Dan Cross

For some time, I have been discomfited by operating systems putting requirements to run specific userspace code onto arbitrary programs.

Sure, some amount of startup code usually exists; even C has to cons up the arguments to `main`, after all. And of course, to execute a system call or access some OS-provided object, allocate memory, and so on, requires some kind of contractual interface with the OS.

But that's not what I'm referring to. It's things like libraries colluding to implement a non-trivial amount of functionality of some system call behind the caller's back, or requirements to use a specific userspace library, calling convention, memory allocator, etc.

Why? Because this limits flexibility for user software, by conflating mechanism and policy. The universe of interesting and useful user programs is vast; by imposing requirements to (say) make system calls through a particular library interface, I'm imposing a particular "shape" onto the design space for the programs that use my system, and thus, pruning away some subset of programs from working on my system.

Consider programming languages: in the context of function calls, they do not all use the same conventions for argument passing and returning values, with respect to stack and register usage. To mandate that proper use of the system must use some specific calling convention (as opposed to system call invocation convention, which is a clear interface boundary) puts a burden on the language implementers and constrains their design space. It shouldn't be necessary to obey all of the rules of the conventions of `libc` to successfully use a system in a program not written in C.

Our cat has started coming into the bedroom every morning at 5am, jumping onto the windowsill, and attempting to communicate with the birds that are outside chirping their morning song. She makes a trilling sound that, I suspect, is meant to mimic the birds' calls and lure them in...so that she can catch and eat them.

Fortunately, no birds were harmed in the making of this cuteness.

Is the C standard a formal document? I suggest that no, it is not. Of course, it's written in a formal register, and it certainly strives for precision, but at the end of the day it is prose and in some places ambiguous. That's not a dig, it's just an inherent limitations of using English to define a programming language.

I rather hope that doesn't happen to Rust. What I'd _much_ rather have happen there is to see the language precisely defined by an operational semantics, that is the source of truth. That could be paired with an explanatory reference document that describes those semantics in prose, as accurately as possible. But at least where ambiguity creeps in (as it surely will in prose) there's an accurate and rigorously defined source that one could consult to address those ambiguities precisely.

@jascha this hits so hard. :-(

"Have I told you about my new idea for an AI-enhanced gardening spade? It measures temperature, moisture level, and nitrogen content, all as you dig! Think Theranos, but for potting soil."

What is it good for?
https://youtu.be/ztZI2aLQ9Sw?si=8xaTAbT__L2Z0Q58

@ricci@discuss.systems agreed!

@robpike@hachyderm.io though I use a Mac as my main workstation, I don't bother with Apple keyboards. Craig Silverstein got me into Kinesis keyboards, which were great for RSI and are significantly higher quality than anything made by Apple in many years (if not ever).

@dan ok, I totally want to know what the bit is.

@chuksawa@discuss.systems yeah, 25th amendment, definitely. This insanity is about to spiral out of control.

@chuksawa@discuss.systems it seems weird to say this, but I wonder if we are a lot closer to a military coup than a lot of people think possible.

@heidilifeldman@mastodon.social much as I appreciate this reporting, I want to point out that these BORTAC and BORSTAR people are _also civilians_. They may be engaging in some kind of quasi-military cosplay, but they're not recognized combatants in an armed conflict, nor members of some military service. They are, therefore, civilians. This is not your fault, of course, but I really wish that journalists would stop referring to general members of the public as "civilians" when writing about these things.

Shout out to @mattblaze@federate.social for aggressively blocking cranks. I appreciate that I don't see their demands to prove a negative on my timeline (or see those only fleetingly), thus keeping my blood pressure at some barely-tolerable baseline.

God speed, Matt. You put up with a lot.

@krismicinski@types.pl @shriramk@mastodon.social @gwozniak@discuss.systems @steve@discuss.systems @jfdm@discuss.systems @csgordon@discuss.systems @lindsey@recurse.social @jeremysiek@types.pl as I've been playing around with these things the last few weeks, doing so as a practitioner who is keenly aware that the a) the world is changing, b) I did not ask it to change, c) it's changing in a way that feels bewildering given my professional experience thus far, I have a few thoughts about the externalities.

First, I think there is a lot of emotion surrounding these things. Beyond the obvious "change can be hard" thing, lots of people have invested significant time and energy over the course of many years of their (innately finite) lives building up skills that are quickly becoming irrelevant. Many people have forged their entire identities and sense of self around being programmers, or software engineers, or whatever the kids are calling it these days. For probably the first time in their lives, they're looking at automation coming for them in a serious way, and it's scary and challenging their sense of self. The fear of, "if I'm not this, what am I?" is real.

I get that. I have deep empathy for that. We _all_ should. Honestly, I'm half there myself.

There's also the sense that it's being driven by a set of out-of-touch members of the billionaire class who are pushing this relentlessly, literally flying above us in their private jets, with no thought to those of us down at the bottom who are going to get crushed under the weight of this juggernaut. We are building the pyramids ever higher for the glory of great pharaoh. What are a few slaves crushed under giant blocks of stone along the way?

Beyond that, if this becomes a de facto necessary part of the production of software, then it seems to me that the means of production of software is going to be controlled by the very few players that can afford to field this technology. And I truly believe the actual cost is much higher than what we're being charged now; they're burning VC money to make it cheap. But happens when that runs out? I keep hearing people talk about "we'll train our own models and run the inference engines locally!" Ok, good luck with that: that's years away. Meanwhile, Google is moving to build small nuclear reactors next to data centers, and that tells me that we're not going to see this outside of the big players any time soon.

Further, energy demands (and water! people always forget the water!) are _increasing_ as these machines get better, not decreasing. Techniques like MoE may lead to them increasing at a slower (non-exponential) rate, but it's still superlinear; eyeballing it, it looks quadratic-ish. Contrast with, say, computer themselves: compare the power draw of ENIAC versus a cell phone; the cell phone is many orders of magnitudes more capable while consuming orders of magnitude _less_ power. Unless we can figure out how to make power consumption sublinear as models increase capability, I don't see how this is at all sustainable in the long term.

Then what?

@shriramk @jschuster @lindsey @krismicinski@types.pl @jfdm @csgordon @jeremysiek it's not meant to be exact. The point being, one develops a very thorough spec before ever attempting to generate a line of code, and spec writing becomes a major focus of the process, with a spec as one of the primary artifacts produced. I never tried Spiral (for that matter, I was never _really_ subjected to Waterfall in its full glory, either), but if you feel that fits better, good to go. In a chat with some colleagues yesterday, most of them were saying that, when using LLMs, they're spending most of their time writing very precise specifications in the form of "prompts." I find that interesting, and very different from from the Agile world of "yolo just write some code, amirite, bruh? lmao."

Definitely feeling some FutureShock at the moment.
https://www.youtube.com/watch?v=JUpidCc7wwY

@shriramk @jschuster @lindsey @krismicinski@types.pl @jfdm @csgordon @jeremysiek yeah. Bryan Cantrill and I were discussing this last week. A trend I've observed is that the LLM does much better when given very precise instructions, sometimes totaling thousands of lines of written text. It reminds me of waterfall development, which never really worked. But why didn't waterfall work? Because the distance between phases was too large; by the time you were in development, it was too late to go back to requirements and iterate through design again. The LLMs sort of short-circuits that process, allowing rapid feedback; does this mean a return to waterfall development styles?

@shriramk @jschuster @lindsey @krismicinski@types.pl @jfdm @csgordon @jeremysiek why is it a foolish position, though? (Serious question: I'm not trying to be either obtuse or confrontational). Most people don't look at the object code in an executable binary, and even most programmers have no working knowledge of an assembly language or machine's instruction set at this point. This is something I was just thinking about while walking over to the corner store. Consider my ham radio example: I consider that a pretty impressive accomplishment, despite all the deficiencies I mentioned earlier. And while I don't think it's possible right now, it is _conceivable_ that at some point in the future, I may be able pop up one of these things and say, "rewrite the Linux kernel in Rust" and it would make a passable go at doing so. But then it struck me, what would be the point? If the premise is that the machine is going to be able to write better code than I can, perhaps not now, but at some point in the future, then what does it matter what notation it uses to do so? The value-adds of Rust, the borrow checker and so on, are there to help human programmers. But if the machine gets good enough to work around our inability to reason about thorny concurrency issues, memory, and type safety, UB, etc, in C, and it does a better job than me _anyway_, then there's no reason to convert it to a language that's optimized for humans. Maybe that's just a big matrix of numbers; meaningful to it, but not to most of the rest of us, just as object code is not meaningful to most of us. Surely the best course of action is to settle on something that's optimized _for the LLM_. Almost all extant programming languages seem to be optimized for the _programmer_ (not you, APL). Perhaps this is similar to what you were referring to when you mentioned the source and intermediate languages?

@krismicinski@types.pl @shriramk@mastodon.social @jfdm@discuss.systems @csgordon@discuss.systems @lindsey@recurse.social @jeremysiek@types.pl "but it can give you a great hollandaise recipe!"

@lindsey @krismicinski@types.pl @shriramk @jfdm @csgordon @jeremysiek I will sit on the top of the roof of this house as the water rises waiting for you all to come along in the lifeboat and rescue me. Please hurry!

@krismicinski@types.pl @shriramk@mastodon.social @jfdm@discuss.systems @csgordon@discuss.systems @lindsey@recurse.social @jeremysiek@types.pl really?! That's depressing.

I find these things _maddening_ to use. It feels like trying to neatly typeset your ideas by by dragging wet toilet paper dipped in ink across a piece of sandpaper.

Are they capable of some impressive things? Yes. Do I think they're a good tool as an augment for a sophisticated user to go faster? Honestly, not really. The NLP aspect is neat; the multiple round-trips through English to back to English are excruciatingly slow, expensive, and inefficient. It's not a good use of my, or frankly the machine's, time, let alone electrical power or water.

Case in point: some colleagues the other day were saying something like, "I just can't get it to use `jq` instead of writing little Python scripts to process JSON....Here's what I put in my CLAUDE.md file: ." I couldn't help but feel like this is exactly the sort of thing where you want the concise precision of a small DSL for assigning weights to tools (and providing templates for those tools' use) to drive how the agent uses them. But you can't do that, because the agent only trades in text.

Like I said, there's clearly a "there" there. But setting aside the moral and ethical issues for a moment, that doesn't mean that the present model of interaction is _good_, let alone that it can't be substantially _better_.

Someone recently said that, "the cloud is just a landlord for your data." If that's the case, then LLMs are like the company store that is leasing you the tools that you use to do your work.

@hiway oh interesting. I assume you've seen this? https://youtu.be/7tG7LZgOb-U?si=B8mnkZUWo7tQQ7x7

@preinheimer@phpc.social yeah. I've played around with these things to see what the hype is about and a few things stick out to me. First, it's obvious they're giving away the product to get people hooked and paying for it with VC money. But even so, a CC Max plan is almost required to get something useful and it's already too stupidly expensive. Are people going to pay for these when it's 10x the current cost? At $2k/mo per seat the calculus changes. Second, these tools just aren't very good. Full stop. They generate mediocre results. Full stop. Seriously, people need to internalize this: the output is not good. That people think that it is kind of amazes me, and also makes me think that most output from humans isn't very good, either. So we're not getting some great leap forward in quality; we're just getting something around or perhaps slightly better than the median, which is already bad. Third, I don't think they actually save all that much time. Yeah, it's kind of nifty to toss the tedious and boring parts at a machine, but they require so much hand-holding to get something merely acceptable that it just feels like shifting the burden from source generation to using imprecise human languages to make a machine do the text generation. I have seen some colleagues do cool things with them, but at a huge cost in terms of effort. If the tools require that much effort, they're not good. For the first time in my professional career, I feel like someone is trying to sell my labor back to me instead of paying me for it. Is there some element of these things that's going to stick around? Sure. But not in their current form, and the hype...oh goodness, it feels like the 1990s all over again.

@snippy@pdx.social @cstross@wandering.shop @graydon@canada.masto.host @aiusepsi@mastodon.london @ysbreker@idlethumbs.social this fails to take into account the context of the era in which these technologies emerged. Bad design? Yes. But it's unclear if it could have been any other way. TCP/IP was the underdog in the early 90s; telcos and the big computer manufacturers were betting on ISO/OSI and their own proprietary goo (DECnet, SNA, etc). The Internet was viewed as a prototype, not a production system, and the design of many of the protocols reflected that. SMTP and authentication? Who needed that when the only people who got access to the network were more or less trusted entities at research organizations? Sure, some precocious undergrads might screw around with forging email, but what's the harm in that? It's a lot less destructive than many campus pranks. Yes yes, I know some folks who were there will vehemently disagree, but I was there too and that was my experience ca 1994. Then Berners-Lee sort of stumbled on the web and it became the Internet's "killer app". The Clinton administration moved the Internet from a research project to something that could be commercialized. OSI was years delayed, complex designed by committee, the tech people hated it (it was designed by the telecom people, not the computer people), TCP was pretty good for most use cases and IP was good enough, and the Internet was already here, now. I remember looking at the web for the first time and thinking, "wow, this is garbage; who would ever use this?" HTTP was clearly an inferior protocol with excessively high overhead, HTML was overly verbose yet anemic to an extreme for representing complicated documents, and the thing wasn't interactive at all unless you were using a NeXT machine; using early web "apps" was like using a glorified 3270 terminal with pictures. But people loved it because it was graphical and had pictures, and in many ways it democratized access to the 'net: you didn't _have_ to internalize the esoterica of the abstruse command-line interfaces of the systems of yore, let alone how to use TELNET or anonymous FTP; that was all hidden away under a point-and-click veneer. So of course it's all built on a teetering house of cards, but when viewed across the evolutionary arc of the whole thing, each little step along the way seemed to make sense locally.

@SteveBellovin @darryl_ramm @mattblaze I used to live 5 blocks from the Macombs Dam bridge; there's no dam there. I think it is most famous because the 155th St overpass was in an Eric B and Rakim video. https://www.youtube.com/watch?v=2TN-kDEKxF0 (The field to the right is now a parking lot. I used to pay to park by Beetle there; every time I went to pick it up the dudes who worked there would be sitting on it smoking weed.)

@SteveBellovin @darryl_ramm @mattblaze the Triborough bridge renaming wasn't that bad, but the one I never got used to was the westside highway. The Joe DiMaggio highway just doesn't have the same ring to it. By the way, it looks like Trump just went ahead and stuck his name on the side of the Kennedy center anyway, legalities be damned: https://www.nbcnews.com/politics/trump-administration/trump-kennedy-center-name-change-congress-rcna250081 What a turd.

@mattblaze@federate.social @SteveBellovin@infosec.exchange it sure seems like a lot of our system relies on not electing wannabe mafia bosses/dictators. Side note: this guy's obsession with self-aggrandizement strikes me as more pathetic than anything else.

@mattblaze@federate.social @SteveBellovin@infosec.exchange ohh...I see what you did there. I trust all of you who understand the law way better than I do that he can't change the name. But does that mean he can't give it a "secondary name" and just declare that to be the "preferred" name? As I understand it, that's what he did with the DoD, who's name is also set by statute?

@mattblaze@federate.social @SteveBellovin@infosec.exchange serious question for those of us who don't understand the mechanics of all of this. With the DoD, apparently what he did was issue an EO that authorized "Department of War" as a secondary name, make that the preferred name for use by the department, but keep the statutory name as the "Department of Defense." Could he try and pull something like that here? Basically "authorize" the use of a secondary name and then just declare that that's the preferred name?

@mattblaze@federate.social sigh. This is what happens when we de-emphasize education, logic, and basic scientific literacy. Does the CDC really have a web page justifying going against the overwhelming medical and scientific consensus, justified by the failure to prove a negative?! Someone please pinch me; I want to wake up from this nightmare. Better yet, punch me in the face.

@byterhymer@mastodon.social @me_@sueden.social @ole@chaos.social I suspect Apple was aware. As I understand it, the single-button mouse was a cost saving measure, and someone (Raskin?) convinced Jobs that a single button was workable.