The eagle-eyed will have spotted the first Helios experiments last week - run by my colleagues and our external partners.

arxiv.org/abs/2511.02125
arxiv.org/abs/2511.03686
arxiv.org/abs/2511.03689

To be fair, the DQI paper didn't claim an advantage for Maxcut but suggested one for a different problem

We see this is a step towards implementing recent quantum Gibbs sampling algorithms on quantum hardware.

Many thanks for the fantastic work to my co-authors Eric Brunner, Luuk Coopmans, Gabriel Matos, Fred Sauvage and Yuta Kikuchi, all @quantinuum.bsky.social

Finally, we analyse algorithmic errors by simulating the full quantum circuits. Circuit simulations with a noise model confirm the trade-off between algorithmic errors and noise, whereby a larger Trotter time step can lead to smaller overall error.

Then we analyse analytically and numerically convergence, mixing time, spectral gap etc. for a mixed-field Ising model and random k-local Pauli jump operators. We also find a certain noise resilience of the protocol at relevant error rates

We propose a simplified Lindblad simulation algorithm combining various earlier ideas, e.g. the Chen et al. quantum thermal state preparation algorithm, QDrift-like randomized jump operators, a single-ancilla protocol for the dissipative term, ETH for good convergence and simpler coherent term etc.

I would include Spiti Valley (Himachal Pradesh, India).

Besides the spectacular landscape, places like Tabo Monastery hold beautiful frescoes some of which date to ~1000 CE

Thank you

it will be interesting to see whether this "classical attack" gives anything for the "optimization by decoded quantum interferometry" paper
arxiv.org/abs/2408.08292

Yes please.