AIWAR Is a Live Board for Reading the AI Race

What went live

AIWAR is a new live service built around one readable question: how is the public AI race moving right now?

The surface is not a generic model leaderboard. It is a dark command board for the U.S.–China AI competition, with the site framing the two headline sides as U.S. Frontier and China Open-Weight. The interface keeps the conflict metaphor visible, but controlled: blue and red teams, a live wire, battle pages, field evidence, and a methodology tab that explains how the board should be read.

The important part is the restraint. AIWAR does not say it can measure the whole AI race. It shows labeled signals and turns them into an editorial view of momentum.

The main board

The home screen opens with AIWAR · LIVE, a search box, and three plain sections: Evidence, Signals, and Methodology. Below it, a Live Wire ticker streams recent GitHub, Hugging Face, and discussion signals with labels like BREAKING and DEVELOPING.

The central panel is the Main Theater. On the left: Blue Team · United States, with U.S. Frontier and models such as GPT, Claude, Gemini, Grok, and Llama. On the right: Red Team · China, with China Open-Weight and models such as DeepSeek, Qwen, Kimi, GLM, Ernie, and Hunyuan.

The page also shows active battles, per-model sentiment chips, and a tug bar labeled Editorial Impact Index · last 14d · reports. In the screenshot, the board is close: U.S. Frontier at +86 and China Open-Weight at +85. The copy underneath matters: higher means gaining ground, last 14 days, editorial index, not a benchmark.

Evidence before ranking

The strongest product decision is that AIWAR treats evidence as the atomic unit. The impact board is surrounded by field cards, source labels, timestamps, confidence labels, and movement reasons.

One card reads as a GitHub item for Claude Code, classified under open weights with U.S. and open-source impact. Another card shows a Hugging Face GLM release on the China Open-Weight side. The cards are not presented as final truth. They carry labels such as UNVERIFIED and are collected with timestamps.

The detail page follows the same rule. A field note about Meta delaying an AI model release is tagged HN, Llama, new AI model, and praise. It shows collected and original dates, then asks whether the event shifts the U.S.–China race. A user call can be attached to one side, but the page still says: too early to call.

Why the methodology matters

The service is careful about what its numbers mean. The methodology page describes the AIWAR Impact Index as an editorial, rule-based signal, not an ELO score, benchmark, or measurement. Each event is tagged by type and severity. The rules then assign signed faction points with confidence and recency controls.

That makes AIWAR closer to an editorial instrument panel than a judge. Export controls, open-weight releases, outages, price cuts, migrations, and adoption signals can move the board. But the site still exposes the rule table and keeps community reactions separate from endorsed reporting.

That distinction is useful. The AI race is noisy. Benchmarks move slowly, product experience changes quickly, and developer sentiment can turn on price, limits, tooling, or trust. AIWAR gives those softer signals a structured place without pretending they are universal truth.

What it avoids

The U.S.–China frame could easily become cheap. AIWAR avoids the worst version of that idea by not using real militarism, nationalist copy, or weapons imagery. The battle language is used as a reading frame for model competition, not as a political slogan.

It also avoids the usual AI-content trap. The product is not a wall of generated comparison articles. It is not another table claiming one model is simply the best. The better claim is narrower: here are the signals, here is how they are labeled, here is how the board moves, and here is where the call is still uncertain.

Why it belongs on the live page

AIWAR fits the Madman Studio live page because it is not just a landing page. It is an operating object: a shipped service with a distinct reading model, a live data surface, a newsletter loop, source cards, voting interactions, and a visible methodology.

The launch is small in the right way. It does not need to settle the AI race. It needs to make the race easier to read from the outside. That is the product: a live board where model releases, open-weight moves, community reactions, and confidence labels sit in one place.

Next work is straightforward: keep feeding the board, tighten verification labels, keep methodology visible, and let users call the matchups without turning calls into facts. The front moves only when the evidence moves.