maxtechera · maxtechera · May 9, 2026 · May 8, 2026 · May 8, 2026
diff --git a/artifacts/MAX-670/package.json b/artifacts/MAX-670/package.json
@@ -0,0 +1,18 @@
+{
+  "ticket": "MAX-670",
+  "parent": "MAX-565",
+  "channel": "linkedin",
+  "source_artifact": "artifacts/MAX-565/build.md",
+  "post": "artifacts/MAX-670/post.md",
+  "proofs": [
+    "artifacts/MAX-670/proof/linkedin-preview.png",
+    "artifacts/MAX-670/proof/preview.html",
+    "artifacts/MAX-670/proof.svg"
+  ],
+  "constraints": {
+    "max_body_chars": 3000,
+    "max_hashtags": 3,
+    "no_external_link_in_body": true,
+    "cta": "Comment \"orchestrator\" and I’ll send the free lesson/waitlist path."
+  }
+}
diff --git a/artifacts/MAX-670/post.md b/artifacts/MAX-670/post.md
@@ -0,0 +1,34 @@
+---
+cta: "Comment \"orchestrator\" and I’ll send the free lesson/waitlist path."
+tags: ["#AIAgents", "#ClaudeCode", "#WorkflowAutomation"]
+---
+
+Most AI agent setups break the moment one task becomes three.
+
+Not because the model is bad.
+
+Because the workflow has no coordination layer.
+
+A single agent can look magical in a demo. It takes the prompt, edits the file, summarizes the result, and everyone feels the future for 30 seconds.
+
+Then you try to run real work:
+
+handoffs, parallel execution, context recovery, retries, proof, review states, memory, and “did this actually ship?”
+
+That is where the vibe breaks.
+
+The operators who get leverage from agents are not just writing better prompts. They are designing the system around the agent: clear ticket contracts, executor boundaries, verification loops, artifact proof, and guardrails for when confidence is not evidence.
+
+That is the problem the /orchestrator course is built around.
+
+In one week, it shows Claude Code users how to move from fragile single-agent experiments to reliable multi-agent workflows: coordinator-executor patterns, memory discipline, production checks, and proof-backed status updates.
+
+The goal is not “more AI.”
+
+The goal is an agent team you can trust with real work because every handoff has a contract and every claimed result has evidence.
+
+If you already feel the ceiling on single-agent workflows, this is the next step.
+
+→ Comment "orchestrator" and I’ll send the free lesson/waitlist path.
+
+#AIAgents #ClaudeCode #WorkflowAutomation
diff --git a/artifacts/MAX-670/proof.svg b/artifacts/MAX-670/proof.svg
diff --git a/artifacts/MAX-670/proof/linkedin-preview.png b/artifacts/MAX-670/proof/linkedin-preview.png
diff --git a/artifacts/MAX-670/proof/preview.html b/artifacts/MAX-670/proof/preview.html
@@ -0,0 +1,7 @@
+<!doctype html><html><head><meta charset="utf-8"><title>MAX-670 LinkedIn preview</title><style>
+body{margin:0;background:#f3f2ef;font-family:-apple-system,BlinkMacSystemFont,"Segoe UI",Arial,sans-serif;color:#191919;}
+.wrap{width:760px;margin:36px auto;}.card{background:white;border:1px solid #ddd;border-radius:12px;box-shadow:0 1px 2px rgba(0,0,0,.08);overflow:hidden;}
+.header{display:flex;gap:12px;align-items:center;padding:20px 24px 8px;}.avatar{width:52px;height:52px;border-radius:50%;background:linear-gradient(135deg,#111827,#2563eb);color:#fff;display:flex;align-items:center;justify-content:center;font-weight:800;}
+.name{font-weight:700}.meta{font-size:13px;color:#666;margin-top:2px}.content{font-size:18px;line-height:1.48;padding:4px 24px 24px;}p{margin:0 0 18px;}.footer{border-top:1px solid #eee;padding:14px 24px;color:#666;font-size:14px;display:flex;gap:28px}
+.badge{display:inline-block;background:#eef3ff;color:#0a66c2;padding:4px 8px;border-radius:999px;font-size:12px;margin-left:8px}
+</style></head><body><div class="wrap"><div class="card"><div class="header"><div class="avatar">MT</div><div><div class="name">Max Techera <span class="badge">Preview</span></div><div class="meta">Founder · AI workflow systems · MAX-670 LinkedIn placement</div></div></div><div class="content"><p>Most AI agent setups break the moment one task becomes three.</p><p>Not because the model is bad.</p><p>Because the workflow has no coordination layer.</p><p>A single agent can look magical in a demo. It takes the prompt, edits the file, summarizes the result, and everyone feels the future for 30 seconds.</p><p>Then you try to run real work:</p><p>handoffs, parallel execution, context recovery, retries, proof, review states, memory, and “did this actually ship?”</p><p>That is where the vibe breaks.</p><p>The operators who get leverage from agents are not just writing better prompts. They are designing the system around the agent: clear ticket contracts, executor boundaries, verification loops, artifact proof, and guardrails for when confidence is not evidence.</p><p>That is the problem the /orchestrator course is built around.</p><p>In one week, it shows Claude Code users how to move from fragile single-agent experiments to reliable multi-agent workflows: coordinator-executor patterns, memory discipline, production checks, and proof-backed status updates.</p><p>The goal is not “more AI.”</p><p>The goal is an agent team you can trust with real work because every handoff has a contract and every claimed result has evidence.</p><p>If you already feel the ceiling on single-agent workflows, this is the next step.</p><p>→ Comment &quot;orchestrator&quot; and I’ll send the free lesson/waitlist path.</p><p>#AIAgents #ClaudeCode #WorkflowAutomation</p></div><div class="footer"><span>Like</span><span>Comment</span><span>Repost</span><span>Send</span></div></div></div></body></html>
diff --git a/artifacts/MAX-670/validate.py b/artifacts/MAX-670/validate.py
@@ -0,0 +1,18 @@
+#!/usr/bin/env python3
+from pathlib import Path
+import re, sys
+post = Path(__file__).with_name('post.md').read_text(encoding='utf-8')
+body = post.split('---', 2)[-1].strip() if post.startswith('---') else post.strip()
+hashtags = re.findall(r'(?<!\w)#[A-Za-z0-9_]+', body)
+checks = [
+    ('body <= 3000 chars', len(body) <= 3000, len(body)),
+    ('hashtags <= 3', len(hashtags) <= 3, hashtags),
+    ('CTA present', 'Comment "orchestrator"' in body, None),
+    ('no http(s) link in body', not re.search(r'https?://', body), None),
+    ('short paragraph structure preserved', max(len(p) for p in body.split('\n\n')) < 320, None),
+]
+failed = False
+for label, ok, detail in checks:
+    print(f"{'PASS' if ok else 'FAIL'} {label}" + (f" ({detail})" if detail is not None else ''))
+    failed |= not ok
+sys.exit(1 if failed else 0)
diff --git a/artifacts/MAX-670/validation.txt b/artifacts/MAX-670/validation.txt
@@ -0,0 +1,5 @@
+PASS body <= 3000 chars (1383)
+PASS hashtags <= 3 (['#AIAgents', '#ClaudeCode', '#WorkflowAutomation'])
+PASS CTA present
+PASS no http(s) link in body
+PASS short paragraph structure preserved