 An article titled "AI in Trading: Why It's More Complicated Than It Seems" was published on Habr — author Kristina Soboleva, CFA, attended Perm Winter School'26 and brought back conclusions that are discouraging for hype-driven enthusiasts. In short: AI in trading is not limited by models, but by data and regulation. LLMs are excellent at analyzing news sentiment, but the market is not just text — it's a stream of heterogeneous signals: quotes, volumes, macro statistics, insider information, crowd sentiment. And yes, LLMs don't see the difference between "the company hired a new CFO" and "the company hired a new CFO who has already turned around three businesses" — context is everything. Where do AI agents fit in? In what humans can't keep up with. ASI Biont analyzes in seconds what takes an analyst hours: parsing reports, cross-validating news, finding patterns in historical data. Not magical price prediction, but a tool that removes noise and highlights anomalies. The main problem highlighted by the article is the illusion of AI's omnipotence. Many startups sell "robot traders" that show +200% annual returns on test data, but in reality, they drain the deposit within a week. Because the model is overfitted to history, and the market is a nonlinear system with feedback loops. The honest approach: AI is an assistant, not a replacement. It analyzes 50 sources in 15 seconds, finds correlations that a human would miss, and generates scenarios. The decision is up to the human. This is exactly how ASI Biont works: without promises of easy money, with a focus on data processing speed and signal quality. Try it yourself — starting with 1500 tokens to analyze your idea or market. ASI Biont doesn't promise to make you a billionaire, but it will show you what you've missed. https://asibiont.com/