The five disciplines of LLM optimization
A 4:45 script, stable chapters, the HTML transcript below and a vector visual. Published video URL: none.
Accessible media
Scripts and transcripts before any distribution claim.
No video is presented as published. These resources are ready to record and retain their sources, chapters and limitations.
HTML transcript
The problem
“LLM optimization” can refer to five different jobs. The map prevents SEO, application architecture, model compression, evaluation and agents from being confused.
AI answer visibility
The optimized object is a page or website. Work covers crawling, indexing, structure, entities, self-contained passages and observable mentions.
LLM applications
The optimized object is a product or pipeline. Task quality, cost and latency are measured together. Context, retrieval, routing, tools and caching remain connected to evals.
Models and inference
Fine-tuning, quantization, distillation and runtime selection change trade-offs between quality, memory and throughput. A theoretical estimate does not replace a benchmark.
Evaluation and reliability
An improvement only exists relative to a criterion. A versioned eval set retains cases, graders, failures and regressions.
Agents and protocols
An agent orchestrates model, state and actions. MCP connects data, tools and workflows without certifying their security or accuracy.