VirusTotal has been using Large Language Models (LLMs) to analyze malware for over a year, starting with macros and scripts. This experience gave us a good grasp of what LLMs can and can't do. But the real challenge was always executables. So, we took on a huge task: disassembling all the binaries and memory dumps in VirusTotal and using LLMs to figure out how they work. In this talk, we'll share what we've learned from this massive project. We'll be upfront about the challenges of using LLMs on complex malware and the wins we've had, including how LLMs provide an approach for pivoting that shows very promising early results. Come hear our story and get a glimpse of the future of malware analysis with AI. We'll have a real talk about how (besides the hype) there are areas where LLMs are making a real difference and what's next in this exciting field.