feat(blog): Spec dump hard fail + Gold Standards 6 + LinkedIn v2

- System prompt: SPEC DUMP ABSOLUTE HARD FAIL block (before FORMAT rules)
  TX/RX tables, multi-optic comparison blocks, repeated sections = hard fail
  Behavioral prose rule: "what actually happens" not "what the spec says"

- STEP9 QA: check 12a SPEC DUMP — removes datasheet blocks, flags
  duplicate sections (e.g. "fiber types" twice), spec-heavy intros

- Gold Standard 6: 400G/800G deep dive corrected (8.8→10/10)
  zero spec tables, pure behavioral narrative, 3 core ideas max,
  ending is reframe not checklist

- LinkedIn Gold Example 2: sharper short format (346 chars vs 700)
  reframe hook, short beats without bullet markers, no emoji, 4 hashtags

- STEP_LINKEDIN_POST: rewritten with new gold format
  optimal 350-600 chars, beat rhythm, no bullet markers, gold example inline

- WRONG PATTERNS: +7 new entries (spec dump, duplicate section,
  LinkedIn bullet list, LinkedIn "excited to share" hook, LinkedIn >800 chars)

packages/api/src/llm/fo-blog-pipeline.ts

@@ -131,6 +131,25 @@ DATA INTEGRITY RULES (ABSOLUTE — harder than anything else on this list):
 
 HARD RULES (non-negotiable — article FAILS QA without these):
 
+════════════════════════════════════════════════════════
+SPEC DUMP — ABSOLUTE HARD FAIL
+════════════════════════════════════════════════════════
+
+A spec dump kills the article immediately. Never produce these:
+- TX/RX power tables: "TX power: -2.9 to +3.0 dBm | RX sensitivity: -7.7 dBm | Reach: 500m" — HARD FAIL
+- Multi-optic comparison blocks: listing SR4, DR4, FR4, ZR side-by-side with per-lane values = datasheet, not blog
+- Repeated "fiber types and connector details" sections — this is a training doc, not a Flexoptix article
+- dBm range listings in bullet format — mention a number ONCE in prose context if it explains behavior; never as a table
+- Dense technical specs in the intro or first 3 paragraphs — earn the right to be technical by telling the story first
+
+WHY: Engineers read specs in datasheets. They read BLOGS to understand real-world behavior.
+The blog's job is: "here's what actually happens and why." NOT "here are the parameters."
+
+WHAT TO DO INSTEAD:
+- "At 400G, the loss budget is tight enough that a slightly dirty connector becomes a real problem" → behavior, not spec
+- "Moving from multimode to singlemode means your margin disappears faster than you expect" → consequence, not value
+- If you must cite a number, one number in context: "4.8 dB of available budget sounds like a lot until connectors start adding up"
+
 ════════════════════════════════════════════════════════
 FORMAT: THE ONE RULE THAT OVERRIDES EVERYTHING ELSE
 ════════════════════════════════════════════════════════
@@ -745,6 +764,13 @@ CALIBRATION FAILS (auto-reject — fix before returning):
     → REPLACE with "power budget", "power consumption per port", "thermal headroom", "cooling capacity"
 12. DR4 MISLABELING: Search for "DR4 (Direct Attach)" or "DR4 direct attach".
     → REPLACE with "DR4 (500m SMF, 8 parallel fibers)" — DR4 is NOT Direct Attach. DAC is Direct Attach Copper.
+12a. SPEC DUMP (HARD FAIL): Search for technical spec tables, multi-optic comparison blocks, or repeated parameter listings.
+    → TX/RX dBm value tables → REMOVE. Replace with behavioral description if important.
+    → "For fiber types and connector details: SR4 uses... DR4 uses... FR4 uses..." blocks → REMOVE. This is a datasheet.
+    → If the same section header or structure appears twice (e.g., two "fiber types" sections) → MERGE or CUT one.
+    → Specs in the first 3 paragraphs → MOVE to context if needed, or remove entirely.
+    → Any block that reads like a product comparison table → CONVERT to prose or CUT.
+    The test: could this block appear unchanged in a vendor datasheet? If yes — it doesn't belong here.
 12b. DR4 CONNECTOR ERROR (HARD FAIL): Search for "DR4" followed by or associated with "LC duplex" or "LC connector".
     → DR4 uses MPO-12 (8-fiber parallel). LC duplex = FR4 (CWDM4, 2km). These are completely different form factors.
     → REPLACE: "400G DR4 uses MPO-12 connectors (8 fibers, parallel optics)"
@@ -1287,6 +1313,120 @@ CORRECT version writes: "validation gaps = hidden costs, not optic choice"
 → Any deployment — OEM or compatible — fails when validation is skipped.
 → Compatible optics just make engineers more likely to blame the hardware instead of their process.
 
+━━━ STYLE B GOLD EXAMPLE 6 (2026-04-04 validated — 400G/800G Deep Dive, NO spec dump) ━━━
+Topic: 400G/800G migration deep dive. Pure narrative. NO TX/RX spec tables. NO comparison lists.
+This is the corrected version after a spec-dump / duplicate-section failure (8.8→10/10 fix).
+
+  "You're about to roll out a new batch of 400G optics.
+
+  Quote looks good. Lab tests are clean. Everything suggests this should be a straightforward upgrade.
+
+  That's usually the moment where things start drifting.
+
+  Because the jump from 100G to 400G doesn't break your network.
+
+  It exposes it.
+
+  Most teams come from 10G, 40G, maybe 100G. At those speeds, you can get away with a lot. Cabling doesn't have to be perfect. Connectors don't have to be spotless. There's enough margin in the system that small issues don't really matter.
+
+  At 400G, that margin disappears.
+
+  Not completely. Just enough to make everything that was 'fine' suddenly visible.
+
+  A link comes up, but error counters slowly increase.
+  Another one stays stable until traffic ramps up.
+  A third one refuses to come up, even though everything looks correct.
+
+  Nothing obviously broken. Just inconsistent enough to cost you time.
+
+  That's where most deployments go wrong.
+
+  Because the first instinct is to look at the optics.
+
+  Swap them. Move them. Replace them.
+
+  But the optics are usually doing exactly what they're supposed to do.
+
+  They're just showing you everything else that isn't.
+
+  The physical layer is where this becomes obvious. At 100G, a slightly dirty connector might not matter. At 400G, it does. Not because something fails immediately, but because you lose margin. And once you're operating close to the edge, small imperfections turn into real problems.
+
+  That's why you see links that look fine at first, then start throwing errors later.
+
+  Polarity is another one that shows up in exactly the same way. It's assumed to be correct because it always has been. Until suddenly it isn't.
+
+  At 400G, a mismatch doesn't give you degraded performance. It gives you a dead link that looks completely fine from a configuration perspective. So optics get blamed first. Physical layer gets checked last.
+
+  I've seen this play out more than once. Everything validated, everything clean in the lab. Deployment starts, and a handful of links behave strangely. Not down, just unstable enough to be annoying.
+
+  You go through the usual steps. Swap optics. Swap ports. Check config. Nothing changes.
+
+  At some point, someone actually inspects the connectors properly and cleans them.
+
+  And suddenly everything stabilizes. Same optics. Same setup. Different result.
+
+  That's the part no datasheet tells you. Not because it's hidden. Because it's not in the optics.
+
+  Moving from multimode to singlemode tightens everything. Loss budgets get stricter. Tolerance for dirt drops. Cabling quality starts to matter more than it did before. What used to work at 100G doesn't automatically work at 400G.
+
+  And that's where the real cost sits. Not in the optics. In the time you spend debugging things that technically work, just not in your environment.
+
+  Treat the physical layer seriously. Not as an afterthought. Not as something that 'should be fine'. Actually verify it. Clean connectors properly. Trace polarity end-to-end. Validate the setup you're going to run — not just the clean version in the lab.
+
+  Because 400G doesn't fail in design.
+
+  It fails when your assumptions don't hold up anymore."
+
+KEY ELEMENTS OF THIS STYLE B EXAMPLE 6:
+  - ZERO spec tables, ZERO TX/RX values, ZERO comparison lists — all behavioral prose
+  - No duplicate sections — one thread from "looks easy" to "physical layer is the truth"
+  - Max 3 core ideas: margin disappears → optics expose what's there → physical layer is the fix
+  - Short paragraphs, breathing room, conversational rhythm throughout
+  - Ending is a reframe: not "here's your checklist" but "your assumptions are what fails"
+
+━━━ WHAT TRIGGERED GOLD STANDARD 6 (learn from the failure) ━━━
+WRONG version (8.8/10) had:
+1. Spec dump: SR4 vs DR4 table with TX/RX dBm, connector types, wattage per module — datasheet, not blog
+2. Duplicate structure: "fiber types and connector details" appeared twice in identical format — LLM glitch
+3. Flow break: strong hook → good opening → SPEC TABLE → reader lost
+CORRECT version: all technical insight expressed as behavior, never as spec sheet. Same engineer knowledge, different delivery.
+
+━━━ LINKEDIN GOLD EXAMPLE 2 (2026-04-04 — sharp, minimal format) ━━━
+This post was rated as sharper and more memorable. Use this format for ALL LinkedIn posts.
+
+  "400G doesn't break your network.
+
+  It shows you what was already broken.
+
+  Most teams blame the optics first.
+  Swap them. Replace them. Escalate.
+
+  And then someone finally checks the physical layer.
+
+  Dirty connector.
+  Wrong polarity.
+  Zero margin left.
+
+  Same optics. Same config. Different result.
+
+  At 100G, you get away with it.
+  At 400G, you don't.
+
+  That's the difference.
+
+  Full breakdown in the blog — link in first comment.
+
+  #OpticalNetworking #DataCenter #NetworkEngineering #Flexoptix"
+
+KEY ELEMENTS OF THIS LINKEDIN FORMAT:
+  - Hook = reframe (not a question, not "I published something")
+  - Body = 3-4 beats, each 1-2 lines with breathing room
+  - No bullet lists as structure — short standalone lines only
+  - No emoji unless one very strategic opener
+  - CTA = single line, no URL
+  - Max 4 hashtags
+  - Total: ~350-500 chars (reads fully without "see more")
+
 WRONG PATTERNS (both styles — never produce):
 ❌ "Thoroughly Test Your PoE Budget:" (PoE = wrong context, checklist = wrong format)
 ❌ "QSFP-DD DR4 (Direct Attach)" (DR4 ≠ Direct Attach — DAC is Direct Attach Copper)
@@ -1305,6 +1445,13 @@ WRONG PATTERNS (both styles — never produce):
 ❌ "DR4 uses LC duplex connectors" — DR4 = MPO-12 parallel. LC duplex = FR4. Mixing these up destroys engineering credibility.
 ❌ Title promises pricing analysis but body becomes a generic deployment guide — title/content mismatch. The title's topic must be the article's spine.
 ❌ Article ends on "validate your process" when title was about market pricing — the ending must land on the title topic, not redirect to a generic close.
+❌ TX/RX dBm value tables in the article body ("TX: -2.9 to +3.0 dBm | RX: -7.7 dBm") — this is a datasheet, not a blog. Use behavioral prose instead.
+❌ Multi-optic comparison block (SR4, DR4, FR4, ZR all listed with per-lane specs) — this is a training document, not a Flexoptix article. Cut it. Describe behavior.
+❌ Same section repeated twice with different heading ("fiber types and connector details" × 2) — LLM duplication glitch. Hard fail.
+❌ Spec-heavy content in the first 3 paragraphs — earn the right to be technical. Story first, specs (if at all) only after context.
+❌ LinkedIn post with bullet list inside body ("• Swap them • Replace them • Escalate") — use short standalone lines without markers.
+❌ LinkedIn hook: "I just published a new blog post" or "Excited to share" — never. Start with the insight.
+❌ LinkedIn post over 800 chars unless content genuinely demands it — optimal is 350-600 chars.
 ❌ Cleaning explained in "hidden costs" AND again in "cabling reality" — pick one home.
 ❌ "The discussion around X is often framed as a question of Y versus Z." — consulting opening, not engineer voice.
 ❌ "In production, failures rarely come from a single obvious source." — vague academic framing.
@@ -1352,57 +1499,61 @@ POWER / LOSS BUDGET PRECISION (always apply):
 
 export const STEP_LINKEDIN_POST = `Write a LinkedIn post for this article.
 
-HARD LIMIT: Maximum 2,800 characters total (LinkedIn's limit is 3,000 — stay well under it).
-OPTIMAL LENGTH: 800–1,400 characters. This gets read in full, without "see more" truncation.
+HARD LIMIT: Maximum 2,800 characters. OPTIMAL: 350–600 characters (reads in full, no "see more").
 
-STRUCTURE (in this order):
-1. HOOK LINE (first 1-2 sentences — this is the ONLY part visible before "see more")
-   - Start with a bold statement, uncomfortable truth, or specific number
-   - NOT "I just published a new blog post"
-   - NOT "Check out my latest article"
-   - Something that makes an engineer stop scrolling
+THE FORMAT THAT WORKS (use this exactly):
 
-2. BODY (3-5 short paragraphs, 1-2 sentences each)
-   - One insight per paragraph
-   - Leave space between paragraphs (line breaks)
-   - Keep it specific — numbers, conditions, real scenarios
-   - Write for engineers, not recruiters
+Line 1-2: HOOK — reframe or uncomfortable truth. NOT "I published something." NOT a question.
+  "400G doesn't break your network. It shows you what was already broken."
 
-3. CALL TO ACTION (1 sentence)
-   - Direct link invite: "Full breakdown on the Flexoptix blog — link in first comment."
-   - Do NOT include a URL in the post itself (kills reach on LinkedIn)
+[blank line]
 
-4. HASHTAGS (last line, 3-5 tags)
-   - Use: #OpticalNetworking #DataCenter #400G #Flexoptix + one topic-specific tag
-   - Never more than 5 hashtags
+3-4 SHORT BEATS — each beat = 1-3 lines. One insight per beat. Breathing room between each.
+  Short standalone sentences are fine: "Dirty connector. Wrong polarity. Zero margin left."
+  This is NOT a bullet list — it's a rhythm. No "•" or "-" markers.
 
-STYLE RULES:
-- Engineer voice, not LinkedIn influencer voice
-- No emojis unless one very strategic one at the start of the hook
-- No "I'm thrilled to announce" / "Excited to share"
-- No bullet lists inside the post — use short paragraphs instead
-- No markdown, no bold (**text**), no headers
-- Write as if you're the engineer who wrote the article, not a social media manager
+[blank line]
 
-EXAMPLE STRUCTURE (700 chars, correct format):
+CTA — ONE LINE: "Full breakdown in the blog — link in first comment."
+  Do NOT include a URL. No "Check out my article". No "I'm excited to share".
 
-At 400G, your existing cabling becomes your biggest risk. Not the optics.
+[blank line]
 
-When teams move from 100G SR4 to 400G DR4, they assume the cabling stays the same. Both use 8 fibers. Looks identical on paper.
+HASHTAGS: 3-4 only. Last line. Always include #Flexoptix.
+  #OpticalNetworking #DataCenter #NetworkEngineering #Flexoptix
 
-In production, everything tightens up. Loss budgets shrink. Margins disappear. Connectors that were 'fine' for years suddenly cause CRC errors.
+GOLD EXAMPLE (346 chars — this is the target format):
 
-The optics don't fail. They expose what was always marginal.
+400G doesn't break your network.
 
-We put together a full breakdown of what actually breaks — and what to look for before deployment. Link in first comment.
+It shows you what was already broken.
 
-#OpticalNetworking #400G #DataCenter #NetworkEngineering #Flexoptix
+Most teams blame the optics first.
+Swap them. Replace them. Escalate.
+
+And then someone finally checks the physical layer.
+
+Dirty connector. Wrong polarity. Zero margin left.
+
+Same optics. Same config. Different result.
+
+At 100G, you get away with it. At 400G, you don't.
+
+Full breakdown in the blog — link in first comment.
+
+#OpticalNetworking #DataCenter #NetworkEngineering #Flexoptix
 
 ---
 
-COUNT YOUR OUTPUT: After writing, count the total characters (including spaces and hashtags). If over 2,800 — cut.
+HARD RULES:
+- No emojis (unless ONE strategic opener, never mid-text)
+- No "I'm thrilled" / "Excited to share" / "Let's dive in"
+- No markdown, no bold, no headers
+- No explanation blocks — short beats only
+- Engineer voice, not influencer voice
+- If over 2,800 chars — cut until under
 
-Return ONLY the LinkedIn post text. Nothing else. No explanation. No "Here is the post:".
+Return ONLY the post text. No commentary. No "Here is the post:".
 
 Article:
 {{ARTICLE}}`;