QuantumNovelty — Scout, Audit, and Falsify Quantum-Computing Research & Patents

📬 Quantum Novelty on Substack — everything new in quantum: fresh papers, tools, patents, and results that survive scrutiny.

Five ways to use it

Pick a workflow.

All five share the same skill catalog and audit-and-falsify framework. Pick by what you need today: broad idea scouting, paper writing, paper review, patent examination, patent drafting, or prosecution prep.

A

Scout a new quantum topic

Start from a topic, seed PDF, or notes file and get a grounded novelty report with references, quote evidence, and a ranked idea ledger.

Surface literature live — CrossRef, arXiv, Semantic Scholar, and optional Serper produce a baseline catalog and synthesis.
Acquire arXiv PDFs — bounded PDF download creates a run-local corpus so the scout can show where it looked.
Build a source KB — seed PDFs, markdown, text, JSON, BibTeX, and optional DOCX are indexed into an offline local KB.
Substantiate claims — every candidate idea gets exact quote evidence, citations, and a claim ledger before ranking.
Report the run — scout_report.md, scout_references.bib, claim_ledger.md, arxiv_corpus/, source_kb/, and substantiation/.

$ chain/run.sh --pipeline scout \
    --topic "Using AI for designing superconducting quantum chips" \
    --source-file seed_paper.pdf \
    --scout-arxiv-max-downloads 10 \
    --llm codex

B

Write a new quantum paper

From a research question to a submission-ready manuscript with audit script attached.

Surface the literature — CrossRef + arXiv + Semantic Scholar feed a Pareto-shaped baseline catalog of current published methods.
Drive a discovery loop — the LLM proposes ansatz / circuit candidates; a strict-domination archive at calibrated ε keeps the non-dominated points.
Falsifiability check — cross-LLM amplitude prediction across two distinct vendors; predictions persisted to disk before truth is computed.
Novelty audit — augmented-baseline merge surfaces rediscoveries the LLM didn't catch; honest negatives become a required Failure Modes section.
Draft the paper — venue-aware section ordering, abstract limits, and required statements per the 11-journal registry.
Review your own draft — 5-voice panel + 11 quantum-CS-specific fallacy checks + Stage-6 CQE before submission.

$ chain/run.sh --pipeline full-pipeline \
    --topic   "LLM-driven UCCSD pruning on H2O" \
    --hamiltonian H2O_4e_4o_8q \
    --journal     npj-quantum-information \
    --quantum-lib mlxq

C

Review an existing paper

Point QuantumNovelty at any published or preprint paper. Get a real reviewer report.

Deep research review — the 8-item audit-and-falsify checklist (Augmented baselines / Strict-domination / Recompute-from-raw / Wilson CIs / Cross-LLM / Honest negatives / Simulator precision / Auditable claims), with PASS / PARTIAL / FAIL per item.
5-voice reviewer panel — EIC + R1 Physics + R2 Novelty + R3 Evidence + Devil's Advocate. Each writes ≥4 paragraphs; the EIC produces a numbered must-fix list.
Logical-fallacy scan — standard fallacies plus 11 quantum-CS-specific (cherry-picked-baseline, ad-hoc-precision-floor, simulator-laundering, pareto-cherry-picked-axes, cross-llm-theatre, …).
Stage-6 CQE — 6-dimension Collaboration Quality Evaluation with geometric-mean composite. A 30/100 on one dimension cannot be averaged away by 95s on the other five.

$ chain/run.sh --pipeline chat \
    --prompt "Review this paper" \
    --paper draft.pdf \
    --journal npj-quantum-information

D

Examine or prosecute a quantum patent

Point QuantumNovelty at a Google Patents URL, publication number, or saved file. Patent examiners get claim‑by‑claim statutory triage; patent attorneys get prior-art, validity, amendment, and Office Action response prep. The output is a simulated USPTO Office Action with structured evidence, claim-compliance review, full-application review, and a legal-review disclaimer.

Load the case — Google Patents URL / publication number / saved file → examiner-ready text, bibliographic metadata, claim count, and kind code.
Frame the posture — A1 published application is treated as under examination; B1/B2 issued patent is treated as validity-risk and response-planning work.
Surface prior art — the deep-research stage pulls candidate anticipating and rendering-obvious references for the claim set, with search notes attorneys or examiners can inspect.
Run the USPTO panel — six voices: Primary Examiner (§ 101 Alice/Mayo), § 102 anticipation, § 103 obviousness (KSR), § 112 enablement / written-description / definiteness, Quantum Technical Specialist, and Supervisory Patent Examiner.
Check claim compliance — USPTO § 112(b) definiteness / antecedent basis / structure, EPO Art. 84 clarity / support / two-part form, and PCT clarity checks via --filing-standard.
Review the full application — specification adequacy, formalities, abstract/title/drawings, required sections, enablement, written-description support, and filing handoff items.
Gate the Office Action — zero-LLM-cost _office_action.json: disposition (allowance / non-final / final), rejected_claims, and rejections_by_statute {101 / 102 / 103 / 112}, parsed from the reconciled rejection block.
Hand off the work product — PATENT_REPORT.pdf, markdown stage outputs, and structured JSON support examiner search notes, attorney response drafts, claim-amendment strategy, and invalidity-risk memos.

$ chain/run.sh --pipeline patent-audit \
    --patent "https://patents.google.com/patent/US…A1/en" \
    --art-unit "AU 2128 (CPC G06N10/00)" \
    --filing-standard multi \
    --llm claude

E

Draft a quantum patent application

Start from an invention disclosure or short topic. QuantumNovelty drafts a filing package for USPTO, EPO, PCT, or multi-standard review: claims, specification, abstract, drawings plan, compliance notes, and filing handoff checklist.

Intake the disclosure — technical problem, quantum modality, control stack, implementation variants, and inventor-confirmation gaps.
Plan prior-art search — keywords, CPC/IPC classes, assignee/inventor angles, MPEP/EPC/PCT rule checks, and targeted documents to retrieve later.
Draft claims — system and method claims by default; computer-readable-medium claims only when the disclosure supports software/control logic.
Draft the application body — title, abstract, background, summary, brief description of drawings, detailed description, examples, and embodiments.
Run compliance review — USPTO § 112(b), EPO Art. 84/two-part form, PCT clarity/support, and full-application formalities under the selected standard.
Package handoff — filing_package.md plus a deterministic manifest for attorney review.

$ chain/run.sh --pipeline patent-draft \
    --disclosure invention_disclosure.md \
    --filing-standard uspto \
    --llm claude

RAG + knowledge bases

Evidence retrieval is a first-class workflow.

QuantumNovelty exposes a portable local KB for exact-quote evidence, and it interoperates with the broader ScienceSkills RAG stack where BGE-M3/Regiment indexing is the normal deep-research path.

QN native

Portable local KB

skills/quantum_kb creates file-backed KBs under quantum-kb/. It ingests PDF, Markdown, TeX, BibTeX, JSON, text, and optional DOCX, then writes plain inspectable index files.

Retrieval: BM25 sparse postings, deterministic hashed vectors, quantum-domain query expansion, metadata boosts, and source-diversity filtering.
Artifacts: search_results.json, quotes_for_prompt.txt, claim_evidence.md, citations.md, quote_fidelity.json, and emma_parity_report.json.
Use case: offline/local quote grounding, deterministic smoke tests, Emma-style perspective reports, and scout substantiation without a server.

SS full stack

BGE-M3 RAG through ScienceSkills

The ScienceSkills deep-research scout provides the heavier RAG path: run-local PDF KBs and per-idea substantiation KBs indexed through Regiment with BGE-M3 by default.

PDF KB mode: --scout-pdf-kb-only --scout-pdf-kb-index-backend bge-m3 downloads/fetches documents, extracts text, indexes them, and runs quote search.
Per-idea KB mode: --scout-idea-kb --scout-idea-kb-index-backend bge-m3 builds a separate evidence pack for each final ranked idea.
Runtime: bin/science_env.sh prefers the DSS-root venv and local HuggingFace cache; SCIENCESKILLS_HF_ONLINE=1 opts into refreshing models.

# QN native portable KB/RAG
$ skills/quantum_kb/run.sh create --kb quantum_papers --name "Quantum Papers" --default
$ skills/quantum_kb/run.sh ingest --kb quantum_papers --source papers/
$ skills/quantum_kb/run.sh index --kb quantum_papers --purge
$ skills/quantum_kb/run.sh substantiate --kb quantum_papers --claim "The scout idea is novel against current QEC decoder baselines."

# ScienceSkills BGE-M3 PDF KB
$ cd artifacts/code/ScienceSkills/agents/workflows/ai_scientific_paper_workflow_chain
$ ./run.sh --pipeline scout --topic "AI-assisted superconducting chip design" \
    --scout-pdf-kb-only --scout-pdf-kb-index-backend bge-m3

# ScienceSkills per-idea BGE-M3 substantiation KBs
$ ./run.sh --pipeline scout --topic "AI-assisted superconducting chip design" \
    --scout-idea-kb --scout-idea-kb-index-backend bge-m3

The catalog

Twenty-five skills. One contract. Deterministic gates included.

Every skill is a directory with a SKILL.md, a run.sh, and a Python driver. Drop a new one in — the chain discovers it on the filesystem walk.

novelty_audit

marquee

The audit-and-falsify framework. Augmented-baseline merge → strict-domination at ε_abs=10⁻¹², ε_rel=10⁻⁹ → recompute-from-raw → Wilson 95 % CIs → honest-negatives enforcement → re-runnable audit_claims.py.

quantum_scout

default discovery

SS-like topic scout: live literature, bounded arXiv PDF acquisition, run-local source KB, exact quote substantiation, claim ledger, references, and ranked novel avenues.

quantum_kb

local RAG

Offline-first local KB: create, ingest, index, search, substantiate claims, run grounded reviews, and generate Emma-style perspectives with verified quotes and citations.

deep_research

7 modes

full · quick · systematic-review · socratic · fact-check · lit-review · review

quantum_paper

10 modes

full · plan · outline-only · revision · revision-coach · abstract-only · lit-review · format-convert · citation-check · disclosure

quantum_reviewer

6 modes

full (EIC + R1 + R2 + R3 + Devil's Advocate) · quick · guided · methodology-focus · re-review · calibration

patent_reviewer

USPTO panel

6-voice examining unit → an Office Action for examiner and attorney workflows. Primary (§ 101 Alice/Mayo) · § 102 anticipation · § 103 obviousness (KSR) · § 112 enablement/definiteness · Quantum Technical Specialist · SPE. Also runs claim-compliance and full-application review under USPTO/EPO/PCT standards.

patent_prior_art

real references

Generates patent-search queries, runs keyless Google Patents/DDGS searches, and writes prior_art_refs.json for grounding §102/§103 analysis in named prior art.

patent_drafter

guided drafting

Invention disclosure → quantum-patent filing package: claims, specification, abstract, drawing plan, USPTO § 112(b) / EPO Art. 84 / PCT compliance review, formalities checklist, and attorney handoff.

logical_fallacies

+11 quantum-CS

cherry-picked-baseline · ad-hoc-precision-floor · conflated-regimes · active-space-handwave · hardware-irrelevant-comparison · asymptotic-only-claim · unit-inflation · simulator-laundering · mapping-by-convenience · pareto-cherry-picked-axes · cross-llm-theatre

cross_llm_prediction

Falsifiability rubric. Refuses single-vendor "consensus". Predictions persisted before truth.

pareto_explorer

Deprecated narrow ansatz/Pareto path for fixed-Hamiltonian reproductions. Use quantum_scout or --pipeline scout for broad novelty discovery.

ablation_designer

4 standard axes: LLM-mutator on/off · commutation-hint · Pareto seeding · cross-vendor.

literature_surfacer

CrossRef + arXiv + Semantic Scholar live HTTP. Emits the Pareto-shaped baseline catalog.

book_acquirer

Anna's Archive client for citations not on arXiv or CrossRef.

process_summary

Stage-6 CQE. Geometric-mean composite of Novelty Rigour · Reproducibility · Methodological Rigour · Falsifiability · Domain Depth · Communication.

evidence_ledger

Zero-LLM permitted-facts ledger plus post-review audit for unanchored citations, quotes, paper claims, and numeric attributions.

requirements_judge

Claim-vs-evidence judge that emits allowed and forbidden claims, with conservative downgrade/reject behavior when the manifest is inconsistent or unparsable.

chat

Natural-language frontend. Pattern-first dispatch; LLM fallback. "Review this paper" → the right skill + mode + flags.

audit_falsify

Strict-domination primitives surfaced so you can compose custom audit pipelines.

journals + quantum_libs

Registries: 11 venues (npj-QI, PRX-Quantum, Quantum, PRL, PRA, Nature Comms…) + 7 libraries (Qiskit, PennyLane, QuTiP, mlxq, Cirq, OpenFermion, no-code).

The chain

Six stages. Resumable. Honest about what didn't run.

1

Scout + literature

Multi-source live search, bounded arXiv PDF acquisition, and optional source-file ingestion.

2

Local evidence KB

Run-local lexical/hashed-vector KB writes exact quote packets, citations, and claim evidence.

3

Novelty audit

Augmented-baseline merge. Ratio recompute. Wilson CIs. Honest-negatives gate.

4

Draft + cross-LLM

Venue-aware draft. Falsifiable amplitude rubric across two vendors.

5

Review panel

5 voices. Quantum-CS fallacy taxonomy applied. Verdict + must-fix list.

6

Process summary

6-dim Collaboration Quality Evaluation. Geometric mean. Honest about gaps.

Resumable

Re-run with the same --outdir: stages whose output exists are skipped. --force re-runs everything.

Token + USD ledger

Every LLM call records input_tokens + output_tokens + cost_usd + model_id via claude --output-format json. Estimated counts (codex) are flagged honestly.

Backend-honest

If a backend fails the call surfaces as an error. No silent fallback. backend_actually_used is written for Claude, Codex, Kimi/Moonshot, and explicit Anthropic API runs.

Evidence-visible scout

The scout writes global_literature/, arxiv_corpus/, source_kb/, substantiation/, scout_manifest.json, and scout_quality.json so users can see where it looked.

Retrieval pre-flight gate

Known-answer probes run before every novelty verdict. Recall < 0.67 downgrades verdicts to "(indicative)" and records the original in verdict_ungated. Skip with --skip-retrieval-preflight.

282-test zero-model suite

All 282 tests run without LLM calls. New modules cover novelty-audit gates, patent-reviewer defect detection, cross-LLM vendor guard, and call_llm resilience/fallback paths.

Patent chain

Patent-audit mirrors the paper flow, but claim-by-claim.

The patent path is built for patent examiners and patent attorneys who need a structured workbench for quantum inventions: prior-art search, statutory issue spotting, Office Action-style reasoning, and a report that preserves every stage output.

1

Patent ingest

URL, publication number, or saved file becomes a reproducible extracted markdown record with claims and metadata.

2

Prior-art surface

Deep-research review over the claim set: anticipating references, obviousness combinations, and search notes.

3

Examiner panel

§§101/102/103/112 voices plus quantum specialist and SPE disposition, reconciled into an Office Action block.

4

Compliance review

Claim clarity and structure under USPTO §112(b), EPO Art. 84, PCT clarity/support, plus full-application formalities.

5

Fallacy scan

Quantum-CS overclaiming, operability leaps, simulator-laundering, and claim/spec reasoning gaps.

6

Report + handoff

_office_action.json, stage markdown, CQE summary, and PATENT_REPORT.pdf for examiner or attorney review.

Worked examples — in the repo

Real papers. Real reviewers. Real cost ledger.

Every example here lives at examples/end_to_end/ and contains the input paper, every stage's LLM output, the backend markers (token counts + USD), and the compiled PDF.

⚠ Disclaimer. All reviews below are generated end-to-end by AI, for academic and demonstration purposes only. We make no claims about the papers under review — four passed real peer review at their journals and one is a public arXiv preprint; nothing here is criticism of the authors or a substitute for human peer review. The combined five-paper report is REVIEWS.pdf.

PRX Quantum

LCU-Trotter — Trotter error compensation with LCU

Zeng, Sun, Jiang, Zhao · arXiv:2212.04566 · PRX Quantum 6, 010359

Panel 6.67/10, minor-revisions; 5 fallacy findings. The physics voice traced boundary-condition gaps in the lattice analysis; the Devil's Advocate pressed the observable-estimation-vs-coherent-simulation framing. Full claude run, $1.94 real USD.

Read the report (PDF)

npj-QI

Flow-VQE — Generative flow-based VQE warm-start

Zou, Rahm, Kockum, Olsson · arXiv:2507.01726

Panel 6.0/10, major-revisions; 7 fallacy findings from the quantum-CS taxonomy (cherry-picked-baseline, pareto-cherry-picked-axes, …). Full claude run, $0.97 real USD.

Read the report (PDF)

Quantum · QML

HW-QML — Hamming-weight preserving circuits for ML

Monbroussou, Mamon, Landman, Grilo, Kukla, Kashefi · arXiv:2309.15547 · Quantum 9, 1745

The quantum-machine-learning entry: trainability, expressivity, barren-plateau analysis of subspace-preserving ansätze. Panel 5.5/10, major-revisions; 6 fallacy findings. Full claude run, $1.63 real USD.

Read the report (PDF)

PRX Quantum · QML

QCNN — Quantum CNNs are classically simulable

Bermejo, Braccia, Rudolph, Holmes, Cincio, Cerezo · arXiv:2408.12739 · PRX Quantum 7, 020304

The QML dequantization landmark: Pauli-path simulation of QCNNs. Panel 6.0/10, major-revisions; 7 fallacy findings. Full claude run, $1.36 real USD.

Read the report (PDF)

preprint · hardware

Majorana tetron — 20-second parity lifetime

Microsoft Quantum · arXiv:2606.03884 · preprint

The hardware entry — first non-gate-model paper through the pipeline, and the showcase for the full verification layer: argument-structure audit, deterministic claims registry, 16-point disclosure audit, and a paragraph-anchored revision plan. Panel 6.33/10, major-revisions; the Devil's Advocate flagged the single-wire scope vs the multi-tetron-array claims. Full claude run, $1.78 real USD across 8 stages.

Read the report (PDF)

scaffolded

Three quantum avenues from a blank page

Surface code · QML kernels · QAOA on MaxCut

Three different topics, three target journals (PRX Quantum / npj-QI / Quantum), three quantum libraries (Qiskit / PennyLane / QuTiP). One drafted with codex when claude hit a nested-CLI collision (since root-caused — nested calls now disable tool use). All three got reject / major-revisions — because they were scaffolded with no real experiments. The reviewer panel is brutal about that, by design.

Open the run

vs. ARC and ARS

What QuantumNovelty adds.

QN is a peer of AutoResearchClaw and academic-research-skills. We cherry-picked the good parts and added a quantum-specific spine. The bottom line first — all three frameworks were run head-to-head on the same PRX Quantum paper (LCU-Trotter, arXiv:2212.04566) with the same backend; the measured rows come from that run (full 40-page comparison PDF):

	QN (QuantumNovelty)	ARS (academic-research-skills)	ARC (AutoResearchClaw)
Review steps	4 default stages (+3 opt-in: novelty-audit, cross-llm, synthesizer → up to 7)	7 agents in 3 phases	2 review stages (of a 23-stage generative pipeline)
Reviewer "agents"	5 voices in one LLM call (EIC + Physics + Novelty + Evidence + Devil's Advocate)	7 independent LLM calls (field-analyst, EIC, methodology, domain, perspective, Devil's Advocate, synthesizer)	1 call simulating Reviewer A + B
Logical fallacies	✓ dedicated stage — the only one with a domain taxonomy: 11 quantum-CS fallacies (simulator-laundering, cherry-picked-baseline, cross-llm-theatre, …) + standard	partial — folded into Devil's Advocate + methodology reviewer	✗ (fabrication / evidence-consistency check instead)
Devil's Advocate	✓ Voice 4 of the panel	✓ dedicated agent	✗
Numeric quality gate	✓ deterministic `_quality_gate.json` (ARC's shape, zero LLM cost)	✗	✓ the original (`score_1_to_10` + threshold)
Editorial decision package	✓ opt-in `--with-synthesizer` (+ CONSENSUS-0 tags for fallacy-only findings)	✓ Phase-2 synthesizer (consensus tags, roadmap, response template)	partial (`required_actions` list)
Measured cost (same paper)	4 calls, $2.15 real USD	7 calls, $3.10 real USD	2 calls, $0.66 real USD
Quantum-specific?	✓ the only one	✗ general academic	✗ general (ML-leaning)

All numbers are from a single all-claude run (Claude Code CLI, exact per-call USD). On LCU-Trotter the three architectures landed in the same revisions band — QN gate 6.0/10 (major revisions), ARC gate 7/10 (accept with minor revisions), ARS editorial decision (Minor Revision); QN's quantum-CS-specific checks make it the strictest. The concrete differences in detail:

Quantum-specific pipeline

The audit-and-falsify framework targets the failure modes that appear in quantum-computing papers: simulator precision artefacts, complex64 vs float64 swaps, Pareto cherry-picking on convenient axes, JW/BK/parity mapping chosen for cosmetic gate counts. ARC's gate stack is venue-agnostic; QN's is quantum-shaped.

Tested on real published papers

Run end-to-end on Flow-VQE (arXiv:2507.01726, npj-QI) and LCU-Trotter (arXiv:2212.04566, PRX Quantum) with the PDF report shipped in examples/ — plus HW-QML, QCNN, and the Majorana-tetron preprint, and three scaffolded avenues across different libraries and venues.

Used to generate a new paper

Not just review. The quantum_paper skill with 10 modes (full / plan / outline-only / revision / revision-coach / abstract-only / lit-review / format-convert / citation-check / disclosure) takes a topic to a venue-formatted draft. cross_llm_prediction lets you bake falsifiability into the paper.

Quantum-CS logical fallacies

11 fallacies that appear in quantum-CS work specifically — not in the general taxonomy. Cherry-picked-baseline. Ad-hoc-precision-floor. Simulator-laundering. Mapping-by-convenience. Pareto-cherry-picked-axes. Cross-llm-theatre. The taxonomy was extracted from real reviewer concerns raised against the framework's motivating study (paper in development).

Venue + library selectors

11-journal registry (npj-QI · PRX Quantum · Quantum · PRL · PRA · PR Applied · Nature Comms · Comms Physics · QST · Physics Letters A · IEEE TQE) with abstract limits, section orderings, required statements. 7-library registry (Qiskit · PennyLane · QuTiP · mlxq · Cirq · OpenFermion · no-code) emitting library-specific code skeletons.

Token + USD ledger in the PDF

Every LLM call's input/output tokens, model snapshot ID, USD cost, and elapsed time get aggregated into the comparison PDF. Two-paper analysis cost $<X> on claude; the PDF shows you the receipt. Codex / estimated rows flagged honestly.

Nested-CLI isolation playbook

QN scrubs CLAUDE_CODE_* and ANTHROPIC_* env from every subprocess, runs from a neutral cwd, passes --no-session-persistence. Hardened against the "tool_use ids must be unique" 400 that bites nested claude calls. When isolation isn't enough, QN refuses to silently swap backends.

Stage-6 CQE with geometric mean

Six dimensions composed by geometric (not arithmetic) mean. A run that scored 95 on five dimensions and 30 on Falsifiability composites at 67, not 84. The publication-blocking weakness cannot be averaged away.

From ARC gate-stack pattern · citation-integrity · compilation-quality · paper-verification · quality-gate · research-decision · knowledge-archive

From ARS modular skill-as-folder pattern · multi-mode skill drivers · literature-surfacer triad (CrossRef + arXiv + Semantic Scholar)

New in QN quantum_scout · quantum_kb · patent_prior_art · novelty_audit · audit_falsify primitives · quantum-CS fallacy taxonomy · cross-LLM falsifiability rubric · venue + library registries · token + USD ledger · Stage-6 CQE · retrieval pre-flight gate · assumption manifests · patent-panel calibration (5/9 granted, 17.6→1.8 rejected claims) · 282-test zero-model suite

Quick start

Pick the path that fits your setup.

All paths get you to the same place: local QuantumNovelty skills running through Claude Code CLI by default, Codex CLI for OpenAI-side checks, or Kimi/Moonshot when you want a direct Anthropic-compatible HTTP backend.

1

Option A — local checkout

$ git clone <repo> QuantumNovelty
$ cd QuantumNovelty
$ python -m venv .venv && source .venv/bin/activate
$ pip install -e .

2

Option B — use your CLI backend

$ which claude   # default: Claude Code CLI (subscription)
$ claude --version
$ which codex    # optional: OpenAI/Codex run
$ export KIMI_ENV_FILE=/path/to/kikm.sh  # optional Moonshot/Kimi

3

Option C — optional patent data connectors

$ # QN works with Google Patents URLs out of the box.
$ # Add external patent-data MCPs separately when you need
$ # broader USPTO, PTAB, MPEP, EPO, or worldwide data.

4

Verify

$ chain/run.sh --list-skills
$ chain/run.sh --list-stages

$ python -m pytest tests/test_smoke.py

What can I do?

Real patent workflows.

Goal	Run this	Output
Scout a new research avenue	`chain/run.sh --pipeline scout --topic "Using AI for designing superconducting quantum chips" --scout-arxiv-max-downloads 10`	Live literature cards, bounded arXiv PDF corpus, optional source KB, exact quote substantiation, claim ledger, references, and scout report.
Find patentability issues	`chain/run.sh --pipeline patent-audit --patent US10614371B2 --filing-standard multi`	Prior-art surface, six-voice Office Action, JSON/PDF report, fallacy scan, CQE summary.
Ground §102/§103 in named prior art	`skills/patent_prior_art/run.sh --topic "quantum error correction decoder" --output runs/prior_art`	`prior_art_refs.json` for `patent_reviewer --prior-art`, plus markdown search notes.
Check claims for compliance	`skills/patent_reviewer/run.sh --mode full --patent patent.html --filing-standard multi`	USPTO §112(b) definiteness / antecedent-basis / structure issues and EPO Art. 84 clarity / two-part-form issues.
Review a full application	`--filing-standard uspto`, `epo`, `pct`, or `multi`	Specification adequacy, enablement, written-description support, formalities, required sections, abstract/title/drawings.
Draft from scratch	`chain/run.sh --pipeline patent-draft --disclosure invention.md --filing-standard uspto`	Guided quantum-patent filing package with claims, spec, abstract, drawings plan, compliance review, and handoff checklist.
Review a quantum paper	`chain/run.sh --pipeline paper-audit --paper paper.pdf --journal npj-quantum-information`	Research review, 5-voice reviewer panel, claims registry, fallacy scan, and CQE summary.

Requirements

claude CLI (default) or codex CLI
Python ≥ 3.11
pdftotext (poppler) and pdflatex (any TeX distribution)
Optional: KIMI_ENV_FILE for Kimi/Moonshot; ANNAS_ARCHIVE_KEY for book downloads; SERPER_KEY for Google Scholar; external patent-data MCPs for broader USPTO/PTAB/MPEP/EPO/worldwide sources

Scout, audit, and falsify quantum‑computing research & patents.

Pick a workflow.

Scout a new quantum topic

Write a new quantum paper

Review an existing paper

Examine or prosecute a quantum patent

Draft a quantum patent application

Evidence retrieval is a first-class workflow.

Portable local KB

BGE-M3 RAG through ScienceSkills

Twenty-five skills. One contract. Deterministic gates included.

novelty_audit

quantum_scout

quantum_kb

deep_research

quantum_paper

quantum_reviewer

patent_reviewer

patent_prior_art

patent_drafter

logical_fallacies

cross_llm_prediction

pareto_explorer

ablation_designer

literature_surfacer

book_acquirer

process_summary

evidence_ledger

requirements_judge

chat

audit_falsify

journals + quantum_libs

Six stages. Resumable. Honest about what didn't run.

Scout + literature

Local evidence KB

Novelty audit

Draft + cross-LLM

Review panel

Process summary

Resumable

Token + USD ledger

Backend-honest

Evidence-visible scout

Retrieval pre-flight gate

282-test zero-model suite

Patent-audit mirrors the paper flow, but claim-by-claim.

Patent ingest

Prior-art surface

Examiner panel

Compliance review

Fallacy scan

Report + handoff

Real papers. Real reviewers. Real cost ledger.

LCU-Trotter — Trotter error compensation with LCU

Flow-VQE — Generative flow-based VQE warm-start

HW-QML — Hamming-weight preserving circuits for ML

QCNN — Quantum CNNs are classically simulable

Majorana tetron — 20-second parity lifetime

Three quantum avenues from a blank page

What QuantumNovelty adds.

Quantum-specific pipeline

Tested on real published papers

Used to generate a new paper

Quantum-CS logical fallacies

Venue + library selectors

Token + USD ledger in the PDF

Nested-CLI isolation playbook

Stage-6 CQE with geometric mean

Pick the path that fits your setup.

Option A — local checkout

Option B — use your CLI backend

Option C — optional patent data connectors

Verify

Real patent workflows.

Requirements

Scout, audit, and falsify
quantum‑computing research & patents.