Markus Halvorsen

I started working in iGaming compliance and licensing in 2010, focusing on operator risk assessment and regulatory frameworks across European markets. Over the next five years, I specialised in multi-jurisdictional licensing structures - MGA, UKGC, KGC - and began reviewing bonus terms against player-protection standards. The work gave me direct visibility into how operators draft their T&Cs, where the legal grey areas sit, and which clauses systematically disadvantage players.

By 2020, I had a recurring frustration: most affiliate review sites operated as paid listing directories. Rankings were sold, criteria were vague, and editorial independence was theatre. The problem was structural - review teams needed scale, scale required money, and money came from operators who expected favourable placement. No accountability, no public formula, no audit trail.

In 2024, I started building CasinoScraper as a solo-founder project. I designed the six-pillar rating formula, built the FastAPI and Next.js stack, wrote the scraping pipelines, and integrated Anthropic Claude for TOS analysis. The platform spent two years in engineering-only mode - no public traffic, just pipeline development and data accumulation. In August 2026, CasinoScraper launched publicly across 38 markets with the full audit infrastructure already running: public formula, monthly re-scoring, daily catalogue expansion, automated TOS analysis, and full version history from day one.

What makes CasinoScraper different is the engineering discipline behind it. This is not a content site with a scoring system bolted on - it is an audit platform that happens to publish content. Every casino is re-scored monthly against fresh data. Every terms-and-conditions document is analysed by Claude against a fixed predatory-clause checklist. Every brand change is versioned. The workflow that would require an editorial team runs on Celery jobs and caching layers instead.

The platform launched across 38 markets on day one because the pipeline architecture was built first and content was generated second. I did not start with five markets and expand gradually - I built the tooling to handle 38 locales, ran it for two years, then made it public. The solo-founder model is not a limitation - it is the design constraint that made the automation necessary and the independence possible.

The editorial bar is also non-negotiable. Every text field on the site is written by me, in first person, with no ghostwriters and no swappable freelancers. When I describe a licensing framework or a withdrawal test, the voice is consistent because it is one voice. Where other platforms say "our team of experts", I say "I" or describe what the system does. That is not a stylistic choice - it is an accountability mechanism.

Most casino affiliate sites operate on a team-driven editorial model. They hire writers to produce reviews, SEO specialists to optimise for keywords, and analysts to assign scores. The output is polished, the update cadence is high, and the rankings are influenced by commercial relationships even when the disclosure says otherwise. The tell is usually in the vague criteria - "overall quality", "player satisfaction", "industry reputation" - and the absence of any public formula that would allow a third party to verify a score.

CasinoScraper inverts that model. The scoring is automated and deterministic. The six-pillar formula is published with exact weights. The rankings update monthly when the data updates, not when a partnership deal changes. The editorial voice is singular and attributed - every review, every market guide, every explanatory paragraph is written by the same person. The platform does not pretend to be a team operation because independence at this scale requires automation, not hiring.

The technical infrastructure also separates this project from typical affiliate sites. Most platforms run on WordPress with a review plugin and manual score entry. CasinoScraper runs on a custom FastAPI backend with Celery pipelines, PostgreSQL storage, Redis caching, and Next.js 15 on the frontend. Scraping happens monthly via ScrapingBee and Playwright. TOS analysis runs on Anthropic Claude. Scores recompute automatically when any input changes. Publication happens daily at 08:00 UTC via a scheduled job. The stack is over-engineered for a review site because this is not a review site - it is an audit platform.