What It Is

Most retail investor content is either generic ("diversify your portfolio") or locked behind an institutional paywall. Finovian sits between those two company-specific analysis on semiconductor and AI infrastructure stocks, written with a real position and a public track record.

Every article makes a specific call. Every call gets checked against actual earnings events. Wrong calls stay visible.

Content Architecture

Each covered company gets a four-layer cluster:

• Layer 1 Business Model: Evergreen. How the company makes money, competitive moat, revenue mix.
• Layer 2 Earnings: Quarterly. Specific numbers, guidance, what actually changed.
• Layer 3 Macro: Event-driven. Tariffs, export controls, cycle shifts that move the stock.
• Layer 4 Insights: Single-angle deep dives. One question, one real position.

Completed clusters: TSMC, Broadcom, AMD. Intel is next.

Track Record System

Every "Finovian's Take" contains:

• A falsifiable prediction not a direction, a number
• A measurable threshold with an exact check-by date tied to a real earnings event
• A binary outcome: correct or incorrect, updated after the event

Predictions are stored in Sanity with expectedValue and actualValue fields. The homepage pulls these dynamically. Nothing gets quietly buried.

Before publishing each Take, I post it on X and submit the URL to the Internet Archive. That timestamps the prediction before the event. Can't rewrite history.

Infrastructure

Deployed on Cloudflare Pages with Astro's static output adapter. Build runs on Cloudflare's global CDN sub-100ms TTFB on US routes.

Custom domain with full security stack:

• SSL/TLS: Full (strict) mode via Cloudflare
• HTTP Strict Transport Security (HSTS) with 1-year max-age
• Security headers via _headers file: X-Frame-Options: DENY, X-Content-Type-Options: nosniff, Referrer-Policy: strict-origin-when-cross-origin
• Content Security Policy scoped to trusted origins only
• Cloudflare Bot Fight Mode enabled

CMS: Sanity with custom document schemas. Track Record predictions use typed fields with validation rules can't publish a Take without specifying a numeric threshold and check-by date.

SEO Implementation

• Narrative-framed meta titles (not data-reporting format that kills CTR even with impressions)
• JSON-LD schemas: Article, BreadcrumbList, FAQPage per page
• IndexNow API submission on every publish Bing and Yandex indexed within hours
• dateModified synced to Sanity _updatedAt for freshness signals
• Bidirectional internal links across all four cluster layers per company

Content Standards

• No em dashes
• No AI vocabulary: pivotal, crucial, delve, tapestry, underscores
• No vague attributions ("analysts say", "experts believe")
• "Finovian's Take" written first-person with specific watch signals and a checkable number