diff --git a/docs/blog/index.html b/docs/blog/index.html
index 732bed2..bccbf3b 100644
--- a/docs/blog/index.html
+++ b/docs/blog/index.html
@@ -31,11 +31,29 @@ <h1>Blog</h1>
                     PyArrow, pandas or NumPy. Measurements, not estimates.
                 </p>
 
-                <!-- DRAFT, not yet listed: parquet-in-aws-lambda-without-pyarrow.html
-                     Add its card here once reviewed. -->
+                <a class="post-card" href="lambda-cost-break-even.html">
+                    <h2>
+                        When rugo makes your Lambda cheaper, and when it
+                        doesn't
+                    </h2>
+                    <p class="post-meta">17 October 2026</p>
+                    <p>
+                        What rugo, PyArrow, DuckDB and Polars cost per million
+                        Lambda calls. rugo costs 15–35% less than PyArrow at
+                        every cold-start rate and, at 1024 MB, beats DuckDB and
+                        Polars once about one call in ten is cold.
+                    </p>
+                </a>
 
-                <!-- DRAFT, not yet listed: lambda-cost-break-even.html
-                     Follow-up to the Lambda post above; list it after that one. -->
+                <a class="post-card" href="parquet-in-aws-lambda-without-pyarrow.html">
+                    <h2>Reading Parquet in AWS Lambda without PyArrow</h2>
+                    <p class="post-meta">17 October 2026</p>
+                    <p>
+                        PyArrow takes 69% of Lambda's 250 MB package limit and
+                        no longer fits a direct upload. rugo takes 5.4%, and
+                        starts 1.5–2.2× faster than PyArrow, DuckDB or Polars.
+                    </p>
+                </a>
 
                 <a class="post-card" href="jsonl-logs-grep-vs-pandas-vs-rugo.html">
                     <h2>
diff --git a/docs/blog/lambda-cost-break-even.html b/docs/blog/lambda-cost-break-even.html
index 09db39d..91897b9 100644
--- a/docs/blog/lambda-cost-break-even.html
+++ b/docs/blog/lambda-cost-break-even.html
@@ -1,24 +1,4 @@
 <!doctype html>
-<!--
-  ============================ DRAFT — DO NOT PUBLISH ============================
-  Follow-up to parquet-in-aws-lambda-without-pyarrow.html, which is itself still
-  an unlisted draft. Publish that one first; this post depends on it.
-
-  Unresolved before this can go live:
-
-  1. Both posts now use the same run (benchmarks/lambda-cost, 17 October 2026)
-     and the same 38.6 MB file, so their numbers should agree. If either is
-     re-measured, re-measure both.
-
-  2. Publish the parent post first.
-
-  3. Check Lambda prices on the day of publishing (us-east-1 figures below).
-
-  All numbers here: measured 17 October 2026, rugo 0.5.4, pyarrow 26.0.0,
-  duckdb 1.5.6, polars 2.0.0, python3.12, us-east-1. The AWS stack is torn down;
-  re-run with benchmarks/lambda-cost.
-  ===============================================================================
--->
 <html lang="en">
     <head>
         <meta charset="UTF-8" />
@@ -48,7 +28,7 @@
                 <header class="post-header">
                     <h1>When rugo makes your Lambda cheaper, and when it doesn't</h1>
                     <p class="post-meta">
-                        Justin Joyce <span class="sep">·</span> [date]
+                        Justin Joyce <span class="sep">·</span> 17 October 2026
                         <span class="sep">·</span> 8 min read
                     </p>
                     <div class="post-tags">
diff --git a/docs/blog/parquet-in-aws-lambda-without-pyarrow.html b/docs/blog/parquet-in-aws-lambda-without-pyarrow.html
index d090cdc..afea36e 100644
--- a/docs/blog/parquet-in-aws-lambda-without-pyarrow.html
+++ b/docs/blog/parquet-in-aws-lambda-without-pyarrow.html
@@ -1,19 +1,4 @@
 <!doctype html>
-<!--
-  ============================ DRAFT — DO NOT PUBLISH ============================
-  Unresolved before this can go live:
-
-  1. Re-measured 17 October 2026 on rugo 0.5.4 (was 0.4.30, 14 August). The
-     Brotli and INT96 zero-row bugs are fixed: Brotli now raises a clear
-     "codec not supported" error, and INT96 timestamps read correctly. Re-check
-     the "What rugo cannot do" list against the current release on the day.
-
-  2. Publish before lambda-cost-break-even.html, which follows on from this post.
-
-  3. The benchmark AWS stack has been torn down. The harness and raw results are
-     in benchmarks/lambda-cost.
-  ===============================================================================
--->
 <html lang="en">
     <head>
         <meta charset="UTF-8" />
@@ -43,7 +28,7 @@
                 <header class="post-header">
                     <h1>Reading Parquet in AWS Lambda without PyArrow</h1>
                     <p class="post-meta">
-                        Justin Joyce <span class="sep">·</span> [date]
+                        Justin Joyce <span class="sep">·</span> 17 October 2026
                         <span class="sep">·</span> 12 min read
                     </p>
                     <div class="post-tags">