The Dinoradar blog
Straight-talking guides on what Claude Code actually costs, how to read your usage, and how to get more out of your AI coding — from the people building a dashboard for exactly that.
How to read your Claude Code usage
Where your usage lives, the four token types that actually drive cost, and the five numbers — cost, cache hit rate, tokens, sessions, burn rate — worth reading in 30 seconds.
Read the guide →Claude Code stuck in a loop: catching a runaway session
A stuck agentic loop can out-spend a normal week in an hour — and nothing stops it by default. What a runaway looks like, why there's no cap, and how to catch one in time.
Read the guide →How to get notified when Claude Code needs you
Claude Code can run for minutes, then quietly wait on you. How the built-in notifications work, why they're easy to miss, and how to get a clear alert that clicks you back to the terminal.
Read the guide →Did Claude Code just leak a secret?
Claude Code reads your files and shell, so an API key can land in a transcript on disk. How it happens, where to look, and how to scan your history for exposed secrets.
Read the guide →How to write a CLAUDE.md that actually works
What belongs in CLAUDE.md and what doesn't, global vs project files, why short and specific beats long, and the habit that turns repeated corrections into rules that stick.
Read the guide →How much does Claude Code actually cost?
Subscription vs API pricing, what a real day of coding costs, why the same task can cost 5× more, and how prompt caching quietly saves you money. A plain-English breakdown.
Read the guide →Claude Code plans: Pro vs Max vs API
Three ways to pay, wildly different fits. What Pro, Max and the pay-as-you-go API each give you, who they're for, and how to tell which matches your real usage.
Read the guide →How to reduce Claude Code costs
Seven practical levers — from a one-setting model swap to lean context and prompt-caching habits — that cut your bill without cutting output quality.
Read the guide →How Claude Code prompt caching works
The biggest saver on your bill runs automatically — and you can't see it. What caching does, why reads cost ~10%, how to keep the cache hot, and how to measure what it saved.
Read the guide →Opus vs Sonnet vs Haiku for Claude Code
Model choice is the single biggest cost lever — a several-times swing on the same task. Which model fits which job, and how to stop expensive models doing cheap work.
Read the guide →The Claude Code context window — and what /clear and /compact do
Every turn re-sends your whole conversation as input, so a bloated
context is a tax on every message. How it works, and keeping it
lean with /clear and /compact.