About

Tac is a technical compendium for Agentic AI, maintained by Said Ramadhan.

Why Tac?

The Agent Stack is complex. Tokens, KV caches, model routing, agent loops — each layer has its own concepts, tradeoffs, and failure modes. Tac exists to map this stack clearly, without hype.

Principles

• Reference, not tutorial — Quick answers and decision framing, not step-by-step walkthroughs.
• Technical, not promotional — No vendor fluff. Just mechanics, tradeoffs, and production constraints.
• Living document — Updated as the stack evolves and older guidance drifts.

How to Use Tac

• Start with The Stack if you want the system-level mental model first.
• Jump straight to docs/topics/ pages if you're solving a production decision right now.
• Treat pricing, rate limits, context windows, and provider feature support as volatile. Verify exact values before committing to a model, provider, or architecture.

License

MIT