Status: what is built and what is left

State on 2026-11-18, taken from the code, the logs and the saved notebook outputs. Dates are the last real run.

  • done: built and run for real.

  • partial: built, part of it unverified or missing.

  • export only: solver input is written but never solved.

  • missing: not built.

What has never been verified

These gaps matter most.

  • No converged AEDT run. Every eigenmode used fast settings (at most 3 passes, delta f 2.8-3.2 % against a 1 % target). A converged eigenmode needs a machine with 32 GB or more. In tutorial 4 the LOM and the FEM chi differ by about 54 %.

  • The kept FEM results are stale. Every kept tutorial result was solved for an earlier revision of its job. 57 of 153 sampled sim groups export differently since the November export changes, so they must be solved again.

  • Palace, Elmer and Sonnet were never solved. Only the input files are written; solving with them is a future development goal. The Linux runner and queue were never used on a server.

  • The FEM loop never closed on target. Tutorial 4 got to about 0.2 % of the targets but reports “on target (FEM): False”.

  • The plugin’s eigenmode run, --aedt gui, and a whole-chip simulation from the GUI were never run. Whole-chip Run LOM of the 12- and 5-qubit tutorial chips takes more than 10 minutes.

  • The wheel was built on 2026-11-17 and installed from the wheelhouse in a scratch environment (qcc selftest: 29 of 29 checks).

  • The fast test suite passed on 2026-11-19 on the release tree, with the PyPI releases of KQCircuits, Quantum Metal, SQcircuit, QuCat and PyAEDT installed: 3132 passed, 270 skipped (the AEDT read-back tests need local AEDT result files and skip without them). The tutorial “saved clean” test was dropped on 2026-11-18: tutorials are saved with their outputs.

  • There is no AEDT screenshot in these docs: capturing one needs AEDT running.

Modules

Module

Page

Status

Last real run / evidence

KLayout plugin (menu, tools, panel, Flow strip)

plugin tour

done

GUI smokes all pass 2026-11-17 (83, 49, 39, 37 checks; 62 cells; 95 tutorial checks); end-to-end with a real non-graphical AEDT Q3D

KLive scripted design (Designer)

scripted design

done

live loop 0.1 s per change 2026-11-02; window handling 2026-11-13

PDK

PDK

done

rules changed 2026-11-17, 105 fab exports compared

Component library and catalog

components

done

62-cell GUI sweep 2026-11-17; 142 cell pages

design.json schema and Python export

routing and DRC

done

105 examples round-trip 2026-11-15

Assembly and routing

routing and DRC

done, edge cases open

every example builds 2026-11-18

Connectivity DRC and netlist

routing and DRC

done

rules added 2026-11-17

Layout to LOM and LOM simulation

LOM

done; model gaps

all 93 notebooks 2026-11-18

Circuit IR, estimators, consensus, QuTiP

LOM, results

done

consensus on real FEM 2026-11-13

Sim-group export

simulation

done for Ansys

angled-crossing ports 2026-11-18

HFSS / Q3D driver (AEDT 2024 R1)

simulation

done with fast settings

real solves 2026-11-03/04, 11-13, 11-14, 11-17

Elmer, Palace, Sonnet

simulation

export only; future development goal

never solved

Runner and server queue

simulation

partial

unit tests only

Result read-back

results

done for AEDT

2026-11-13

Reports and plots

results

partial

LOM and consensus plots yes; convergence, S21, EPR as tables only (plots being built)

Dashboard tracker

dashboard

done

all tutorials 2026-11-13; t04 served 2026-11-18

Feedback and co-design

feedback

partial

one real FEM loop (tutorial 4, 2026-11-13)

LOM-first synthesis

LOM

done for Xmon + lambda/4 on one feedline

examples/lom_from_layout.ipynb

Fab export

fab

done; wafer assembly being built, mask assembly missing

105 exports 2026-11-17

qcc CLI

CLI reference

done

the reference page lacks design junctions and design placement-junction

Wheel and packaging

getting started

partial

wheel built and installed in a scratch env; 0.0.1b0 was the first PyPI release, 0.0.2 the current one

Docs site

the docs

partial

rewritten as a section guide 2026-11-19; the design.json field reference is still missing

The three design routes

Route

Status

Evidence

Code-first in Jupyter, live in KLayout

done

tutorials 0-4, checker runs 2026-11-18

Drag and drop in the plugin to netlist, circuit.json, design.py

done

tutorials 1-4 redrawn in the GUI 2026-11-17, layout difference 0, 0, 0 and 3.9 µm squared

KLayout, render, sim parameters, LOM

done for Q3D

end-to-end smoke 2026-11-17; eigenmode from the GUI never run

LOM first (targets to layout)

done for one template

Xmon + lambda/4 on one feedline

Milestones of the original plan

M

Milestone

State

M0

environment and skeleton

done; CI on GitHub Actions (lint, tests on Ubuntu and Windows, docs, release)

M1

PDK and mask export

done

M2

components with ports

done

M3

design model and builder

done

M4

live loop

done

M5

KLayout plugin

done except freeform shape to component

M6

simulation builder

done for Ansys; flip-chip two-face export missing

M7

Linux runner, Elmer, Palace

not done (export and queue code only)

M8

results store and circuit IR

done (SQLite rows of JSON; no Parquet or NPZ arrays)

M9

extraction and four adapters

done

M10

pyEPR, QuTiP, consensus

done; eigenmode never converged on this PC

M11

surrogates

partial (power-law calibration, no FEM sweeps)

M12

co-design optimiser

partial (emulated loop; one real FEM loop)

M13

feedback UI

done (dashboard, plugin tabs); per-instance hover values missing

What is left, ranked

#

Item

Kind

Size

1

Simulation PC. Solve the kept FEM set again on the current library (tutorial 0: 3 jobs, tutorial 4: 9, the subsets of 1-3) and keep it

never verified

M (3-4 h of AEDT)

2

Simulation PC. One converged FEM run (eigenmode + EPR, driven) for tutorials 0 and 4 on a 32 GB machine or the server

never verified

M

3

Simulation PC (needs the FEM results of 1-2). Calibrate the LOM against the kept FEM: hanger kappa per kind, TransmonCross claws, claw-to-ground C, chi

built, partial

M

4

LOM robustness: floating unconnected pins, QuCat at 8 or more junctions, whole-chip LOM of 21 qubits (simulation PC for the big chips)

not built

M-L

5

Simulation PC. Close the FEM feedback loop on target and document “at most N FEM iterations”

never verified

M

6

Simulation PC. Q3D nets above 8000 points: unite pieces in AEDT after import

not built

S-M

7

Moved to Future development goals below

8

Simulation PC (FEM sweeps). Co-design levels M11 and M12: FEM-sweep surrogates, HFSS sign-off with a bias record (Palace verify is a future goal)

not built

L

9

Simulation PC. Flip-chip two-face FEM export

not built

M

10

Docs: field reference for design.json and the other file formats, regenerate the CLI page, refresh design images and spec numbers

partial

M

11

Plots for convergence, S21 with fit, EPR participations, mesh extrapolation

being built

S

12

Moved to Future development goals below

13

Install from PyPI on a fresh PC and the plugin from Salt

wheel installs in a scratch env; PyPI and Salt not yet checked

S-M

14

One green fast test suite. “Tutorials saved clean” is settled: the tutorials keep their outputs

done on the release tree (2026-11-19, 3132 passed)

S

15

Mask and wafer assembly

being built (wafer assembly)

M

16

Router edge cases: node near a corner, waypoint between r and 2r (the tiny-leftover zigzag is done, fixed 2026-11-18)

built, partial

S

17

LOM model requests: pad-to-pad coupler line C, StarQubit pad arcs, IDC kappa at 1.2 µm gaps, Purcell filter between series caps, SQUID static flux bias

not built

M

18

Analysis extras: SQcircuit T1 and T-phi, QuTiP mesolve, TLS surface participations, asymmetric SQUID cells, NbN per element

not built

M each

19

Eigenmode mesh reuse for driven runs

experiment

S

20

Remaining designs (hyperbolic lattice, flip-chip TSV, quick demo mask, community specs, UT Austin wireless JPA)

not built

M

21

Dashboard Fab step; per-instance values in KLayout

per-instance values being built; Fab step not built

S

(S, M, L: small, medium, large.) Items marked Simulation PC need 32 GB of RAM or more; the checklist is the simulation PC page.

In progress

Being built and not yet verified:

  • Lumped-junction EPR from the LOM and from the eigenmode runs, and the convergence, S21 fit and EPR plots.

  • Wafer assembly of dies on 4 and 6 inch wafers.

  • Per-instance values in KLayout.

  • Custom components: a custom qubit and a spiralmon, a spiral capacitor and a tunable resonator.

  • Packaging: wheel with dependencies, qcc selftest, install scripts, the developer guide.

Future development goals

Not part of the simulation PC work; only the input files are written today.

Item

State

Size

Linux runner, Palace and Elmer solves, Palace EPR reader

never verified, not built

L

Sonnet and Elmer result readers and a qcc-sonnet test

not built

S-M

Palace verify step of the co-design levels

not built

M

Plugin

Open items:

  • Never solved from the GUI: an eigenmode run from the Simulate dialog, --aedt gui, a whole-chip simulation.

  • Live PCells: KLayout asks to save on close after the swap; shape-parameter instances stay static; importing the library inside KLayout needs approved environment changes.

  • No per-instance values on hover; the circuit picture is crowded on big chips; mirrored placement is not in the smokes.

  • Freeform shape to component capture is not built. Snapping a component to a line is not built.

  • Node snapping offers the start port’s angle only. The termination of an existing free end is edited in the JSON.

  • A node closer to a corner than the bend’s cut fails in KQC; only the preview warns.

  • One unreproduced click flake in 8 runs; sim groups are not drawable (Simulate makes them); half-built routes have no handles.

  • Xmon default junction “Trilayer Junction” (0.1 µm test finger) gives L_J of about 202 nH (kept for now).

  • Metal StarQubit: pins on solid metal, the coupler arm meets ground at the pin.

Library requests from the notebook rewrite

Collected while rewriting the v2 design notebooks (design notebooks); none is built. The numbering follows the order the requests came up: R1 to R11 and B-R1 to B-R11 from the first and second groups of notebooks (R1 to R4 apply to both), B4-R1 to B4-R5 from the amplifier notebooks. The notebook lists are not exhaustive.

Request

One line

Notebooks that need it

R1

Reuse a solved eigenmode mesh in a driven run (export_job(..., mesh_from=...))

every notebook with an eigenmode group

R2

After feedback the drawn junction keeps its fingers: set junction_params.lj_nH too, or flag it in the DRC

m51, a5, a6, qubit_test_chip, filled_qubit_test_chip, flipmon_test_chip, xmon_cap_sim, KQC SQUID chips

R3

Confirm codesign pairs a qubit with a lambda/2 readout that ends on a coupled pad, finger capacitor or tee

m51, a5, sqchip_em, demo, demo_twoface, munchqubits

R4

A multi-pair loop.iterate that keeps the FEM minus LOM offset

all notebooks with an iterate cell

R5

loop.analyse is slow on 17-100 qubits (every estimator on every pair); the notebooks pass group=

m53, demoqpu17, tut222, ibm_falcon_27q

R6

Feedback for qubits without a readout (qubit-only surrogate)

tut13, tut222, m54, tut443

R7, B-R2

epr_values with no readout per qubit (accept readouts={q: None}); a qubit-only form

tut13, m54, iswap1-iswap3, a7, two_transmon, simple

R8

Feedback bounds for designs whose targets are outside the PDK R_n window (pdk_lj_bounds(..., keep_outside=True))

a4, a7, m53, the five community chips, tutorial_5

R9

Add C-matrix mutual capacitances as branches when no route or coupling declares them

component_examples

R10

LOM model for TunableCoupler02 (series SQUID in the line); the coupled group fails in QuCat

iswap2

R11

Drop unreachable targets (e.g. alpha) in target_copy or loop.feedback

google_tc_cell, google_tc_2x2, ibm_bowtie_5q, ibm_falcon_27q

Driven with one port

driven_values fits notches only; add an S11 or transmission-peak fit (also B-R2, B-R8)

sqchip_em, ibm_bowtie_5q, gatemon_test_chip, shaping, simple

StarQubit junction port

The exporter makes no polygon for port 2, so no eigenmode job

m54

B-R1

LOM model for a lambda/2 in transmission between two IDCs (a wire today), so feedback has no pair

gatemon_test_chip

B-R3

Kept Q3D results should count for an L_J-only revision (a geometry hash beside the design hash)

every Results 5 that changes only L_J

B-R4

design_copy(...).save(path, name=...): a revision keeps the design name, so its FEM jobs collide

all notebooks with an iterate cell

B-R5

Check that an 8-resonator line’s eigenmode setup solves at least 8 modes

resonator_test_chip

B-R6

KQC qubit cells (Swissmon, DoublePads, …) lack qcc_category = "qubit", so layout_step counts them as “other”

KQC qubit chips

B-R7

Flip-chip groups simulate one face only (face_stack); a two-face stack is needed

demo_twoface, quality_factor_twoface, crossing_twoface

B-R9

Zoom views of one-PCell chips from refpoints, not coordinates read off pictures

the eight one-PCell test chips

B-R10

Snapshot without the ground grid (with_grid chips rebuild about 0.9 M holes per segment)

with_grid KQC chips

B-R11

Chip launchers know only SMA8, ARD24 and DC24

sample_holder_test

B4-R1

Keep a driven job’s Touchstone file with the kept results (or fit reflection resonances), so S11 can be read again later

the amplifier JPAs

B4-R2

A q_min argument for driven_from_eigenmode (its mode pick needs Q > 100; the inductive JPA’s Q is about 71)

twpa_sim_jpa

B4-R3

Check how fill treats the JTWPA’s grounded resonator pads

jtwpa_4wm

B4-R4

A one-supercell KI-TWPA extraction instead of meshing the whole 18.5 mm line

ki_twpa_serpentine

B4-R5

Decide on a chip-sized sim group for a driven S11 (a design change, so not made)

unsw_kipa