Tech Talk with Tyler·1d agoI Tried Using Jev for RAG. It Didn’t Work.I had what seemed like a reasonable idea: what if an AI searched through documents the way a person would?
Tech Talk with Tyler·Oct 29When words and actions don’t match up: can a model say one thing and then do another?I started with a strange question: would an AI hesitate to delete another AI?
Tech Talk with Tyler·Oct 25The $10 Research Lab: How to Make Scientific Discoveries on Your LaptopAI-powered scientific research used to require millions of dollars. I built an autonomous research system that generated 150 ideas in five…
Tech Talk with Tyler·Oct 13The Quantization Scaling Law I Never Knew Until TodayApparently, the bigger an AI model gets, the less it cares about losing 75% of its weight precision.
Tech Talk with Tyler·Oct 11Can a Neural Network Learn to Merge Other Neural Networks?We tested two ways to combine what independently trained neural networks have learned, and discovered that a merger network with just 73…A response icon1A response icon1
Tech Talk with Tyler·Oct 2If a Neural Network Weight Only Holds About 5 Bits of Information, Why Do We Train in 16-Bit?TL;DR: Weights contain around 5 bits of information. We train at 16 to 32 bits because it makes the loss landscape smoother, but we can…A response icon1A response icon1
Tech Talk with Tyler·Oct 2Can 1-Bit AI Save Power? I Tested It on a Tiny ChipShort answer: yes. But there’s a catch.
InObviously SatirebyTech Talk with Tyler·Sep 30Attention Is All You Need to Check Whether the Fridge Is Open
Tech Talk with Tyler·Sep 18Enterprise RAG: Why Getting AI to Work With Company Knowledge Is Harder Than It LooksEnterprise RAG sounds easy.