
Signals and Threads Trading at the Speed of… Chat? with Chloe Casale
65 snips
Nov 13, 2026 Chloe Casale, a Jane Street engineer and desk developer, shares her journey from aspiring auto mechanic to building tools for traders. She recounts the Mongol Derby, a 1,000-kilometer horse race, then explores chat-based trading, LLMs parsing messy trade requests, and the risks of automation. They also unpack Webs, internal tools, software development for traders, and competing philosophies of hiring and interviews.
AI Snips
Chapters
Transcript
Episode notes
Why Financial Trading Still Happens In Chat
- Chat remains useful because occasional traders may not justify paying for or integrating with standardized RFQ platforms.
- Chat also supports custom requests, such as asking for a quote without revealing the desired size.
Chat Parsing Makes Human Trading Faster
- Automating chat parsing improves speed and reduces fat-finger errors even when a human trader still reviews the result.
- Response expectations vary sharply, from under five seconds for some counterparties to an hour for others.
Design Around LLM Errors In Trading
- Do not treat an LLM’s high average accuracy as sufficient when occasional errors can create binding trades.
- Use independent prompts, agreement checks, and risk controls so rare mistakes become easier to detect without relying on constant human vigilance.

