Claude-focused AI architecture
QuantumRead's own route for designing and building with Claude: prompting that holds up in production, context and retrieval, tools and MCP, agents, evaluation, and the security and architecture decisions around them. Early version: written by QuantumRead and not yet reviewed by a domain expert.
Free, no account needed. The check tells you which of these you can already do.
Click a topic to read what it is. Click its box to mark it Learning, Done or Skipped - your own note, kept in this browser. QuantumRead counts evidence separately, and only from what you have proved.
Getting dependable results from a model: instructions, context and output shape.
Proved by: Explain it
Giving a model the right knowledge at the right moment, and knowing what it cannot hold.
Proved by: Explain it
Letting a model act through tools, and the boundaries that keep that safe.
Proved by: Explain it
Systems that plan and take steps, and the ways they go wrong.
Proved by: Explain it
Knowing whether it works, and noticing when it stops working.
Proved by: Explain it
Untrusted input, data handling, least privilege and risk you can describe.
Proved by: Explain it
Designing the whole system: patterns, trade-offs and what it costs to run.
Proved by: Explain it
Where to learn this
Everything opens in a new tab. The licence is printed beside each one, because a link is not permission to copy.