diff --git a/config/previously-recommended.md b/config/previously-recommended.md
index 728a46b..e174462 100644
--- a/config/previously-recommended.md
+++ b/config/previously-recommended.md
@@ -111,3 +111,20 @@ https://www.smashingmagazine.com/2025/07/design-patterns-ai-interfaces/
 https://mbrenndoerfer.com/writing/rag-architecture-retriever-generator-design-patterns
 https://www.youtube.com/watch?v=5ABcYSWPVr8
 >>>>>>> Stashed changes
+
+# Week 13 — 2026-06-14 (round 1 — retroactively logged; these were posted to Notion but never written to this log)
+https://www.anthropic.com/engineering/built-multi-agent-research-system
+https://aakashgupta.medium.com/building-multi-agent-systems-will-be-a-must-have-pm-skill-in-2026-30e52fd732cb
+https://www.lennysnewsletter.com/p/we-replaced-our-sales-team-with-20-ai-agents
+https://www.latent.space/p/andon
+https://www.microsoft.com/en-us/research/blog/magenticlite-magenticbrain-fara1-5-an-agentic-experience-optimized-for-small-models/
+https://arxiv.org/abs/2510.24937
+https://arxiv.org/abs/2603.02050
+
+# Week 13 — 2026-06-21 (round 2)
+https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents
+https://learn.microsoft.com/en-us/microsoft-copilot-studio/guidance/multi-agent-patterns
+https://arxiv.org/abs/2512.11979
+https://agentic-design.ai/patterns/ui-ux-patterns
+https://www.news.aakashg.com/p/practical-ai-agents-pms
+https://www.oneusefulthing.org/p/what-it-feels-like-to-work-with-mythos
diff --git a/logs/run-2026-06-21.md b/logs/run-2026-06-21.md
new file mode 100644
index 0000000..64c66ce
--- /dev/null
+++ b/logs/run-2026-06-21.md
@@ -0,0 +1,39 @@
+# Run Log — 2026-06-21
+Week: 13 (second pass — frontier still at Week 13; no items checked in Week 13 of the checklist, so the current-week pointer has not advanced since the 2026-06-14 run)
+Topic: Week 13 · Multi-agent orchestration (design lens)
+Idempotency guard: An existing "Weekly Picks — Week 13" page was found, but it was generated 2026-06-14 (7 days ago) — outside the 3-day window — so this run proceeded with all-new picks rather than skipping.
+Sources checked: ~22 (Microsoft Research RSS direct; 8 blog/Substack feeds attempted direct + via rss2json proxy; 16 web searches)
+Items collected: ~30 surfaced; ~12 serious candidates after dedup
+Items after hard filters: ~12 (excluded all URLs in the Weekly Checklist, previously-recommended.md, and the prior briefing pages Weeks 6–13 — notably last week's 7 Week-13 picks)
+Items after topic gate: ~10
+Must Read: 2 (score range 87–92)
+Recommended: 3 (score range 76–87) — includes 2 "Discovered" design-lane items folded into Recommended (the requested layout has no separate Discovered section)
+Wild Card: yes (trigger: contrarian — single frontier agent vs. multi-agent fleet)
+Discovered: 2 (arXiv "Internet of Agents / HAX" framework; agentic-design.ai pattern catalog)
+Dropped: low-trust SEO/listicle articles and videos that dominated web-search fallback results; arXiv 2606.18716 (couldn't verify content — WebFetch blocked)
+
+Final 7-slot diversity check (6 picks used; cap is 7):
+- Product Lane (≥2 required): 2 — Microsoft Copilot Studio multi-agent patterns; Aakash Gupta "Practical AI Agents for PMs"
+- Design Lane (≥2 required): 2 — "Designing the Internet of Agents (HAX)"; agentic-design.ai UI/UX patterns
+- Pure-engineer-audience (≤2 allowed): 1 — Anthropic "Effective context engineering for AI agents"
+- Wild Card (general audience): 1 — Mollick "What it feels like to work with Mythos"
+Diversity rule satisfied. Chose 6 over 7 deliberately (quality over quantity) — no 7th item cleared the bar without padding with low-trust content.
+
+Picks:
+1. ⭐ Anthropic — Effective context engineering for AI agents — 92 — https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents
+2. ⭐ Microsoft Learn — Multi-agent orchestration patterns & best practices (Copilot Studio) — 87 — https://learn.microsoft.com/en-us/microsoft-copilot-studio/guidance/multi-agent-patterns
+3. 🟠🔵 Designing the Internet of Agents (HAX) — arXiv 2512.11979 — 87 — https://arxiv.org/abs/2512.11979
+4. 🟠🔵 Agentic Design Patterns — UI/UX & Human-AI Interaction — 80 — https://agentic-design.ai/patterns/ui-ux-patterns
+5. 🟠 Aakash Gupta — Practical AI Agents for PMs: Tool Guide — 76 — https://www.news.aakashg.com/p/practical-ai-agents-pms
+6. 🃏 Ethan Mollick — What it feels like to work with Mythos — 72 — https://www.oneusefulthing.org/p/what-it-feels-like-to-work-with-mythos
+
+Errors/Notes:
+- YouTube RSS: FAILED — all channel feeds + rss2json proxy returned HTTP 403 (wholesale datacenter-IP block). Web-search fallback for YouTube returned only low-trust SEO videos; none met the bar, so no video pick this week (mild content-type-mix weakness, accepted in favor of quality).
+- Blogs/RSS: DEGRADED — direct feed fetches AND the rss2json proxy both returned 403 this run (broader block than the 2026-06-14 run, where the proxy still wasn't needed differently). Microsoft Research direct feed worked. Coverage rebuilt via 16 targeted web searches against named Tier-1 sources. WebFetch was effectively unusable for content verification all run — characterizations rely on corroborated web-search snippets.
+- X: skipped (no tool budget after fallback searches).
+- PushNotification tool: NOT AVAILABLE in this environment (ToolSearch found no such tool). Per STEP 7 fallback, noted here and continued. Katherine was notified via the in-page Notion @mention instead, which the task instruction designated as sufficient. No Google Calendar event created (per task instruction).
+- Dedup-log gap repaired: last week's Week-13 picks (2026-06-14) had never been written to previously-recommended.md; logged them retroactively this run alongside the new picks.
+- Source files config/sources.md, config/previously-recommended.md, and prompts/weekly-briefing.md still contain unresolved merge-conflict markers (Updated upstream / Stashed changes) from an earlier stash. Left untouched — not introduced by this run and out of scope. Flag for Katherine to resolve.
+
+Notion page: https://app.notion.com/p/38682e5a5fa381d68fd4d63a0d68a1e4
+Duration: ~single run