiki.
reply