// Progress log · from the repositories' commits

What we did,
day by day.

Every result, correction and release, newest first. Filter by area; each entry links to its details.

Each entry is labelled: resultcorrectionreleasefollow upmilestone.

18 Nov 2026
resultQFT and QCDSymbolic computation

More instruments for quantum field theory in PQCF

More instruments for quantum field theory in PQCF. Feynman diagrams: every diagram of a theory with its symmetry factor and fermion sign, for QED and QCD equal to FORM's generator in all 24 cases compared (self-energies to three loops, vertices to two; 324 diagrams for the three-loop gluon self-energy, in 0.1 s), and exact against zero-dimensional field theory. Feynman rules from a Lagrangian: the QED and QCD vertices equal Peskin and Schroeder's term by term, and each propagator is the inverse of its quadratic Lagrangian. The running of QCD: the beta function and the quark mass anomalous dimension to five loops, exactly; alpha_s and quark masses at any precision, against mpmath to 28 digits and against CRunDec, the field's reference program, within its own accuracy. The one-loop box D0 at any precision, against mpmath and LoopTools (QCDLoop returns no number for two of the four configurations above threshold in our ARM build). How to reproduce: diagrams and rules; the running of QCD; one-loop integrals.

17 Nov 2026
resultDistributed computingGenomics

GenHub2 runs across machines

GenHub2 runs across machines. On a lab's two nodes it gives the same results, nearly twice as fast, and the tree built from 36 tiles is byte for byte the single-task one. Through the FVLLMP federation, its tasks go from DGX-1 to DGX-2 and come back identical, checked by hash.

16 Nov 2026
resultQFT and QCDSymbolic computation

First instruments for quantum mechanics and quantum field theory in PQCF

First instruments for quantum mechanics and quantum field theory in PQCF: angular-momentum algebra (Wigner symbols exactly) and SU(N) colour factors with N symbolic; perturbation series to high order (the anharmonic oscillator against Bender and Wu); spectra to 30–38 digits (the quartic oscillator, hydrogen); Dirac traces in d dimensions, equal to FORM on 60 traces of up to 12 matrices; one-loop integrals at any precision: the two-point functions equal mpmath to 35 digits and LoopTools to 1e-11, the three-point function is checked in 40 configurations against mpmath's direct integral (28 digits) and contour deformation (1e-12). On the way: in our builds on ARM, QCDLoop is wrong in 6 of those configurations and LoopTools in 7, all where the Källén function is negative; PQCF agrees with mpmath there. Details in the comparison's page; How to reproduce the results.

16 Nov 2026
resultPhotonicsSymbolic computation

Quantum optics in PQCF, at any precision

Quantum optics in PQCF, at any precision: operator algebra, master equations and quantum trajectories, exact steady states and spectra (also as rational functions of the parameters), cumulant expansions, collective atoms and input-output networks. Against QuTiP 5.3.1 on the same core: on four problems with closed forms PQCF has 36 correct digits where QuTiP has 5 to 16; QuTiP is faster at time evolution, 12 to 280 times on the driven Jaynes–Cummings model, PQCF faster on the steady state of resonance fluorescence. The driven Kerr cavity and the Kerr resonator with two-photon drive and loss equal their exact published solutions (Drummond and Walls; Bartolo et al.) to 30 digits; cumulant equations agree with QuantumCumulants.jl to 1e-13, the collective-atom solver with QuTiP's PIQS within PIQS's single precision. Details in the comparison's results; How to reproduce the results.

15 Nov 2026
correctionQFT and QCDSymbolic computation

Improved holographic QCD corrected

Improved holographic QCD corrected: the 6–15% difference from the paper was our error, one parameter of the model (a = 5/16 instead of 3/8). The glueball ratios now agree with the paper within 0.6%. Details in Holographic QCD in PQCF.

15 Nov 2026
resultQFT and QCDSymbolic computation

PQCF's symbolic engine reaches beyond FORM's field

PQCF's symbolic engine reaches beyond FORM's field. Polynomial gcd and factorisation are measured against FORM: the gcd is level with it, factorisation 4 to 8 times slower. Series, special functions, curvature, tensors and eigenvalue solvers to 30 digits are new. Seven holographic computations, from the action to the result, are checked against their papers, and six agree. For one (improved holographic QCD) our glueball ratios are 6–15% lower than the paper's, and we have not found the reason. Superseded the same day: the cause was our error, and the corrected ratios agree with the paper within 0.6% (the entry above). Details in Holographic QCD in PQCF.

14 Nov 2026
follow upSymbolic computation

PQCF's symbolic engine, faster where FORM was ahead

PQCF's symbolic engine, faster where FORM was ahead: results kept in Rust come back faster, powers multiply by merging, the symbolic permanent runs in Rust. Old against new on the same core: powers 1.6 to 2.2 times faster, sparse products 1.6 to 1.9, permanents 2.9 to 5.3 (now as fast as FORM or faster); the same answers, exactly. Details in Large-scale symbolic computation; every comparison on this page in Benchmarks.

14 Nov 2026
milestoneDistributed computingGenomics

GenHub2 starts: virus classification for genetic labs on a federation of lab nodes, the first application …

GenHub2 starts: virus classification for genetic labs on a federation of lab nodes, the first application built on PQCF, open source under Apache 2.0. It is the proving ground for PQCF's distributed task layer, the same layer quantum simulation workloads will run on. Phase A (one node) is done: on 593 held-out genomes of 30 viruses the species is right for 95.1%, 1,000 genomes are classified in 0.7 s. Seven more methods added and measured the same day (protein search, VeryFastTree, UShER, more Nextclade datasets, Hostile, iVar, Kraken 2 + Bracken); with proteins, the species is right for 96.5%. Details in GenHub2 starts.

13 Nov 2026
resultGenomicsPhotonics

Can PQCF help genome research?

Can PQCF help genome research? Several research papers used, and Gaussian boson sampling, coherent Ising machines, the GBS graph kernel and genome encoding measured on a virus, a bacterium and a human chromosome against the best classical methods. Classical exact methods solve these tasks in under a second; the photonic routes need hardware that does not exist or lose; genomes do not compress as quantum states. Details in Can PQCF help genome research?

12 Nov 2026
resultPhotonics

PQCF head to head with The Walrus, Piquasso, Perceval, DeepQuantum and Strawberry Fields on the same …

PQCF head to head with The Walrus, Piquasso, Perceval, DeepQuantum and Strawberry Fields on the same machine and core: all 51 cross-checks agree; PQCF's boson sampler is the fastest at every size, its permanent, torontonian and Fock kernels are behind the best by 1.4 to 3 times; a Piquasso bug on ARM found. Now run weekly; details in PQCF against the photonic simulators.

12 Nov 2026
resultQFT and QCDSymbolic computation

The three-loop β-function of QCD recomputed, every coefficient exact

The three-loop β-function of QCD recomputed, every coefficient exact: diagrams, Feynman rules and colour in FORM 5, the integrals reduced by PQCF at one and two loops and by FORM's Forcer at three, the renormalisation by PQCF; the results published with the code. Details in The three-loop β-function of QCD.

9 Nov 2026
resultQFT and QCDSymbolic computation

Four-loop QCD integrals from FORM's own Forcer checks, reduced by PQCF

Four-loop QCD integrals from FORM's own Forcer checks, reduced by PQCF: every answer exact, against Forcer's results and against the numbers FORM's checks assert. Forcer, built for these integrals, is 17 to 25 times faster on four loops (75 to 150 in our first run); details in Four-loop integrals from QCD.

8 Nov 2026
releaseDistributed computingQFT and QCDSymbolic computation

Feynman-integral reduction in PQCF, measured against FIRE on its own two-loop double box

Feynman-integral reduction in PQCF, measured against FIRE on its own two-loop double box: 1.21 seconds against 7.99 on one core, and 3.73 against 11.99 for a harder target, with the same answers. It is done by solving numerically modulo primes and reconstructing the exact result, with only the equations the target needs, the diagram's symmetries, denominator guessing and a learned elimination. The first attempt, on 7 November, was far slower than FIRE. Released as PQCF 1.4.0, the reduction now part of the library (release notes); the details in Reducing Feynman integrals.

7 Nov 2026
releaseSymbolic computation

PQCF 1.3.0 released

PQCF 1.3.0 released: the symbolic engine measured against FORM 5. Compiled programs for the resultant benchmarks of FORM's own papers now need fewer operations than FORM's best optimisation level, O4, on all three (2,741, 10,243 and 34,799 against 2,936, 10,722 and 35,419), through Horner schemes, shared subexpressions and a parallel tree search over variable orders. FORM still gets there faster. Expansion: dense products about 3 times faster than FORM, sparse products and powers 2.2 to 4.3 times its time (from up to 32), the interferometer's symbolic matrix 1.1 to 1.5. Backward compatible: release notes; the results and charts in Large-scale symbolic computation.

5 Nov 2026
releasePhotonics

PQCF 1.2.0 released, beyond the specification's roadmap

PQCF 1.2.0 released, beyond the specification's roadmap. PQCF can now be used from Python: a client package with a circuit builder, run, sample, exact probabilities and gradients, and conversion to and from Perceval and Piquasso, talking to PQCF over JSON. Density-matrix gates are 5 to 25 times faster: each gate is prepared once instead of for every row and column, with identical results. The 4-mode case with up to 20 photons now takes 1.2 seconds instead of 12. Backward compatible: release notes.

5 Nov 2026
releasePhotonicsSymbolic computation

PQCF 1.1.0 released

PQCF 1.1.0 released: Phase 4 of the specification's roadmap is complete. New: threshold detection of displaced light; feed-forward; gradients in one forward and one backward pass, also through loss, and exact parameter-shift rules; noise models (partially distinguishable photons, multi-photon sources, dark counts, thermal loss); permanents and hafnians on an NVIDIA GPU, every term identical to the CPU's bit for bit (permanents faster from 24 × 24, hafnians no faster on our GB10); density matrices stored by photon-number blocks (cutoff 20 on 4 modes in 0.18 GB instead of 1.81 GB). Every change is backward compatible: release notes; the results and charts in What the roadmap added. The benchmarks re-run on 1.1.0, and the comparison with FORM re-run and extended with the literature's multiplication benchmarks.

2 Nov 2026
milestoneSymbolic computation

A goal: large-scale symbolic computation in PQCF, measured against FORM 5 every week until the end of the …

A goal: large-scale symbolic computation in PQCF, measured against FORM 5 every week until the end of the project, with the results public. Two algorithm changes brought the symbolic permanent and polynomial powers from thousands of times slower than FORM to between FORM's speed and 22 times slower. An optional bridge to FORM handles larger polynomial work.

1 Nov 2026
releaseDistributed computingPhotonics

PQCF 1.0.0-rc.1: Phases 0–3 done and the interface frozen

PQCF 1.0.0-rc.1: Phases 0–3 done and the interface frozen. Every requirement of the specification is implemented and tested on x86_64 and ARM.

30 Oct 2026
milestonePhotonics

New project: PQCF, a photonic quantum computing framework (Erlang/OTP with Rust kernels), started in a …

New project: PQCF, a photonic quantum computing framework (Erlang/OTP with Rust kernels), started in a public repository; the first commit has the application skeleton and the random-number stream with its test.

29 Oct 2026
resultPhotonics

Photonic schemes, phase 7

Photonic schemes, phase 7: one mode in continuous variables — GKP readout under shift noise (exact, sampled and closed form, in q and in p), cat parity under photon loss, the homodyne-heralded comb; QxCV (Elixir) and cv.py (Python) agree; the review's Fig. 5 read from its PDF; a sign mismatch between its Eqs. 22 and 25.

29 Oct 2026
resultPhotonics

The review's 2.7 % loss figure traced to Bartolucci et al.

The review's 2.7 % loss figure traced to Bartolucci et al. (2023) and reproduced from their formulas: the encoded 6-ring network with Bell-pair-boosted fusion, 10.4 % per four-photon fusion.

29 Oct 2026
resultPhotonicsQuantum computing

Phase 6 at circuit level

Phase 6 at circuit level: surface-code memories with heralded erasure, decoded with the heralds by QxDecode and PyMatching, and by PyMatching without them — the heralds cut the logical error about twentyfold at d = 5, 4 % erasure.

29 Oct 2026
resultPhotonicsQuantum computing

Phase 6, erasure decoding

Phase 6, erasure decoding: QxDecode takes heralded erasures; pure erasure crosses at 50 % (d = 3–9), mixed erasure and flips shift the threshold as expected, agreeing with PyMatching; QxStab samples heralded erasure, checked against Stim.

29 Oct 2026
resultPhotonics

Photonic schemes, phase 5

Photonic schemes, phase 5: detector POVMs checked; the NS and CZ gates against detector efficiency (resolving and bucket); photon loss on a dual-rail qubit shown to be an exact erasure.

29 Oct 2026
resultPhotonicsQuantum computing

Photonic schemes, phase 4

Photonic schemes, phase 4: measurement-based computation on small clusters; the review's Fig. 10 chain gives a CZ, its Eq. 56 star a CNOT with Pauli byproducts.

29 Oct 2026
resultPhotonics

Photonic schemes, phase 3

Photonic schemes, phase 3: type-I and type-II fusion photon by photon; the n + m − 1 graph rule on state vectors; the cost of growing a chain — 7 Bell pairs per qubit at best among the strategies simulated, not the review's 6.5.

29 Oct 2026
resultPhotonics

Photonic schemes, phases 0–2 of our plan

Photonic schemes, phases 0–2 of our plan: QxFock (Elixir) and fock.py (Python), checked three ways with Perceval; the HOM dip; KLM's NS gate and CZ (1/4, 1/16, fidelity 1); the review's Fig. 8 angle sign found wrong against KLM's paper.

29 Oct 2026
resultQuantum computing

Night: the load test — QxDecode against union-find in C++ and Python and PyMatching, under open-loop load …

Night: the load test — QxDecode against union-find in C++ and Python and PyMatching, under open-loop load on the same cores. The hypothesis is falsified: 12–20 times less throughput than C++, a 99th percentile over 1 ms from 80 % load; the per-syndrome process is the bottleneck. Our first design's closed-loop load was itself flawed (Little's law), and is replaced.

29 Oct 2026
resultQuantum computing

Evening: QxDecode, our union-find decoder, compared with PyMatching on the same syndromes — 6 to 37 % more …

Evening: QxDecode, our union-find decoder, compared with PyMatching on the same syndromes — 6 to 37 % more logical errors, 14 to 50 times slower per core; a peeling bug found by the comparison and fixed.

29 Oct 2026
follow upQuantum computing

Later still: Shor names each failure and shows its exact distribution; power checks for every cross-check; …

Later still: Shor names each failure and shows its exact distribution; power checks for every cross-check; exact tests on the logical error rates and with the observable; the NumPy engine's 20-qubit jump profiled (OpenBLAS threads); the research question's design tightened.

29 Oct 2026
follow upQuantum computing

Later: Shor's algorithm in general (N = 15 or 21, any a); QxSim checked against Aer's dynamic circuits; …

Later: Shor's algorithm in general (N = 15 or 21, any a); QxSim checked against Aer's dynamic circuits; exact tests against Stim; the qx comparison published; the results called reproductions, and our research question stated.

29 Oct 2026
resultQuantum computing

03:06 — Grover on 2 to 10 qubits; the full distance-3 surface code; QxStab with gate-level noise, checked …

03:06 — Grover on 2 to 10 qubits; the full distance-3 surface code; QxStab with gate-level noise, checked against Stim; the threshold study and benchmarks (f0b6d50). 02:20 — Shor's circuit compiled for a = 7, teleportation with corrections, seeds, the repetition sweeps (80faa5d). 01:29 — full correction cycles for three codes (94fa43a).

28 Oct 2026
milestoneQuantum computing

17:18–21:42 — the site moved to its own host; the circuits as a live grid; the process models and their …

17:18–21:42 — the site moved to its own host; the circuits as a live grid; the process models and their photon pictures (f2eb632 … b6b3352).

May 2026
milestoneQuantum computing

The ten demo circuits and the BEAM service, on an earlier version of this site

The ten demo circuits and the BEAM service, on an earlier version of this site.

23 Mar 2026
milestoneQuantum computing

Quanterall becomes a quantum computing research company

Quanterall becomes a quantum computing research company (below).

Mar 2026
milestoneQuantum computing

Work starts: quantum demos on the BEAM, begun on Craig Richards' qx

Work starts: quantum demos on the BEAM, begun on Craig Richards' qx.