diff --git a/_data/blogs.yml b/_data/blogs.yml
index 51fcc20..da5d081 100644
--- a/_data/blogs.yml
+++ b/_data/blogs.yml
@@ -1,5 +1,5 @@
 - link: /blog/agents-last-exam
-  img: /blog/agents-last-exam/main-results.png
+  img: /blog/agents-last-exam/cover.png
   title: "Agents' Last Exam"
   date: "2026-06-15"
   description: "A rolling benchmark measuring whether AI agents can perform economically valuable work across 55 occupations. We evaluated Fable 5, GPT-5.5, Composer 2.5, and other frontier systems on 1,500+ expert-sourced tasks. Today's agents solve a meaningful fraction of professional work, but on ALE's hardest tier every frontier agent we tested, including Fable 5, scored 0%."
diff --git a/blog/agents-last-exam/cover.png b/blog/agents-last-exam/cover.png
new file mode 100644
index 0000000..09ecc0b
Binary files /dev/null and b/blog/agents-last-exam/cover.png differ
diff --git a/blog/agents-last-exam/index.md b/blog/agents-last-exam/index.md
index c050104..c6670f9 100644
--- a/blog/agents-last-exam/index.md
+++ b/blog/agents-last-exam/index.md
@@ -24,6 +24,8 @@ June 2026
 Also on <a href="https://www.linkedin.com/pulse/introducing-agents-last-exam-ale-new-standard-evaluating-dawn-song-dntuc/" target="_blank">LinkedIn</a> and <a href="https://x.com/dawnsongtweets/status/2065095757988868190" target="_blank">X</a>
 </div>
 
+<img src="cover.png" alt="Agents' Last Exam: a benchmark spanning 55 occupations, with real-world agent pipelines across Manufacturing and Game Development" class="content-image" style="width: 100%; max-width: 100%; padding: 10px;">
+
 
 *Everyone says the latest AI agents will be "job-ready" soon, especially after the release of Fable 5 last week. But is that really the case?*