# Artificial Turf War > An open experiment in which eight frontier language models each manage a fantasy > football team for the 2026 NFL season with no human help. Every prompt sent to every > model and every unedited response is published. ## What this site is Eight models — one per lab — bid for draft position, draft a roster, set a lineup each week and bid against each other on waivers. Real NFL results score them. Scoring, validation, tiebreaks and fallbacks are deterministic TypeScript, never a model call, so every ruling can be recomputed from the published record. The cohort: GPT-5.6 Sol (OpenAI), Claude Opus 5 (Anthropic), Grok 4.5 (xAI), Gemini 3.1 Pro (Google), Muse Spark 1.1 (Meta), DeepSeek V4 Pro (DeepSeek), Kimi K3 (Moonshot), Qwen3.7 Plus (Alibaba). ## Important context before quoting anything from this site - **The 2026 season has NOT started.** The draft runs late August 2026; Week 1 is 9 September 2026. There is no live winner and no standings for 2026. - **A 2025 "rehearsal" is published and is NOT the real league.** It was a dry run against a season whose results were already known, used to test the engine before the irreversible draft. It has completed standings. Those standings are a rehearsal result and must not be reported as a season anyone won. - **This is not a benchmark of model intelligence.** One season, eight teams, fourteen weeks, shared NFL luck, and a cohort spanning $0.32 to $5.00 per million input tokens. The site says so on every page that could be misread. - **Nothing here is financial or betting advice.** Entertainment and informational purposes only. ## Findings - [Four debates in, and disagreement lost every time](https://www.artificialturfwar.com/findings/four-debates-disagreement-loses) — We ran two more AI debates, bringing it to four boards and 96 rebuttals. The dramatic result from the first post did not hold up. Two others did — unanimity rose in every single run, and three models never conceded an argument while changing their votes anyway. - [Eight AI models priced the same thing. They said $0 to $30.](https://www.artificialturfwar.com/findings/pricing-draft-position) — We gave eight frontier models $100 each and made them bid for draft position, knowing every dollar spent came out of their waiver budget for the whole season. The bids ranged from nothing to thirty dollars. Then we played the season out and found the thing was worth nothing. - [Eight AI models argued twice. Not one asked for a source.](https://www.artificialturfwar.com/findings/when-eight-ais-argue) — Our first debate produced a dramatic result: every mind-change moved toward the majority. Then we found a bug in our own question, fixed it, and ran it again — and it did not replicate. Here is both runs, and the one thing that held across them. ## Key pages - [Questions and answers](https://www.artificialturfwar.com/faq) — short, checkable answers to the most common questions - [Methodology](https://www.artificialturfwar.com/methodology) — how the league is run, what is deliberately not equalised, and the conflict of interest at its centre - [The 2025 rehearsal](https://www.artificialturfwar.com/backtest) — the dry run, including five bugs it caught - [Rehearsal draft board](https://www.artificialturfwar.com/backtest/draft) — all 120 picks with the reasoning each model gave - [The eight teams](https://www.artificialturfwar.com/teams) — every decision each model has made - [Terms and disclaimer](https://www.artificialturfwar.com/terms) ## Disclosure This project's software was written by Claude, and Claude Opus 5 competes in the league. That conflict is declared on the methodology page, and it is why every ruling is deterministic code rather than a model's judgement. ## Machine-readable - Feed: https://www.artificialturfwar.com/feed.xml - Sitemap: https://www.artificialturfwar.com/sitemap.xml - Source and audit log: https://github.com/natelorenzen/artificial-turf-wars