feat: add 2 more gold-standard blog training articles (13 total)

- blog-012: technology_deep_dive — coherent vs direct-detect decision framework
- blog-013: market_alert — transceiver price cycle, when to buy

Training set now covers: market_alert(2), comparison(1), technology_deep_dive(4),
tutorial(3), hype_cycle(1), buying_guide(1), migration_guide(1) — 13 total

blog-training-data/blog-012-coherent-vs-direct-detect-decision.md

@@ -0,0 +1,40 @@
+---
+title: "Coherent vs. Direct Detect: The Decision Your Network Will Make for the Next Decade"
+type: technology_deep_dive
+target_audience: technical
+score: 9/10
+---
+
+There's a moment in every network upgrade cycle when you have to decide whether your capacity problem is a data center problem or a WAN problem. That decision determines whether you're buying direct-detect optics or coherent optics, and getting it wrong means buying the wrong technology at the wrong price for the wrong reasons.
+
+The difference is not complexity — it's physics.
+
+Direct-detect transceivers convert light directly to electrical signal at the receiver. The transmitter sends intensity-modulated light; the receiver detects whether the light is on or off (NRZ) or determines the amplitude level (PAM4). No phase information, no carrier recovery, no DSP. This is every QSFP28 100G SR4, DR4, LR4 module you've ever touched. The optics are simple. The manufacturing cost is low. The interoperability is near-universal.
+
+Coherent transceivers transmit using the phase and amplitude of the optical carrier, encoding information in both in-phase and quadrature components, plus two polarizations. At the receiver, a local oscillator laser mixes with the received signal to recover the carrier phase. The DSP processes the electrical signal and recovers the original data. This process — called coherent detection — can be combined with powerful error correction, dispersion compensation, and nonlinear mitigation to achieve spectral efficiency that direct detect can't approach. 400G ZR modules carry 400G over 80km of uncompensated fiber using two polarizations and QPSK modulation. Direct-detect can't do that.
+
+The reason this matters operationally: coherent optics are not better versions of direct-detect optics. They solve different problems. Using coherent optics for an intra-data-center link is like renting a semi-truck to move a box — you're paying for capability you don't need. Using direct-detect for a 400km terrestrial WAN link is physically impossible regardless of price.
+
+The decision tree looks like this.
+
+For links under 10km, within a data center, between adjacent buildings, or on dark fiber with no DWDM: use direct-detect. 10G SR/LR, 25G SR/LR, 100G SR4/LR4/DR4, 400G SR8/DR4/LR4/FR4. The ecosystem is mature, compatible optics are widely available and cheap, and the performance is more than sufficient. No DSP complexity, no coherent penalties, no platform-specific tuning.
+
+For links over 80km, on DWDM infrastructure, crossing metro or long-haul fiber, or requiring line-side amplification: you need coherent. 100G DWDM pluggables (100G ZR, or older CFP/CFP2 coherent), 400G ZR (OIF implementation agreement, 80km target), 400G ZR+ (extended reach, vendor-specific, 600-2000km). These are platform-specific products. The performance depends on the DSP firmware. Interoperability exists but is tested on specific pairs, not assumed universally.
+
+The grey zone is 10-80km. This is where the argument is genuinely complex. Direct-detect LR4 covers 10km. Direct-detect ER4 covers 40km. For some metro deployments, that's enough and the simpler optic wins on cost and operational simplicity. For higher spectral efficiency or higher reliability over that range, coherent 400G ZR makes sense. The answer depends on your fiber budget, amplification infrastructure, and whether you're running DWDM.
+
+The operational complexity difference is real and often understated. A direct-detect link is up or it's down. The DOM data shows TX power and RX power. If both are in spec, the link works. If they're out of spec, you replace the optic. The troubleshooting flow is two steps.
+
+A coherent link has pre-FEC BER, post-FEC BER, constellation diagrams, OSNR margin, PDL, nonlinear noise figures, phase noise metrics, and DSP convergence state. The link can be "up" — passing traffic — while accumulating correctable errors at a rate that will eventually exceed FEC capacity. The diagnosis requires understanding what normal looks like for your specific fiber plant and traffic load. You need more sophisticated monitoring. You need engineers who understand coherent signal processing, or at least understand which parameters indicate which failure modes.
+
+That's not an argument against coherent. It's an argument for having the right tooling and people before deploying it at scale. Organizations that deploy 400G ZR without training, without enhanced monitoring, and without documenting baseline OSNR for every span will have incidents they can't diagnose.
+
+The compatible optic question is different for coherent. For direct-detect 400G DR4, buying compatible is straightforward — the MSA standard is well-defined, the optic is commodity, and every reputable compatible vendor has thoroughly tested it. For coherent, the situation is more nuanced. OIF 400G ZR is a published standard with compliance testing. Compatible 400G ZR modules exist and from reputable vendors they work correctly on tested platform pairs. But the "tested platform pairs" part matters more than it does for direct-detect. A compatible 400G ZR module should be validated on your specific linecard firmware and your specific peer device firmware, not just on the generic platform family. This testing exists — ask for the test matrix.
+
+The price delta matters at scale. A direct-detect 400G DR4 from a tier-1 compatible vendor: $25-40. A coherent 400G ZR: $500-1200 from OEM, $200-400 from compatible. For a 32-port switch spine, that's $800-1280 vs $6,400-12,800 in optics per box. For DCI, coherent is clearly worth it. For a data center fabric, direct-detect is clearly worth it.
+
+The upgrade path to 800G also differs by technology. 800G direct-detect (SR8, DR8, LR8 emerging) extends the same physics with more lanes or higher baud rate. 800G coherent (800G ZR, 1.2T extended reach) similarly extends coherent capabilities. The two tracks don't converge. If you're building DCI infrastructure for 400G ZR today, you're on the coherent roadmap for 800G. If you're building data center fabric with DR4, you're on the direct-detect roadmap.
+
+The decision is not about which technology is better. It's about which technology solves your specific problem. Metro and long-haul WAN: coherent, no compromise. Data center fabric within a campus: direct-detect, no compromise. DCI within a metro at 10-80km: evaluate based on fiber plant, amplification, and DWDM infrastructure. Everything else: the physics tells you which one to use, and the physics doesn't change based on which vendor you're talking to that quarter.
+
+Buy the right tool. Deploy with appropriate monitoring. Train the people who will operate it. That's the whole framework.

blog-training-data/blog-013-price-drop-timing-when-to-buy.md

@@ -0,0 +1,40 @@
+---
+title: "When to Buy: Reading the Transceiver Price Cycle Before It Reads You"
+type: market_alert
+target_audience: sales
+score: 9/10
+---
+
+Every network upgrade has a procurement window. Buy too early and you're paying innovation-phase prices for technology that'll be 40% cheaper in 18 months. Buy too late and supply pressure hits you six weeks before you need to cut over. Understanding where you are in the price cycle for the technologies you're deploying is worth more than any volume discount a vendor will offer you.
+
+The price cycle for optical transceivers follows a pattern that's been consistent enough across enough generations of technology to be predictable. Not precisely predictable — this isn't a formula, and supply chain disruptions can distort the timing — but the direction is reliable.
+
+A new optic type launches with limited supply and heavy manufacturing cost. 400G ZR in 2019 was $8,000-12,000 per module. Not because it cost that to make, but because only two manufacturers could make it, and they could charge that. The buyers at that price were building infrastructure that made the economics work at high cost: major cloud providers with very specific capacity requirements, telco operators with constrained fiber paths.
+
+Then multi-vendor qualification happens. More manufacturers certify to the spec. Compatible vendors commission the same DSP chipset. Volume builds. Price falls to $1,500-2,500. Early enterprise adopters buy here, willing to pay a moderate premium for technology that's now proven.
+
+Then commodity entry. The price floor for a mature MSA standard transceiver is roughly manufacturing cost plus margin. For coherent DSP-based optics, manufacturing cost is higher than for direct-detect. For simple SR4 with commodity VCSEL arrays, manufacturing cost is very low. When you see prices compress and multiple vendors offering similar pricing, you're at or near the floor. That's where compatible optics thrive — the technology is understood well enough that rigorous testing can verify compliance, and the price advantage over OEM is substantial.
+
+Right now, in mid-2026, here's where key technology categories sit in this cycle.
+
+100G SR4, LR4, CWDM4: commodity floor. Multiple manufacturers, widespread availability, stable prices for 3+ years. Compatible optics from tier-1 vendors: $22-35 for SR4, $45-75 for LR4. OEM list price 8-12x more. There's no price catalyst that will materially change this. Buy compatible, buy from someone who can guarantee the EEPROM data and batch test reports.
+
+400G DR4 (direct-detect): mature, approaching floor. $25-40 compatible, down from $180 two years ago. Still some room to fall — the DSP architecture was simplified and VCSEL lanes are now commodity. By end 2026, $18-25 is realistic for volume buyers. The decision: if you need units now, $25-35 is a good price. If you can wait 6 months, you might save 20%. The risk: supply shortages happen. At $25/module, the inventory carrying cost of buying now is low. Don't wait for marginal savings on commodity optics.
+
+400G LR4 (direct-detect): still declining. Currently $60-120 compatible, depending on vendor and certification status. ETA to floor: 12-18 months as volume scales. The 10km reach makes this a volume segment for enterprise campus deployments. Buy what you need for current projects; avoid overstocking in anticipation of deployment that's 12+ months out.
+
+400G ZR (coherent): early descent from peak. OEM at $800-1,200, compatible at $250-500 depending on platform validation status. The pattern suggests floor is around $150-250 in 24-36 months as multi-source manufacturing matures. If you're deploying DCI today, you pay today's price. If your DCI deployment is 12+ months out, there's meaningful savings to capture by waiting. The caveat: ZR is platform-validated, not just spec-compliant, so verify that the compatible you're evaluating has test data for your specific line cards.
+
+800G OSFP SR8/DR8: innovation phase. OEM $2,000+, limited compatible availability. This is early-adopter pricing. The use case is specific: AI/ML fabric, very high-density pod-scale switching. If you're building this now, it's because you have the workload that justifies the price. If you're evaluating 800G for 18 months from now, prices will be meaningfully lower.
+
+The supply-side risk factor is currently elevated for specific SKUs. The fab capacity constraints that disrupted 400G supply in 2022-2023 have mostly cleared, but the AI infrastructure buildout is creating localized pressure on high-end pluggables. 800G OSFP and 400G ZR+ are seeing supply-side pressure. Standard 400G DR4 and 100G are not.
+
+What this means for procurement: the commodity-tier decision (anything 100G, standard 400G DR4/SR4) should be based on project timing and storage logistics, not price speculation. The price isn't going to move enough to justify delayed procurement for active projects. Buy when you need it, from a vendor with consistent supply.
+
+For premium tiers (400G ZR, 400G ZR+, 800G): the price curve matters more. If the deployment timeline is flexible, the savings from waiting 12-18 months can be 30-50% on the coherent side. If the deployment timeline is fixed, don't wait — but negotiate hard on volume pricing and validate that your supplier has confirmed allocation.
+
+The other factor nobody discusses: the cost of the wrong decision. Buying early at peak prices is a known, quantifiable cost. The delayed deployment cost — traffic not flowing, capacity not available, customers waiting — is usually much larger. Most procurement teams optimize for the known cost (module price) while underestimating the unknown cost (deployment delay). When in doubt, buy at current prices and deploy on schedule.
+
+The pricing data we track across 60+ vendors shows price movements in real time. When prices for a specific SKU start dropping across multiple vendors in the same 30-day window, that's a signal that the manufacturing cost has been hit and competition is driving toward the floor. That's the data signal that justifies waiting. One vendor dropping prices while others hold steady is promotional, not structural.
+
+Watch the data. Know your timeline. Don't buy equipment for a deployment that's 18 months out at today's innovation-phase prices. Don't delay an active deployment to catch a price floor that may not arrive on your schedule. The pricing cycle is predictable in direction, not in timing.