Complete pre-release sanity checks and repair hosted mathematics

.github/workflows/hosted-render.yml

@@ -0,0 +1,33 @@
+name: Inspect hosted mathematics
+on:
+  pull_request:
+    paths:
+      - '**.md'
+      - 'tools/inspect_hosted_render.py'
+      - '.github/workflows/hosted-render.yml'
+  workflow_dispatch:
+permissions:
+  contents: read
+jobs:
+  hosted-render:
+    runs-on: ubuntu-latest
+    timeout-minutes: 12
+    env:
+      RENDER_REF: ${{ github.event.pull_request.head.sha || github.sha }}
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          ref: ${{ env.RENDER_REF }}
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.13'
+      - run: python -m pip install playwright==1.55.0
+      - run: python -m playwright install --with-deps chromium
+      - run: python tools/inspect_hosted_render.py --ref "$RENDER_REF"
+      - uses: actions/upload-artifact@v4
+        if: always()
+        with:
+          name: hosted-render-${{ env.RENDER_REF }}
+          path: build/hosted-render/
+          retention-days: 14
+          if-no-files-found: warn

.github/workflows/verify.yml

@@ -20,4 +20,5 @@ jobs:
           cache: pip
       - run: python -m pip install -r requirements.txt
       - run: python check_repository.py
+      - run: python tools/test_release_checks.py
       - run: python verify.py

AGENTS.md

@@ -30,6 +30,8 @@ Follow `docs/coaching.md` as the single protocol. Every round has Revise → Lea
 
 Credit conditional variance, entropy-fluctuation speed limits, Hamiltonian averaging, reversible compression, concentration inequalities, and prior routing/initialization work. Add foundational sources to both `docs/related-work.md` and `references.bib`. Keep journal-target judgments and development diaries out of reader-facing science. Preserve assumptions, proof domains, source provenance, and load-bearing limitations when removing redundancy.
 
-Run `python check_repository.py` and `python verify.py`. The repository checker includes `tools/check_learning.py`, which validates the map, dependencies, source links, and learning entrypoints. The ten numerical suites remain separate. Replace tracked numerical results only after a relevant executed source change; furnishing alone does not justify new scientific provenance. Passing checks is not outside endorsement.
+Run `python check_repository.py`, `python tools/test_release_checks.py`, and `python verify.py`. The repository checker includes `tools/check_learning.py`, which validates the map, dependencies, source links, and learning entrypoints. Release checks also validate reading-page tables, equation tags, and tracked certificate integrity. The ten numerical suites remain separate. Replace tracked numerical results only after a relevant executed source change; furnishing alone does not justify new scientific provenance. Passing checks is not outside endorsement.
 
-Use GitHub dollar-delimited math with blank lines around displays, no display equations inside tables, and local navigation links. Inspect rendered README, learning pages, and representative math-heavy science pages for macro errors, missing subscripts, bra–ket rendering, and overflow. Test the actual GitHub renderer when accessible; identify a local MathJax preview as local, not as hosting verification. Report which checks actually ran.
+For mathematical pages, prefer GitHub fenced math blocks and its backtick-protected inline math syntax, with blank lines around display blocks and no displays inside tables. Use supported upright commands such as `\mathrm{Tr}` and brace starred subscripts as `_{\ast}`. This repository's hosted checks exposed Markdown consumption of backslashes and a rejected `\operatorname` macro; local MathJax success alone is insufficient. These are hosting-specific precautions, not universal restrictions on TeX. The [official syntax](https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/writing-mathematical-expressions) describes protected forms.
+
+Inspect rendered README, learning pages, and representative math-heavy science pages for macro errors, missing subscripts, bra–ket rendering, and overflow. Use the separate `Inspect hosted mathematics` workflow to check actual GitHub at a pinned commit and retain screenshots. A local MathJax preview is local, not hosting verification. Report which checks actually ran; blocked access or unfinished rendering is not a pass.

README.md

@@ -19,46 +19,46 @@ The single educational anchor is **John Watrous, *The Theory of Quantum Informat
 
 ## Main result
 
-Each module contains $n\ge1$ payload qubits and $s\ge0$ helpers. Every module-qubit–bus interaction has operator norm at most $J$; all contacts may act simultaneously. Internal operations confined to either module or the bus have zero charged duration. Both interfaces remain bounded. There are no measurements, resets, or extra initialized qubits.
+Each module contains $`n\ge1`$ payload qubits and $`s\ge0`$ helpers. Every module-qubit–bus interaction has operator norm at most $`J`$; all contacts may act simultaneously. Internal operations confined to either module or the bus have zero charged duration. Both interfaces remain bounded. There are no measurements, resets, or extra initialized qubits.
 
-For helpers independently prepared in the known state $\rho_p=(1-p)|0\rangle\langle0|+p|1\rangle\langle1|$, with $0\le p\le1/2$, the optimal fixed-error interaction time is
+For helpers independently prepared in the known state $`\rho_p=(1-p)|0\rangle\langle0|+p|1\rangle\langle1|`$, with $`0\le p\le1/2`$, the optimal fixed-error interaction time is
 
-$$
+```math
 T_{\epsilon,p}(n,s)
 =\Theta_\epsilon\!\left(
 \frac{n}{J\sqrt{(n+s)\{1+s[1-h_2(p)]\}}}
 \right).
-$$
+```
 
-For every fixed $0<\epsilon\le0.1$, the constants are uniform in $n,s,p$. The error is a complete-operation half diamond distance, including arbitrary reference-entangled payload/bus inputs and the fixed helpers' inaccessible preparation references. The physical helper density matrix returns exactly in the construction; remaining correlations and bus errors are included in the error budget. Compression, routing, and decompression share one invariant high-probability sector, so initial preparation error does not accumulate gate by gate.
+For every fixed $`0\lt \epsilon\le0.1`$, the constants are uniform in $`n,s,p`$. The error is a complete-operation half diamond distance, including arbitrary reference-entangled payload/bus inputs and the fixed helpers' inaccessible preparation references. The physical helper density matrix returns exactly in the construction; remaining correlations and bus errors are included in the error budget. Compression, routing, and decompression share one invariant high-probability sector, so initial preparation error does not accumulate gate by gate.
 
 The [full proof](docs/mixed-helpers.md) distinguishes zero error from fixed nonzero error. A correlated-state counterexample shows why entropy deficit alone is not a complete bandwidth invariant for arbitrary helper states.
 
 ## Exact endpoints
 
 On identical hardware with the same payload,
 
-$$
+```math
 T_{\mathrm{clean}}(n,s)
 =\Theta\!\left(\frac{n}{J\sqrt{(n+s)(s+1)}}\right),
-$$
+```
 
-$$
+```math
 T_{\mathrm{unknown}}(n,s)
 =\Theta\!\left(\frac{n}{J\sqrt{n+s}}\right).
-$$
+```
 
-These exact tasks restore helpers and an arbitrary bus, including reference correlations. The changed resource is the initial helper promise, not extra wiring. At $s=0$, the time is $\Theta(\sqrt n/J)$. Any full-rank mixed preparation has the unknown-helper scale at zero error; the fixed-error law is not uniform as $\epsilon\to0$.
+These exact tasks restore helpers and an arbitrary bus, including reference correlations. The changed resource is the initial helper promise, not extra wiring. At $`s=0`$, the time is $`\Theta(\sqrt n/J)`$. Any full-rank mixed preparation has the unknown-helper scale at zero error; the fixed-error law is not uniform as $`\epsilon\to0`$.
 
 ## Quantitative realization and scope
 
-After concentration into $r$ near-clean leader bits, balanced blocks need only
+After concentration into $`r`$ near-clean leader bits, balanced blocks need only
 
-$$
+```math
 K_{\mathrm{exact}}=2^{\lceil(n+s)/(r+1)\rceil-1}
-$$
+```
 
-exact interaction intervals per Ising gate. Synchronizing each block's translations suffices because the contact is an additive sum. For fixed nonzero polarization and $s=\Theta(n)$, this gives $O(n)$ total **exact contact intervals**. Global internal compression is still free in this theorem, and no efficient elementary implementation of every encoding is implied.
+exact interaction intervals per Ising gate. Synchronizing each block's translations suffices because the contact is an additive sum. For fixed nonzero polarization and $`s=\Theta(n)`$, this gives $`O(n)`$ total **exact contact intervals**. Global internal compression is still free in this theorem, and no efficient elementary implementation of every encoding is implied.
 
 Exact rational spectral certificates demonstrate finite-size advantages over every allowed unknown-helper protocol at the same complete-operation accuracy, not just over one selected implementation. The [resource analysis](docs/accuracy-and-resources.md) also shows that a fixed sum of contact norms removes the parametric initialization advantage, while a fixed quadratic contact budget retains it.
 

check_repository.py

@@ -8,6 +8,7 @@
 from urllib.parse import unquote, urlsplit
 from verify import SUITES
 from tools.check_learning import check_learning
+from tools.check_release import check_release
 
 ROOT = Path(__file__).resolve().parent
 CONTACT = 'This repository serves as a record of the work and a guide for the author’s self-directed learning. For discussion or potential collaboration, please contact Ruge Lin at [gogoko699@gmail.com](mailto:gogoko699@gmail.com).'
@@ -61,6 +62,7 @@ def main() -> None:
     assert keys and len(keys) == len(set(keys)), 'Missing or duplicate bibliography keys'
     assert set(keys).issubset({x.upper() for x in anchors((ROOT/'docs/related-work.md').read_text())})
     learning = check_learning(ROOT)
+    check_release(ROOT)
     print(f'PASS: {len(markdown)} reading files, {link_count} local links, {len(sources)} Python files, {len(SUITES)} verified suites, {len(keys)} references and {learning["milestones"]} milestones.')
 
 if __name__ == '__main__':