Tanner A Keyes

You'd be amazed how much more you win when you *actually* do ten times more than your closest competitor.

This year I just want to meet like minded people that help me get to my goals and where I want to be in life 🙏🏼

Always remembered, never forgotten.

🚨🚨POWERFUL🚨🚨 #NFL STAR AJ BROWN POSTED A 9-MINUTE VIDEO SPEAKING ABOUT MENTAL HEALTH AFTER RONDALE MOORE COMMITTED SUICIDE. “Don’t take the easy way out. I once thought that was the way.” Brown once nearly took his own life. A must-watch video 🙏🙏

Most men get their first set of flowers at their funerals. Harsh reality.

Vikings WR Rondale Moore was found deceased in the garage of his property with a suspected self-inflicted gunshot wound. Moore was only 25 years old 💔🙏

NFL wide receiver Rondale Moore died Saturday, according to a statement released by Louisville football coach Jeff Brohm, who coached Moore during his college career at Purdue. He was 25.

RIP Rondale Moore One of the best Wide Receivers to come out of Louisville metro. He was only 25 years old.

@JoshuaPoddoku Thank you! :)

@Keyes_Tanner BRAVO: you leveled up on these AI concepts

Built a RAG system from scratch and learned a lot about vector search, multi-tenancy, and embedding architectures. Here's what I discovered:

**The Stack Additions:** - `@xenova/transformers` (ONNX model runtime) - Cross-encoder models (Xenova/* repos) - TypeORM entities (Conversation, Message) - Token estimation utility - Context window optimization logic

4. Token estimation doesn't need to be perfect, just conservative 5. Conversation persistence improves UX dramatically Both features work together: better retrieval + better context = better answers.

Just shipped two major features that dramatically improved my RAG system: 1. Cross-encoder reranking for better retrieval accuracy 2. Full conversation management with context window optimization Here's what I learned building them: