What If We Could Debug LLMs Instead of Retraining Them?
Exploring whether larger AI models could help maintain smaller ones by debugging their weights instead of always retraining them.
LLM
The Bitter Lesson, LLMs, and Learning From Experience
LLMs inherit knowledge from human-generated data, while reinforcement learning helps AI learn from action and feedback. The future may combine both.
Jev Is the Missing Piece in Production AI Systems
Production AI should not send every event to a reasoning model. Jev offers a cheap decision layer that can route, score, and escalate work first.
Caching in the AI Era
LLM caching isn’t response caching. It’s KV state caching: here’s how it works, what it costs, and how to structure prompts to actually get hits.