diff --git a/.github/workflows/gpy-tutorials.yml b/.github/workflows/gpy-tutorials.yml
index 8cd204c..b40de60 100644
--- a/.github/workflows/gpy-tutorials.yml
+++ b/.github/workflows/gpy-tutorials.yml
@@ -1,7 +1,6 @@
 # Execute GPy/*.ipynb against a pinned GPy (CIP-0003).
 #
-# GPy pin: 1.15.0 candidate (SheffieldML/GPy#1235, branch release/1.15.0).
-# Switch the install line to `gpy==1.15.0` once that release is on PyPI.
+# GPy pin: 1.15.0 on PyPI (SheffieldML/GPy#1235 merged, v1.15.0).
 # Do not float on GPy devel for the release-gate job.
 #
 # East Timor in GPyCrashCourse is loaded from datasets_mirror (PyPI pods 0.1.18
@@ -13,9 +12,8 @@
 # climin is installed so SVI.ipynb and optimizer-implementation.ipynb are
 # required (CIP-0003 skip only applies when the extra is absent).
 #
-# Do not add a "latest PyPI GPy" job until 1.15.0 is published. Current PyPI
-# is 1.14.2; these notebooks target 1.15, and GitHub still paints
-# continue-on-error jobs as failing checks.
+# execute-latest installs current PyPI GPy without a version pin. It matches
+# execute while 1.15.0 is latest, and becomes a canary after the next release.
 name: GPy tutorials
 
 on:
@@ -54,10 +52,10 @@ jobs:
         with:
           python-version: "3.11"
 
-      - name: Install GPy 1.15.0 candidate, plotting, Jupyter
+      - name: Install GPy 1.15.0, plotting, Jupyter
         run: |
           python -m pip install --upgrade pip
-          python -m pip install "gpy[plotting] @ git+https://github.com/SheffieldML/GPy.git@refs/pull/1235/head"
+          python -m pip install "gpy[plotting]==1.15.0"
           python -m pip install nbclient ipykernel jupyter_client ipywidgets pods seaborn pandas climin
           python -c "import GPy, climin; print('GPy', getattr(GPy, '__version__', '?')); print('climin', climin.__file__)"
 
@@ -74,3 +72,40 @@ jobs:
           name: executed-gpy-notebooks
           path: executed-gpy/
           if-no-files-found: ignore
+
+  execute-latest:
+    runs-on: ubuntu-latest
+    timeout-minutes: 90
+    env:
+      MPLBACKEND: Agg
+      MPLCONFIGDIR: ${{ github.workspace }}/.mplconfig
+      PYTHONUNBUFFERED: "1"
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+
+      - name: Install latest PyPI GPy, plotting, Jupyter
+        run: |
+          python -m pip install --upgrade pip
+          python -m pip install "gpy[plotting]"
+          python -m pip install nbclient ipykernel jupyter_client ipywidgets pods seaborn pandas climin
+          python -c "import GPy, climin; print('GPy', getattr(GPy, '__version__', '?')); print('climin', climin.__file__)"
+
+      - name: Register kernel
+        run: python -m ipykernel install --user --name python3
+
+      - name: Execute GPy tutorials
+        run: bash scripts/run_gpy_tutorials.sh --output-dir executed-gpy-latest
+
+      - name: Upload executed notebooks
+        if: failure()
+        uses: actions/upload-artifact@v4
+        with:
+          name: executed-gpy-latest-notebooks
+          path: executed-gpy-latest/
+          if-no-files-found: ignore
diff --git a/backlog/infrastructure/2026-10-24_gpy-tutorials-ci-runner.md b/backlog/infrastructure/2026-10-24_gpy-tutorials-ci-runner.md
index 24267e5..1059121 100644
--- a/backlog/infrastructure/2026-10-24_gpy-tutorials-ci-runner.md
+++ b/backlog/infrastructure/2026-10-24_gpy-tutorials-ci-runner.md
@@ -1,7 +1,7 @@
 ---
 id: "2026-10-24_gpy-tutorials-ci-runner"
 title: "Add nbclient runner and GitHub Actions for GPy tutorials"
-status: "In Progress"
+status: "Completed"
 priority: "High"
 created: "2026-10-24"
 last_updated: "2026-10-24"
@@ -25,7 +25,7 @@ tags:
 
 ## Description
 
-Add `scripts/run_gpy_tutorials.sh` (nbclient, `MPLBACKEND=Agg`, per-notebook timeout, climin skip policy) and a GitHub Actions workflow: Python 3.12, install GPy pinned to the 1.15.0 candidate (`gpy==1.15.0` once published; until then `release/1.15.0` or the #1235 SHA) plus `gpy[plotting]`, Jupyter (`nbclient`, `ipykernel`). Upload executed notebooks on failure. Point the README GPy tutorial section at the workflow badge.
+Add `scripts/run_gpy_tutorials.sh` (nbclient, `MPLBACKEND=Agg`, per-notebook timeout, climin skip policy) and a GitHub Actions workflow: Python 3.11, install GPy pinned to `gpy[plotting]==1.15.0` plus Jupyter (`nbclient`, `ipykernel`). Upload executed notebooks on failure. Point the README GPy tutorial section at the workflow badge. A second `execute-latest` job installs unpinned `gpy[plotting]`.
 
 Skip `SVI.ipynb` and `optimizer-implementation.ipynb` when climin is not installed, with the skip documented. `index.ipynb` has no required code. Fail the job if a required notebook raises.
 
@@ -35,6 +35,8 @@ Skip `SVI.ipynb` and `optimizer-implementation.ipynb` when climin is not install
 - [x] GitHub Actions workflow pins GPy and runs the script on this repository
 - [x] README GPy tutorial section shows the workflow badge
 - [x] Job does not use Plotly Cloud credentials
+- [x] `execute` installs `gpy[plotting]==1.15.0` from PyPI
+- [x] `execute-latest` installs current PyPI `gpy[plotting]`
 
 ## Implementation Notes
 
@@ -52,4 +54,4 @@ Run with cwd such that `import GPy` resolves the library, not this repo's `GPy/`
 
 Task created as Ready after CIP-0003 was accepted.
 
-Added `scripts/run_gpy_tutorials.py` / `.sh` and `.github/workflows/gpy-tutorials.yml` (Python 3.12, GPy from SheffieldML/GPy#1235, climin skip). Local execute against the installed GPy passed required notebooks (climin pair skipped). First Actions green run is still pending.
+Added `scripts/run_gpy_tutorials.py` / `.sh` and `.github/workflows/gpy-tutorials.yml` (Python 3.11, climin required). PR #24 `execute` went green against the GPy 1.15.0 git candidate. After v1.15.0 on PyPI, pin `execute` to `gpy[plotting]==1.15.0` and restore `execute-latest`. Marked Completed.
diff --git a/cip/cip0003.md b/cip/cip0003.md
index 3abe508..8081bc3 100644
--- a/cip/cip0003.md
+++ b/cip/cip0003.md
@@ -70,19 +70,17 @@ ahead of a green run.
 
 - **Command:** `nbclient` or `jupyter execute`, `MPLBACKEND=Agg`, per-notebook
   timeout, kernel from the job env.
-- **GPy pin:** install GPy from a version or git SHA recorded in the workflow
-  (`gpy==1.15.0` once published; until then the `release/1.15.0` branch or
-  #1235 SHA). Do not float on GPy `devel` for the release-gate run.
+- **GPy pin:** `gpy[plotting]==1.15.0` from PyPI (v1.15.0 published
+  2026-10-24). Do not float on GPy `devel` for the release-gate run.
 - **Python:** 3.11 (GPy 1.15 supports it; climin 0.1a1 still `import imp`,
   which Python 3.12 removed).
 - **Extras:** `gpy[plotting]` plus Jupyter (`nbclient`, `ipykernel`), `pods`,
   and `climin` so SVI/optimizer notebooks are required. East Timor in
   `GPyCrashCourse` is loaded from `lawrennd/datasets_mirror` until
-  `pods.datasets.east_timor` is on PyPI. Do not run a second job against
-  current PyPI GPy until 1.15.0 is published: 1.14.2 still uses the Plotly
-  matplotlib path these tutorials left, and `continue-on-error` still shows
-  as a failed GitHub check. After the release, switch the pin to
-  `gpy==1.15.0` (or add a latest job that is the same install).
+  `pods.datasets.east_timor` is on PyPI. A second Actions job
+  (`execute-latest`) installs current PyPI `gpy[plotting]` without a
+  version pin. While 1.15.0 is latest the two jobs match; after the next
+  GPy release, `execute-latest` is the canary.
 - **Where:** GitHub Actions **in this repository** (the files under test live
   here). GPy may later add a complementary job that checks out a **pinned SHA
   of this repo**; that is GPy's CIP, not this one.
@@ -100,7 +98,7 @@ Other directories stay unexecuted until a later CIP.
 2. GitHub Actions workflow: Python 3.12, install pinned GPy + extras, run the
    script, upload executed notebooks on failure.
 3. Record the GPy pin and climin skip policy in the workflow comments.
-4. First green run against the 1.15.0 candidate; comment on GPy#1235.
+4. First green run against the 1.15.0 candidate; comment on GPy#1235 after PyPI pin.
 5. Point README GPy tutorial section at the workflow badge.
 
 ## Backward Compatibility
@@ -119,7 +117,8 @@ check when they edit `GPy/` tutorials.
 - [x] Manual execute against GPy 1.15.0; inventory failures
 - [x] Accept CIP after review
 - [x] Runner script + Actions workflow
-- [ ] Green run on GPy 1.15.0 candidate
+- [x] Green run on GPy 1.15.0 candidate (PR #24)
+- [x] Pin `execute` to PyPI `gpy[plotting]==1.15.0`; restore `execute-latest`
 - [x] README badge
 - [ ] Close CIP