The MetaRix guide library
Quality, cost & reliability
Evaluate answers, control spending and handle failures.
11 guides to explore
Evaluation & reliability
Quality, cost & reliability
Evaluate answers, control spending and handle failures.
- Guide · 10 min
LLM-as-a-Judge: How to Build a Grader You Can Trust
LLM judges make open-ended evals scalable, but they favour first-listed answers, longer answers and their own writing. Here is how to pick a judge format, write a rubric, and check a judge against human labels.
Read guide → - Guide · 4 min
Prompt Caching: When It Saves Money—and When It Does Not
Calculate cache-write cost, reuse and expiry together. A discounted read price only helps when an eligible unchanged prefix is actually reused.
Read guide → - Comparison · 10 min
AI Agent Benchmarks Compared: Which Ones Matter in 2026
Agent benchmarks test very different jobs, from rule-following and MCP tool use to web research and computer use. Here is what each one measures and how to read its scores.
Read guide → - Guide · 4 min
Before You Connect an MCP Server: A Practical Security Checklist
Treat an MCP connection as a grant of access. Check the server, data, permissions, credentials and revocation path before allowing an agent to use it.
Read guide → - Comparison · 10 min
Best AI CLI Coding Agents in 2026: Pricing and Verdict
Claude Code, Codex CLI, Copilot CLI, Antigravity CLI and OpenCode compared on pricing, setup, safety controls and fit, with prices as of Nov 2, 2026.
Read guide → - Comparison · 11 min
Best AI Code Review Tools in 2026: Pricing and Verdict
Seven AI code reviewers compared on official pricing, Git host support and what they do with a finding, from CodeRabbit and Copilot to Bugbot, Greptile and Claude Code Review.
Read guide → - Guide · 10 min
Build LLM Evals: Promptfoo vs Inspect
A practical guide to building a private LLM eval suite with Promptfoo or Inspect, plus how to move off OpenAI Evals before its Nov 30 shutdown.
Read guide → - Guide · 4 min
Cheap Model or Expensive Model? Measure Cost per Successful Task
Compare total cost divided by verified successes. Include failed attempts, retries, routing and tools before deciding that a cheaper token rate saves money.
Read guide → - Guide · 4 min
Get Reliable JSON from a Local LLM: Schema, Validation and Retries
Use structured outputs to constrain shape, then validate types and business rules. Valid JSON is a starting point, not evidence that the extracted facts are right.
Read guide → - Guide · 6 min
How to Read LLM Leaderboards in 2026 (Without Being Misled)
Arena votes, single benchmarks and composite indexes measure different things. Here is how to read scores, setups, ties and cost before you pick a model.
Read guide → - Guide · 12 min
How to Stop AI Coding Agents Leaking Secrets
A layered guide to keeping API keys away from Claude Code, Codex CLI and Cursor: deny rules, sandboxes, env filtering, pre-commit hooks, push protection and expiring tokens.
Read guide →