Large language models can compress parts of a research workflow, but they are not trading research assistants. Their useful role is decision support: organising evidence, generating alternative scenarios, checking consistency and documenting assumptions. Every output still requires source verification, market-data validation and human accountability. Do not paste confidential information into public models or use these workflows for unattended trade execution. The screens glowed with numbers that almost didn’t feel real. Just last Friday, the S&P 500 hit a new high—6,966. Not long ago, numbers like that seemed out of reach. The week had been quietly confident, shaped by softer jobs data that kept everyone hoping for rate cuts. Meanwhile, copper shot past $13,000 a tonne on the LME. A few months before, people would’ve laughed at that price. But relentless demand from AI infrastructure, electrification, and new geopolitical supply scares pushed it there.
Markets don’t stop and announce when they’re changing direction. They just speed up, and if you’re still busy dissecting yesterday’s headlines, you get left behind. For pros and serious private investors, the real edge isn’t having more data. We all see the same feeds. The difference is whether evidence is verified, time-stamped and tested against credible alternatives.
Over the last year and a half, I’ve made conversational large language models part of my daily process. At first, it was just a small boost. Now I can run through different market scenarios, and connect signals that used to seem unrelated. What used to take hours now takes minutes.
This piece lays out the framework I use myself: 5 specific, reliable ways I blend technicals, fundamentals and macro trends data.
If you manage capital professionally, this will challenge and refine your approach.
