Stubsack: weekly thread for sneers not worth an entire post, week ending 28th December 2025

God damn that's good.

this made me cackle

very nice

Before: You were eaten by a grue.

After: Oops, All Grues!

If you want a warm and fuzzy Christmas contemplation, imagine turducken production at scale

This doesn't directly answer your question but I guess I had a rant in me so I might as well post it. Oops.

---

It's possible to write tools that make point changes or incremental changes with targeted algorithms in a well understood problem space that make safe or probably safe changes that get reviewed by humans.

Stuff like turning pointers into smart pointers, reducing string copying, reducing certain classes of runtime crashes, etc. You can do a lot of stuff if you hand-code C++ AST transformations using the clang / llvm tools.

---

Of course "let's eliminate 100% of our C code with a chatbot" is... a whole other ballgame and sounds completely infeasible except in the happiest of happy paths.

In my experience even simple LLM changes are wrong somewhere around half the time. Often in disturbingly subtle ways that take an expert to spot. Also in my experience if someone reviews LLM code they also tend to just rubber stamp it. So multiply that across thousands of changes and it's a recipe for disaster.

And what about third party libraries? Corporate code bases are built on mountains of MIT licensed C and C++ code, but surely they won't all switch languages. Which means they'll have a bunch of leaf code in C++ and either need a C++ compatible target language, or have to call all the C++ code via subprocess / C ABI / or cross-language wrappers. The former is fine in theory, but I'm not aware of any suitable languages today. The latter can have a huge impact on performance if too much data needs to be serialized and deserialized across this boundary.

Windows in particular also has decades of baked in behavior that programs depend on. Any change in those assumptions and whoops some of your favorite retro windows games don't work anymore!

---

In the worst case they'd end up with a big pile of spaghetti that mostly works as it does today but that introduces some extra bugs, is full of code that no one understands, and is completely impossible to change or maintain.

In the best case they're mainly using "AI" for marketing purposes, will try to achieve their goals using more or less conventional means, and will ultimately fall short (hopefully not wreaking too much havoc in the progress) and give up halfway and declare the whole thing a glorious success.

Either way ultimately if any kind of large scale rearchitecting that isn't seen through to the end will cause the codebase to have layers. There's the shiny new approach (never finished), the horrors that lie just beneath (also never finished), and the horrors that lie just beneath the horrors (probably written circa 2003). Any new employees start by being told about the shiny new parts. The company will keep a dwindling cohort of people in some dusty corner of the company who have been around long enough to know how the decades of failed code architecture attempts are duct-taped together.

The short answer is no. Outside of this context, I'd say the idea of "code modifications algorithmically at scale" is the intersection of code generation and code analysis, all of which are integral parts of modern development. That being said, using LLMs to perform large scale refactors is stupid.

Not just pinkies, my friend, we are promising with all fingers, at scale!

so, ever watched Godzilla? and then did a twofer with a zombie movie? I think that’s essentially the plot here

I'm a lot more afraid of the recursive effect of this joke. We could have at scale at scale.

randomly placed and statistically average, just like real rivers!

Aldaquaque is where I took that bleft toin

Albrequerre

Hey, look, I found someone who believes things can have feelings:

this made my day, thx

A pal: “I take back everything I ever said about humanoid robots being useless”

And… same

JavaScript isn't even the worst Brendan Eich misdeed!

I strongly disagree. I would much rather write a JavaScript or C++ application than ever touch Rob Pikes abortion of a language. Thankfully, other options are available.

Its standard crypto libraries are also second to none.

I'm also a big fan of the concurrency implementation, I wish other languages made it so easy to use green threads & channels.

More like

Would you like to know more

I mean, sure

Here's a 13,000-word retrospective

Ah, nah fam

Upvoted, but also consider: boycotts sometimes work. BDS is sufficiently effective that there are retaliatory laws against BDS.

nonviolent civil disobedience and direct action are just tactics that work in specific circumstances and can achieve specific goals. pretty much every violent movement for change was supported by non-violent movements. and non-violence often appears in a form that is unacceptable to the slatestarcodex contingent. Like Daniel Berrigan destroying Vietnam draft cards, or Catholic Workers attacking US warplanes with hammers, or Black Lives Matter activists blocking a highway or Meredith and Verity Burgmann running onto the pitch during a South African rugby match.

Yeah, dont see me linking to this post by Gelderloos as fully supporting his other stances, more that the whole way Scott says nonviolent movements are better isnt as supported as he says (and also shows Scott never did much activism). So more anti scott than pro Peter ;).

I really don't know how he can fail to see the irony or hypocrisy at complaining about people trading made up probabilities, but apparently he has had that complaint about P(doom) for a while. Maybe he failed to write a call out post about it because any criticism against P(doom) could also be leveled against the entire rationalist project of trying to assign probabilities to everything with poor justification.

Super-prediction is difficult, especially about the super-future. —old Danish proverb

P(hole)

good for you ig. ozempic is actually small enough (and profitable enough) to make it synthetically, but novo process is to make linear precursor by fermentation, purify that, then tack on it side chain and N-terminal H-His-Aib- using regular peptide chemistry methods. no such luck with retatrutide tho, it has to be entirely synthetic. the real big deal however will be about small-molecule drug that targets this receptor, because this means pills instead of injections from day 1

@saucerwizard Do you drink alcohol, and if so, what has semaglutide done to your desire to drink? (I know my alcohol consumption crashed by about 80% when I went on Rybelsus—the oral formulation of semaglutide, the GLP-1 agonist in Ozempic™—and it's a common enough side-effect that it's undergoing clinical trials as an anti-addiction medication.)

Yeah, it's not like reviewers can just write "This paper is utter trash. Score: 2" unless ML is somehow an even worse field than I previously thought.

They referenced someone who had a paper get rejected from conferences six times, which to me is an indication that their idea just isn't that good. I don't mean this as a personal attack; everyone has bad ideas. It's just that at some point, you just have to cut your losses with a bad idea and instead use your time to develop better ideas.

So I am suspicious that when they say "constructive feedback", they don't mean "how do I make this idea good" but instead "what are the magic words that will get my paper accepted into a conference". ML has become a cutthroat publish-or-perish field, after all. It certainly won't help that LLMs are effectively trained to glaze the user at all times.

terrible state of house CO detectors is not an excuse for putting a CO generator in the living room.

Going from lazy, sloppy human reviews to absolutely no humans is still a step down. LLMs don't have the capability to generalize outside the (admittedly enormous) training dataset they have, so cutting edge research is one of the worse use cases for them.

I think I read that post and thought it was incredibly naive, on the level of "why does the barkeep ask if I want a drink?" or "why does the pretty woman with a nice smile want me to pay for the VIP lounge?" Cheap clanky services like forums and mailing lists and Wordpress blogs can be maintained by one person or a small club but if you want something big, smooth, and high-bandwidth someone is paying real money and wants something back. Examples in the original post included geocities, collegeclub.com, MySpace, Friendster, Livejournal, Tumblr, Twitter and those were all big business which made big investments and hoped to make a profit.

Anyone who has helped run a medium-sized club or a Fedi server has faced an agenda item like "we are growing. Input of resources from new members is not matching the growth in costs and hassle. How do we explain to the new members what we need to keep going and get them to follow up? "

There is a whole argument that VC-backed for-profit corporations are a bad model for hosting online communities but even nonproffits or Internet celebrities with active comments face the issue "this is growing, it requires real server expenses and professional IT support and serious moderation. Where are those coming from? Our user base is used to someone else invisibly providing that."

@o7___o7 @BlueMonday1984

> AI is going to democratize the way people don't know what they're doing

Ok, sometimes you do got to hand it to them

oof ow my bones why do my bones hurt?

–a man sipping from a bottle labeled “bone hurting juice”

this was literally our elevator pitch for Pivot to AI

energy goes in, energy goes out. you can't explain that

Power! Unlimited power! I say as I shoot Force Electrons into the jedi knights.

Ok, lot to respond to here. I'm familiar with the relevant mathematics.

First, I think it's clear from the litany of failed attempts that you can't write down a finite set of rules that tells us what is true about the world. At least to me, it's also intuitively clear that you can't write down such a set of rules. That is not, without considerable auxiliary claims at least, a consequence of gödel's incompleteness theorem, nor does gödel's incompleteness theorem follow from it.

The essential issue here is that the incompleteness theorem deals with formal statements and formal reasoning in formal languages. There is a significant gap between the perfectly acceptable reasoning we use every day to understand the world around us, which if it can be written down at all often requires us to use informal language, and the sort of thing the incompleteness theorem addresses. There are real philosophical claims to be made and argued, which in at least implicit form go far back and have yet to be pinned down. For example, the sorites paradox can be understood as an (informal) proof that induction can fail in informal arguments. The whole thing, rather than being clarified, becomes more and more hopelessly complex the more one thinks about it.

I agree that inasmuch as objectivism pretends to formality it makes itself vulnerable to mathematical theorems, and surely would evaporate on contact with them. But the failure of the pretense to formality itself renders the issue moot.

Second, the question of what is or is not a "consequence" of this or that theorem is, given the nature of implication, a little difficult to pin down in borderline cases (are not all theorems a consequence of "T -> T"?). I'm perfectly fine with calling cantor's theorem, the halting theorem, the incompleteness theorem, etc instances of lawvere's fixed point theorem. But there is significant work required to take the hypotheses of some of these types of theorems and maneuver things into such a position as to apply the FPT. So I don't think it's a consensus opinion.

Third, if we want to describe a postmodern movement in mathematics, while I'm not sure about the dates, I get what you're going for and it makes sense to me. But I think the description you're putting forth here gives way too much weight to theorems. The movement toward a post-modern sensibility is imo much more marked by a movement toward guiltlessly abstract definitions and axioms. Consider the centuries of the torment that mathematicians experienced trying to justify or explain what negative numbers or complex numbers really are. In contrast, in the 20th century one has the definition of a scheme in algebraic geometry, a kind of space characterized by functions on it which are not really functions and which can be 0 at every point and yet not the zero function. What is the meaning of such a thing? Well, it is up to the individual mathematician to accept their own metaphors explaining that matter. Totally unthinkable a century prior. Examples of this sort of thing abound (for example, test functions in functional analysis). The movement toward such things was doubtless urged on by the impossibility theorems you refer to but goes far beyond them and is far more impactful imo.

I totally agree. The linked PauseAI leader still doesn't realize the full extent of the problem, but I'm kind of hopeful they may eventually figure it out. I think the ability to simply say this is bullshit (about in group stuff) is a skill almost no lesswrongers and few EAs have.