Get more out of every AI tool
The settings worth changing, tips that cut token costs, and prompts you can copy, for 112 AI tools. Free, no sign-up.
- Tool guides
- 112
- Tips
- 785
- Last reviewed
- September 2026
Claude (Anthropic)
Settings guide · reviewed September 2026
- Model
- Haiku 4.5 → quick tasks | Sonnet 5.5 → everyday + code | Opus 5.5 / Fable 5.1 → hardest reasoning
- Projects
- Store files and instructions once per project instead of re-pasting
- Extended thinking
- On for analysis and planning, off for simple questions
- max_tokens (API)
- Always set it; 300–800 for chat replies, more for documents
Tip 1 of 7
Put standing context in a Project: upload reference files and write project instructions once, so you stop pasting the same background into every chat.
Popular guides
The tools people open most, one from each category.
- ClaudeChat / LLMPut standing context in a Project: upload reference files and write project instructions once, so you stop pasting the same background into every chat.Low burn
- CursorCodeWrite project Rules once (stack, style, patterns to avoid). Cursor applies them to every Agent request, which keeps generated code consistent.Medium burn
- MidjourneyImageV8.2 is the default model (since July 2026). Check the version selector before comparing results with older prompts.Medium burn
- PerplexitySearch / ResearchUse Perplexity when you need current information with sources; open the citations for anything you will rely on.Low burn
- ElevenLabsAudio / VoicePick the model for the job: Eleven v4 for expressive voiceovers and audiobooks, v4 Turbo or Flash v2.5 for low-latency agents and bulk work (elevenlabs.io docs, September 2026).Medium burn
- RunwayVideoGen-4.5 is Runway's current model, for text-to-video and image-to-video; it uses 12 credits per second (60 for 5 s, 120 for 10 s).High burn
What the burn rating means
Every tool gets a rating for how many tokens or credits it uses on default settings. Our rating, based on the published plans and how each tool behaves out of the box. The higher the burn, the more a settings change is worth.
- Low burn
- Efficient with sensible settings. Free tiers go a long way.For example Claude, Gemini, Perplexity
- Medium burn
- Settings matter. A few changes save real money.For example ChatGPT, Cursor, GitHub Copilot
- High burn
- Expensive on defaults. Read the guide before a long session.For example Runway
A short email when tools change
Pricing changes, new settings and one prompt worth keeping. We are collecting the list now and will write when the first issue is ready. Unsubscribe in one click.