AI-powered trading bots are quickly gaining traction across the crypto market, causing both excitement and anxiety among traders looking to automate their strategies. But industry experts say most people still misunderstand what these bots can and cannot do, and why specialized trading AI works so differently than general-purpose tools like ChatGPT.
This week’s episode of Byte-Sized Insight takes a deep dive into the rise of AI trading tools, the hype behind them, and the risks investors should consider before entrusting their capital to automated systems.
win the market
Brett Singer, head of sales and research at Glassnode, and Nodari Kolmakhidze, chief financial officer and partner at Cindicator, which built Stoic.AI, are two experts directly at the helm of the data, algorithms, and traders that will shape the next generation of AI-driven strategies.
Singer explained that the real power of AI in trading is not magical decision-making. Rather, it’s data processing.
“People create models that allow them to explore the entire database and develop and create these trading strategies within a day or two.”
He said Glassnode’s new Claude-powered MCP server has made advanced analytics much more accessible, saying, “You can pull it directly from the database and answer very complex questions…in minutes, seconds.”
Related: OpenAI Announces ChatGPT ‘Pulse’ — Will It Help You Trade Cryptocurrency?
But Singer cautioned that most AI bots are still inadequate in realistic market conditions. “For the most part, they didn’t beat the market,” he said, noting that many companies relied on shallow backtesting and single-signal strategies that lacked the robustness used by professional quant desks.
General-purpose AI and specialized AI
Winning the market may not be the domain of general-purpose AI models like the wildly popular ChatGPT. In fact, you’re more likely to do so with highly specialized bots designed specifically for the task. Kolmakhidze, who builds specialized trading AI, draws a line between chatbots and models designed for the marketplace.
“There’s a big difference between a specialized training model and a general-purpose one,” he said, arguing that it’s unrealistic to expect a text-trained chatbot to execute a profitable strategy. He emphasized that trading is notoriously difficult, even for top hedge funds.
Kolmakhidze also warned that many traders expect AI bots to become automatic profit machines.
“The biggest misconception is that AI bots are like money printing machines…that’s not the case.”
Market regimes change and even the strongest models can quickly break down as volatility and momentum structures change. “They’re good at predicting the past, but not the future,” he said, stressing the need for careful monitoring and long-term evaluation.
Both experts ultimately agreed that the future is not about AI replacing traders. AI is enhancing this. As Singer puts it, today’s AI functions like a “24-hour employee or intern,” but still requires human judgment.
For the full interview, listen to the full episode of Byte-Sized Insight on Cointelegraph’s podcast page, Apple Podcasts, or Spotify. Don’t forget to check out Cointelegraph’s full lineup of other shows.
magazine: Everyone hates GPT-5, shows AI can’t fix social media