> For the complete documentation index, see [llms.txt](https://kvants.gitbook.io/kvants-whitepaper/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://kvants.gitbook.io/kvants-whitepaper/the-problem-with-news-trading-agents.md).

# The Problem with News-Trading Agents

Why headline-reactive agents need quantitative models, sizing, validation, and risk structure before they can be trusted as software.

Headline-reactive agents can summarize events quickly, but speed is not the same as a testable edge. A news-only agent often lacks a falsifiable signal definition, repeatable sizing rule, cost model, and validation history.

A quantitative AI trading agent starts from the opposite direction. The agent has a model it can inspect: entries, exits, risk limits, sizing, costs, and failure modes. The AI layer helps author, explain, and stress the model, but the model remains visible to the user.

Kvants Studio is designed around that distinction. The user builds an explicit strategy graph, checks it against historical data, records assumptions, and exports a frozen package. AI is a research and authoring aid, not a replacement for user judgment.