A technical reference for Agentic AI. Learn how the stack fits together, or jump straight to the production problem you're solving.
Each topic answers a decision a builder actually faces.
| You're dealing with… | Topic |
|---|---|
| Choosing a model family or provider | Foundation Models |
| Unpredictable LLM costs | Tokens & Cost |
| Standard model failing on hard problems | Reasoning Models |
| Need to choose between RAG and fine-tuning | Fine-Tuning |
| Need a smaller model to mimic a larger one | Fine-Tuning |
| You're dealing with… | Topic |
|---|---|
| Agent breaking at long context | Context Windows |
| Deciding between managed APIs and self-hosting | LLM Serving |
| Wanting to cut input costs significantly | Prompt Caching |
| Inconsistent or unreliable outputs | Sampling |
| Response time too slow | Latency |
| Rate limits hitting in production | Rate Limits & Concurrency |
| Long context eating VRAM | KV Cache & Quantization |
| Retrieval quality worse than expected | Embeddings & Vector Stores |
| Wondering why MoE models are cheap | MoE Architecture |
| You're dealing with… | Topic |
|---|---|
| Picking or building an agent framework | Agent Frameworks |
| Deploying a ready-made autonomous agent | Autonomous Agent Systems |
| Single agent not enough | Orchestration |
| Tools called incorrectly or unreliably | Tool Use & Function Calling |
| Agent losing track in long conversations | Context Management |
| You're dealing with… | Topic |
|---|---|
| Agent output quality is unpredictable | Evaluations |
| Agent misbehaving on untrusted content | Prompt Injection & Security |
If you prefer layer-first navigation, start at The Stack. If you want project scope and editorial intent, see About.
