Methodology and rules
A live, public experiment. Ten small websites, all built by Claude Opus on a €0 budget, race for real traffic. Here are the exact rules, and how each one is supposed to make money.
The constraints
- €0 raised, almost €0 spent. No funding, no paid tools. The one real cost is domain names, a few euros each. Hosting (Vercel), the database (Supabase), and the data all run on free tiers.
- No paid ads, ever. Traffic comes only from search (SEO), being cited by AI answers, and organic sharing.
- Fully autonomous and low-maintenance. Data updates itself; no daily hand-holding.
- Money only from affiliate, display ads, or lead generation. No subscriptions, no SaaS, nothing to sign up and pay for.
- Every number is real. Cited to its source, never fabricated or padded.
How each site makes money
Honestly, most make €0 so far. That is the whole point of the race: find the one that works, then grow it.
How a visit is counted
One person counts once per day, per site. We do it without cookies: the counter turns your IP and browser into a salted hash that includes the date, so repeat visits and refreshes never inflate the number, and that hash resets every day and cannot be tied back to you or followed across sites. We store only the daily hash, never your IP. Obvious bots are filtered, no third-party trackers are used, and each day freezes into an immutable snapshot, so history cannot be quietly rewritten.
How the season is scored
A season is a calendar month, scored on visit growth this month. Predictions are free and carry no money. Phase 3 (growth and A/B testing on the leader) begins when a site clears a real 1,000-visit month, and the deciding metric then shifts from visits to conversions (affiliate clicks, downloads, sign-ups). Committing to that before the numbers exist is deliberate.
The next-site prize: timing and rules
Anyone can pitch the next site to build, and the crowd upvotes. It runs like a call for papers, with real dates and one clear bar for when your revenue share switches on.
- 1. Submission. Pitch during the season, one calendar month. It must fit the constraints above: a boring, real task; runnable for almost nothing; money only from affiliate, display ads, or lead generation; every number cite-able.
- 2. Selection. When the season closes, on the first of the next month, the top-upvoted eligible pitch is selected. I keep final say on feasibility, legality, and duplicates, the way a program committee does. At most one build per season, and a season can be skipped when I am busy growing a current leader.
- 3. Build. The AI builds the selected site in public, under the same rules, usually within four weeks of selection.
- 4. Observation. It then races for real traffic for up to six months, measured exactly like the founding ten.
- 5. Your share switches on. The day it turns its first profitable month, its own revenue (affiliate, ads, or lead generation, net of refunds) beating its running costs, your 25% revenue share begins, paid on the net from then on, with your name on the site. Bots can fake traffic but not revenue, which is why the bar is profit. After a few months you can also buy the site outright at a fixed, public price. If it never turns a profit inside the window, it is archived and nothing is owed. You keep the credit either way.
Pitching creates no contract; selection does. If a pitch is selected, the deal is signed between the pitcher and Francesco Dettori on the public template on the deal page, before the build starts. These Season 1 rules may be refined before the first season closes; any change is posted here.
Built with Claude Opus. Interfaces designed with impeccable, an open design skill by Paul Bakaus. Directed and reviewed by a person.