The title of this post changed as I was reading it. "It looks like the 'JVG algorithm' only wins on tiny numbers" is a charitable description. The article is Scott Aaronson lambasting the paper and shaming its authors as intellectual hooligans.
Agree. Scott is exactly correct when he just straight calls it crap.
It's inaccurate to say it wins on small numbers because on small numbers you would use classical computers. By the time you get to numbers that take more than a minute to factor classically, and start dreaming of quantum computers, you're well beyond the size where you could tractably do the proposed state preparation.
What do you mean? The original 2019 supremacy experiment was eventually simulated, as better classical methods were found, but the followups are still holding strong (for example [4] and [5]).
There was recently a series of blog posts by Dominik Hangleiter summarizing the situation: [1][2][3].
the reason people pay attention to him is that he does a good job publicizing both positive and negative results, and accurately categorizing which are bullshit
While I think the idea that claiming one can "precompute the xr mod N’s on a classical computer" sounds impractical there are a subset of problems where this might be valid. According to computational complexity theory, there's a class of algorithms called BQP (bounded-error quantum polynomial time).
Shor's algorithm is part of BQP. Is the JVC algorithm part of BQP, even though it utilizes classical components? I think so.
I believe that the precomputational step is the leading factor in the algorithm's time complexity, so it isn't technically a lower complexity than Shor's. If I had to speculate, there will be another class in quantum computational complexity theory that accommodates precomputation utilizing classical computing.
I welcome the work, and after a quick scroll through the original paper, I think there is a great amount of additional research that could be done in this computational complexity class.
There is a genuinely interesting complexity class called BQP/poly, which is pronounced something like “bounded-error quantum polynomial time with classical advice” (add some more syllables for a complete pronunciation).
The JVG algorithm is not a high quality example of this or really anything else. If you think of it as “classical advice”, then it fails, because the advice depends on the input and not just the size of the input. If you think of it as precomputation, it’s useless, because the precomputation involved already fully solves the discrete log problem. And the JVG paper doesn’t even explain how to run their circuit at respectable sizes without the sheer size of the circuit making the algorithm fail.
It’s a bit like saying that one could optimize Stockfish to run 1000x faster by giving it an endgame table covering all 16-or-fewer-piece-positions. Sure, maybe you could, but you also already solved chess by the time you finish making that table.
JVC isn't BQP. it's exp time (I.e. worse than factoring without a quantum computer at all). it takes the only step of shors algorithm that is faster to run on a quantum computer and moves it to a classical computer
I didn't get the quantum hype last year. At least with AI, you can see it do some impressive things with caveats, and there are bull and bear cases that are both reasonable. The quantum hype training is promising the world, but compared to AI, it's at the linear regression stage.
People get taken by the theoretical coolness and ultimate utility of the idea, and assume it's just a matter of clever ideas and engineering to make it a reality. At some point, it becomes mandatory to work on it because the win would be so big it would make them famous and win all sorts of prizes and adulation.
QC is far earlier than "linear regression" because linear regression worked right away when it was invented (reinvented multiple times, I think). Instead, with QC we have: an amazing theory based on our current understanding of physics, and the ability to build lab machines that exploit the theory, and some immediate applications were a powerful enough quantum computer built. On the other side, making one that beats a real computer for anything other than toy challenges is a huge engineering challenge, and every time somebody comes up with a QC that does something interesting, it spurs the classical computing folks to improve their results, which can be immediately applied on any number of off-the-shelf systems.
> People get taken by the theoretical coolness and ultimate utility of the idea, and assume it's just a matter of clever ideas and engineering to make it a reality. At some point, it becomes mandatory to work on it because the win would be so big it would make them famous and win all sorts of prizes and adulation.
Good description. Commercial fusion power seems to be in the same category currently.
The next step once you have enough thinkers working on the problem is to start pretending that commercial success is merely a few years away, with 5 or 10 years being the ideal number.
my understanding is that they factored 15 using a modular exponentiation circuit that presumes that the modulus is 3. factoring 15 with knowledge of 3 is not so impressive. Shor's algorithm has never been run with a full modular exponentiation circuit.
The very first demonstration of factoring 15 with a quantum computer, back in 2001, used a valid modular exponentiation circuit [1].
The trickiest part of the circuit is they compile conditional multiplication by 4 (mod 15) into two controlled swaps. That's a very elegant way to do the multiplication, but most modular multiplication circuits are much more complex. 15 is a huge outlier on the difficulty of actually doing the modular exponentiation. Which is why so far 15 is the only number that's been factored by a quantum computer while meeting the bar of "yes you have to actually do the modular exponentiation required by Shor's algorithm".
would other mersenne numbers admit the same trick? if so, factoring 2047 would be really interesting to see. it's still well within the toy range, but it's big enough that it would be a lot easier to believe that the quantum computer was doing something (15 is so small that picking an odd number less than sqrt(15) is guaranteed to be a correct factorization)
> (yes, the authors named it after themselves)
The same way the AVL tree is named after its inventors - Georgy Adelson-Velsky and Evgenii Landis... Nothing peculiar about this imh
This might not be something entirely obvious to people outside of academia, but the vast majority (which I'm only weakening a claim of "totality" in order to guard against unknown instances) of entities that bear the name of humans in the sciences do so because other people decided to call them by that name.
From another view, Adelson-Velsky and Landis called their tree algorithm "an algorithm for the organization of information" (or, rather, they did so in Russian --- that's the English translation). RSA was called "a method" by Rivest, Shamir, and Adleman. Methods/algorithms/numbers/theorems/etc. generally are not given overly specific names in research papers, in part for practical reasons: researchers will develop many algorithms or theorems, but a very small proportion of these are actually relevant or interesting. Naming all of them would be a waste of time, so the names tend to be attached well after publication.
To name something after oneself requires a degree of hubris that is looked down upon in the general academic community; the reason for this is that there is at least a facade (if not an actual belief) that one's involvement in the sciences should be for the pursuit of truth, not for the pursuit of fame. Naming something after yourself is, intrinsically, an action taken in the seeking of fame.
Same with RSA and other things, I think the author's point is that slapping your name on an algorithm is a pretty big move (since practically, you can only do it a few times max in your life before it would get too confusing), and so it's a gaudy thing to do, especially for something illegitimate.
But also note that naming an algorithm, in and of itself, is fine; it's naming it after yoursel(f,ves) in the initial paper that's a sign of crackpottery.
* Named by: Probably fine but heavily weighted on the grandiosity of the title.
* Named after: Almost certainly fine (unless it's something like "X's Absolute Drivel Faced Garbage That Never Works Because X Kidnapped My Dog And Is A Moral Degenerate Algorithm", obvs.)
* Named by yoursel(f,ves) after yoursel(f,ves): In the initial paper? Heavy likelihood of crackpottery. Years later? Egotistical but strong likelihood of being a useful algorithm.
The title of this post changed as I was reading it. "It looks like the 'JVG algorithm' only wins on tiny numbers" is a charitable description. The article is Scott Aaronson lambasting the paper and shaming its authors as intellectual hooligans.
Agree. Scott is exactly correct when he just straight calls it crap.
It's inaccurate to say it wins on small numbers because on small numbers you would use classical computers. By the time you get to numbers that take more than a minute to factor classically, and start dreaming of quantum computers, you're well beyond the size where you could tractably do the proposed state preparation.
Well, the reviewers missed it too.
What reviewers? It's not a peer reviewed article.
Ok.
Scott Aaronson is the guy who keeps claiming quantum supremacy is here every year so he's like the proverbial pot calling the kettle black.
What do you mean? The original 2019 supremacy experiment was eventually simulated, as better classical methods were found, but the followups are still holding strong (for example [4] and [5]). There was recently a series of blog posts by Dominik Hangleiter summarizing the situation: [1][2][3].
[1]: https://quantumfrontiers.com/2026/01/06/has-quantum-advantag...
[2]: https://quantumfrontiers.com/2026/01/25/has-quantum-advantag...
[3]: https://quantumfrontiers.com/2026/02/28/what-is-next-in-quan...
[4]: https://arxiv.org/abs/2303.04792
[5]: https://arxiv.org/abs/2406.02501
the reason people pay attention to him is that he does a good job publicizing both positive and negative results, and accurately categorizing which are bullshit
All I know is he keeps being wrong about quantum supremacy but maybe this is the year he finally gets his wish.
he's been right about it. quantum supremacy was achieved in 2023 (but only for incredibly useless problems)
Yeah I think GP might now prefer his statement(s) to have been about "quantum _advantage_". Which is the modish term after all.
Scott References the top comment on this previous HN discussion
https://news.ycombinator.com/item?id=47246295
While I think the idea that claiming one can "precompute the xr mod N’s on a classical computer" sounds impractical there are a subset of problems where this might be valid. According to computational complexity theory, there's a class of algorithms called BQP (bounded-error quantum polynomial time).
Shor's algorithm is part of BQP. Is the JVC algorithm part of BQP, even though it utilizes classical components? I think so.
I believe that the precomputational step is the leading factor in the algorithm's time complexity, so it isn't technically a lower complexity than Shor's. If I had to speculate, there will be another class in quantum computational complexity theory that accommodates precomputation utilizing classical computing.
I welcome the work, and after a quick scroll through the original paper, I think there is a great amount of additional research that could be done in this computational complexity class.
There is a genuinely interesting complexity class called BQP/poly, which is pronounced something like “bounded-error quantum polynomial time with classical advice” (add some more syllables for a complete pronunciation).
The JVG algorithm is not a high quality example of this or really anything else. If you think of it as “classical advice”, then it fails, because the advice depends on the input and not just the size of the input. If you think of it as precomputation, it’s useless, because the precomputation involved already fully solves the discrete log problem. And the JVG paper doesn’t even explain how to run their circuit at respectable sizes without the sheer size of the circuit making the algorithm fail.
It’s a bit like saying that one could optimize Stockfish to run 1000x faster by giving it an endgame table covering all 16-or-fewer-piece-positions. Sure, maybe you could, but you also already solved chess by the time you finish making that table.
JVC isn't BQP. it's exp time (I.e. worse than factoring without a quantum computer at all). it takes the only step of shors algorithm that is faster to run on a quantum computer and moves it to a classical computer
I mean, considering that no quantum computer has ever actually factored a number, a speedup on tiny numbers is still impressive :P
I didn't get the quantum hype last year. At least with AI, you can see it do some impressive things with caveats, and there are bull and bear cases that are both reasonable. The quantum hype training is promising the world, but compared to AI, it's at the linear regression stage.
It's a variation of nerd snipe. https://xkcd.com/356/
People get taken by the theoretical coolness and ultimate utility of the idea, and assume it's just a matter of clever ideas and engineering to make it a reality. At some point, it becomes mandatory to work on it because the win would be so big it would make them famous and win all sorts of prizes and adulation.
QC is far earlier than "linear regression" because linear regression worked right away when it was invented (reinvented multiple times, I think). Instead, with QC we have: an amazing theory based on our current understanding of physics, and the ability to build lab machines that exploit the theory, and some immediate applications were a powerful enough quantum computer built. On the other side, making one that beats a real computer for anything other than toy challenges is a huge engineering challenge, and every time somebody comes up with a QC that does something interesting, it spurs the classical computing folks to improve their results, which can be immediately applied on any number of off-the-shelf systems.
> People get taken by the theoretical coolness and ultimate utility of the idea, and assume it's just a matter of clever ideas and engineering to make it a reality. At some point, it becomes mandatory to work on it because the win would be so big it would make them famous and win all sorts of prizes and adulation.
Good description. Commercial fusion power seems to be in the same category currently.
The next step once you have enough thinkers working on the problem is to start pretending that commercial success is merely a few years away, with 5 or 10 years being the ideal number.
The problem is that it's an exponential slowdown on large numbers.
Hey hey, 15 = 3*5 is factoring.
my understanding is that they factored 15 using a modular exponentiation circuit that presumes that the modulus is 3. factoring 15 with knowledge of 3 is not so impressive. Shor's algorithm has never been run with a full modular exponentiation circuit.
The very first demonstration of factoring 15 with a quantum computer, back in 2001, used a valid modular exponentiation circuit [1].
The trickiest part of the circuit is they compile conditional multiplication by 4 (mod 15) into two controlled swaps. That's a very elegant way to do the multiplication, but most modular multiplication circuits are much more complex. 15 is a huge outlier on the difficulty of actually doing the modular exponentiation. Which is why so far 15 is the only number that's been factored by a quantum computer while meeting the bar of "yes you have to actually do the modular exponentiation required by Shor's algorithm".
[1]: https://arxiv.org/pdf/quant-ph/0112176#page=15
would other mersenne numbers admit the same trick? if so, factoring 2047 would be really interesting to see. it's still well within the toy range, but it's big enough that it would be a lot easier to believe that the quantum computer was doing something (15 is so small that picking an odd number less than sqrt(15) is guaranteed to be a correct factorization)
> (yes, the authors named it after themselves) The same way the AVL tree is named after its inventors - Georgy Adelson-Velsky and Evgenii Landis... Nothing peculiar about this imh
This might not be something entirely obvious to people outside of academia, but the vast majority (which I'm only weakening a claim of "totality" in order to guard against unknown instances) of entities that bear the name of humans in the sciences do so because other people decided to call them by that name.
From another view, Adelson-Velsky and Landis called their tree algorithm "an algorithm for the organization of information" (or, rather, they did so in Russian --- that's the English translation). RSA was called "a method" by Rivest, Shamir, and Adleman. Methods/algorithms/numbers/theorems/etc. generally are not given overly specific names in research papers, in part for practical reasons: researchers will develop many algorithms or theorems, but a very small proportion of these are actually relevant or interesting. Naming all of them would be a waste of time, so the names tend to be attached well after publication.
To name something after oneself requires a degree of hubris that is looked down upon in the general academic community; the reason for this is that there is at least a facade (if not an actual belief) that one's involvement in the sciences should be for the pursuit of truth, not for the pursuit of fame. Naming something after yourself is, intrinsically, an action taken in the seeking of fame.
Adelson-Velsky and Evgenii Landis were not the ones who named their tree the "AVL tree".
In my "crackpot index", item 20 says:
20 points for naming something after yourself. (E.g., talking about the "The Evans Field Equation" when your name happens to be Evans.)
https://math.ucr.edu/home/baez/crackpot.html for the curious (on that version it's item 25 though :o )
The last line of the introduction
> By doing so, we aim to provide a novel paradigm [...]
also made me think of item 19 on your list:
> 10 points for claiming that your work is on the cutting edge of a "paradigm shift".
I'm sad though that you didn't call it the "Baez crackpot index"...
I find it especially strange that two of the authors gave their first name to the algorithm.
Like RSA?
RSA was also not given that name by its authors, the name came later, which is usually the case.
In the original paper they do not give it any name: https://people.csail.mit.edu/rivest/Rsapaper.pdf
Same with RSA and other things, I think the author's point is that slapping your name on an algorithm is a pretty big move (since practically, you can only do it a few times max in your life before it would get too confusing), and so it's a gaudy thing to do, especially for something illegitimate.
The RSA authors didn’t name the algorithm after themselves.
Leonhard Euler has entered the chat: https://en.wikipedia.org/wiki/List_of_topics_named_after_Leo...
Nothing on that list has been named that way by Euler himself of course.
if you’re Euler you get a pass
Named after != named by
> Named after != named by
But also note that naming an algorithm, in and of itself, is fine; it's naming it after yoursel(f,ves) in the initial paper that's a sign of crackpottery.
* Named by: Probably fine but heavily weighted on the grandiosity of the title.
* Named after: Almost certainly fine (unless it's something like "X's Absolute Drivel Faced Garbage That Never Works Because X Kidnapped My Dog And Is A Moral Degenerate Algorithm", obvs.)
* Named by yoursel(f,ves) after yoursel(f,ves): In the initial paper? Heavy likelihood of crackpottery. Years later? Egotistical but strong likelihood of being a useful algorithm.