Jake Xuereb

"You're a cargo cult scientist,
A real crank turning nihilist,
I only found your citations--
On Google Squalor"

Avidly preparing for my next academic rap battle, that's how tenure works right?

Really cool, nice to see the use of dirty qubits

(3/3) The "quantumness" of thermal machines is an actively debated topic - showing that this "t" value is related to the quantumness of the heat exchange seems like an exciting challenge.

A computational TUR might lurk here...

(2/3) Interestingly, it seems one can only beat the classical probabilistic sample complexity if the the two scenarios lead to probe ground state populations that are different enough. We call this "t" the distinguishability parameter in the manuscript.

(1/3) Updated on the arXiv this morning! No more painful typos and new appendices. One comparing our sample complexity with that of classical probabilistic approaches and another providing a recipe for oracles to prepare thermal machines.

arxiv.org/abs/2505.15887

View of the Maltese Archipelago from Gozo this morning 🇲🇹

YouTube video by CrystalRoseCreations Mitchell and Webb: "Are we the baddies?"

Reminiscent of

youtu.be/ToKcmnrE5oY?...

Easy to feel powerless in times like these.

Trying to remember that science is a political act and a force that brings people together from all over the world out of excitement for the beauty of nature.

In this sense, science is an act of defiance against those that wish to divide humanity.

This goes hard.

(7/n) This was only possible and fun thanks to a great group of people.

Thanks to Jinming, an undergrad intern who kind of inspired the project. My close collaborator Ben Stratton for the weekly Zooms. Alberto for making our math solid. And Pharnam and Marcus for the guidance.

(6/n) We continue to do a whole host of things, like prove a Carnot like limit on how much you can cool per step of a protocol

But our last contribution is a method to optimise cooling unitaries by showing they can be reduced to perfect matchings on bipartite graphs.

(5/n) While machines with fixed energy gaps in their qubits - as those traditionally considered in algorithmic cooling, are never reducible!

(4/n) In particular we're able to derive a condition which allows one to energetically design useful cooling machines which are never reducible as n increases.

Here we show that machines whose energies grow exponentially in the no. of qubits are always reducible to a single qubit 😬

(3/n) We term this, k-reducibility, when a cooling scenario involving an n qubit machine can be reducd to one on n-k qubits -- meaning k of your qubits were useless!

This is fully characterised by our inequalities which show that it's the energetic structure of machines which underpins reducibility

(2/n) When starting to looking into this problem we were interested in understanding "when is cooling hard?" i.e. when is the unitary required to cool something complex.

E.g. in cooling a qubit with two others, when does cooling require a bipartite or tripartite interaction?

(1/n) Today on the arXiv, we examine the folklore that cooling a quantum system with access to another requires one to order their joint state eigenvalues in decreasing order and find that this sorting is actually dictated by a simple set of inequalities.

scirate.com/arxiv/2506.1...

It's conceivable to to think of a universe where free energy would take on a different functional form - changing equilibration rates (or whether they equilibriate at all and to what fixed point...) and ultimately the rate at which information propagates through systems and becomes objective.

Maybe @itsmealeale.bsky.social and @mathsmire.bsky.social would also enjoy weighing in here

There would be no emergent causal order to the universe without the production of entropy.

The total order on equilibrium states and partial orders on quantum thermal states dictate which state transformations are possible.

Without equilibration nothing hangs out long enough to form structure.

Totally like not tweeting about this for any specific reasons but the surname Jozsa as in Deutsch-Jozsa is spelled "zs" not "sz" in case like I dunno you like umm put an arxiv posting with this name in the title...

A Universal Constraint on Computational Rates in Physical Systems Conventional computing has many sources of heat dissipation, but one of these--the Landauer limit--poses a fundamental lower bound of 1 bit of entropy per bit erased. 'Reversible Computing' avoids thi...

It is somewhat strange to see @quantamagazine.bsky.social highlighting a 2022 preprint without any citations

arxiv.org/abs/2208.11196

Not that citations or being published is an indicator of quality - but when the preprint is being used to substantiate the claims of a start-up, it's questionable?

I think a lot about how Aaron Swartz committed suicide while facing a probably long jail sentence for download some of JSTOR, but no one at Meta is going to face any repercussions for using all of Libgen to train its AI.

Deep congratulations to the whole team but especially to Florian on what I'm sure is only his first Nature Physics paper.

I am also super happy to see the Malta - Vienna connection crystallized in such a paper facilitated by @aspects-quantum.bsky.social

Thanks so much for the shout out and the kind words @felixled.bsky.social

(7/7) I enjoyed this solo project and found the implications, that you can do some quantum computations without coherences and using thermal states, thought provoking to say the least.

Thanks to all my friends for encouraging me on to give this a shot on my own - wouldn't have done it without you.

(6/7) Most importantly, since we've encoded our solution into a property of a mixed state - we need samples! Thankfully and surprisingly, if an energetic condition is satisfied, the no. of samples required to solve Deutsch-Josza is independent of the problem size!

(5/7) Fascinatingly, the probe temperature and energy must scale with the size of the problem to be sensitive to it - but not too much! Or it'll become insensitive to the difference between the different scenarios.

This is my favorite insight, giving a notion of thermo cost for a query.

(4/7) The temperature of the probe after query is now sensitive to the property of the function! We examine this for the Deutsch-Josza problem where the oracle uses an exponential no. of thermal qubits for encoding and the Bernstein-Vazirani problem where they can get away with a linear number.

(3/7) Here we exchange the de-excited probe + all excited machine state with its opposite - excited probe + no excitations in machine.

Because of the oracle's encoding, the all excited state contains information about the sum of the function's output!

(2/7)Imagine an oracle with access to N thermal qubits whose energy gaps it can change to encode a Boolean function.

Can an agent exchange heat with this oracle's thermal machine using a single qubit and learn properties of the function? Yes, using the famous virtual qubit subspace swap.