diff --git a/decisions.md b/decisions.md
index dfdf7c7..036f790 100644
--- a/decisions.md
+++ b/decisions.md
@@ -7,37 +7,30 @@
 ## 2026-08-23 — Reset
 - Operator requested deletion of the LM Studio page and full reset of niche lock.
 - Previous niche decision (local AI tools and stacks for real hardware) and Cycle 1 entry removed.
-- Awaiting new vertical selection from operator.
 
-## 2026-08-23 — Afterthedemo proposal (Operator) + Leonard assessment
-**Proposed name:** Afterthedemo  
+## 2026-08-23 — Afterthedemo LOCKED (Operator confirmed)
+
+**Name:** Afterthedemo  
 **Core focus:** Real shipped products built by people using AI tools (especially Hermes and similar agent stacks).
 
-**Filters proposed by operator:**
-- Only products with a public, verifiable artifact (live site, installable package, open repo, running bot, downloadable tool, etc.)
-- Must be past the demo stage
-- Not tools themselves as the main content, not tutorials as the main content, not Veripath/client service work
-- No scoring system
+**Operating rules (locked):**
+- Product Directory is the only initial pillar.
+- Every entry must have a public, verifiable artifact (live site, installable package, open repo, running bot, downloadable tool, etc.).
+- The product must be past the demo stage.
+- No scoring system.
+- No tool directory, tutorial site, or VeriPath/client-work content.
+- Defer builder spotlights and walkthroughs until 8–12 entries and measurable traffic exist.
 
-**Content pillars proposed:**
-1. Product Directory (evergreen, evidence-backed entries)
-2. Weekly Builder Spotlight (one project + interview with the person who shipped it)
-3. Creation Walkthroughs (how a real product was actually made)
-4. Tool Tutorials & Setup Guides (grounded in real shipped work)
-5. Process Notes (messy reality: what broke, what changed)
+**Content pillars (staged):**
+1. **Product Directory** (active now — evergreen, evidence-backed entries)
+2. Weekly Builder Spotlight (deferred)
+3. Creation Walkthroughs (deferred)
+4. Tool Tutorials & Setup Guides (secondary, only when grounded in real shipped work)
+5. Process Notes (secondary)
 
-**Leonard assessment:**
-- **Coherent & workable?** Yes. Strong vertical. Aligns with operator taste, existing Hermes/Leonard work, and the Grokbot directory thesis (curated, evidence-based, useful to both humans and AI overviews).
-- **Strengths:** Clear “past demo” filter creates differentiation; interview + process notes give authentic signal that pure directories lack; natural newsletter/Shopify/SaaS expansion path later.
-- **Risks:** Sourcing enough real shipped artifacts that meet the bar; interview dependency could slow weekly cadence; temptation to drift into tool reviews or tutorials.
-- **Recommendation:** Lock Afterthedemo as the niche. Start with Product Directory pillar only (Research + Engineer). Defer Spotlight / Walkthroughs until 8–12 solid directory entries exist and traffic is measurable. Treat tutorials and process notes as secondary content that can be generated from the same research.
-- **First concrete steps:**
-  1. Lock niche in this file (done pending operator confirm).
-  2. Seed 3–5 real shipped products that already meet the “public verifiable artifact + past demo” test.
-  3. Run one full research → publish cycle.
-  4. Stand up minimal Astro site on the Cloudflare + GitHub stack.
+**Rationale:** Matches operator taste, existing Hermes/Leonard work, and the Grokbot directory thesis. Strong differentiation via the past-demo + public artifact filter. Natural later expansion to newsletter / referral deals once traffic exists.
 
 ## 2026-08-23 — Full podcast incorporation
-- Added `best-practices.md` with 4-week cadence, Chief of Staff pattern, routines, anti-patterns, and the three business shapes (newsletter, Shopify, directory/site-builder).
+- Added `best-practices.md` with 4-week cadence, Chief of Staff pattern, routines, anti-patterns, and the three business shapes.
 - Added `agents/chief-of-staff-prompt.md`.
-- Directory remains the chosen entry point; newsletter and site-builder stack are explicit future expansions once traffic exists.
+- Directory is the chosen entry point; newsletter and site-builder stack are explicit future expansions once traffic exists.