Caveman Mode (Output Brevity)

Caveman mode reduces Pando’s output tokens by constraining expression — it cuts filler, restatements, and unnecessary prose while preserving code, commands, paths, errors, and verification output. This is a token-saving feature that helps reduce costs and keep responses focused.

Caveman mode does not reduce reasoning, tool use, testing, or verification — only how the AI expresses itself.

How to Use

Slash Commands

Type these in the chat input:

Command	What it does
`/caveman lite`	Enable light brevity — removes filler and preambles
`/caveman full`	Enable standard brevity — concise answers, no unnecessary explanation
`/caveman ultra`	Maximum brevity — fragments and bullets only, no paragraphs
`/caveman off` or `/caveman-finish`	Disable caveman mode

Settings

Set a default mode that applies to all new sessions:

Web UI: Settings > Token Optimization > Caveman Mode

TUI: Settings > Tools > Caveman

Configuration file:

[Caveman]
# "" (empty) = off, "lite", "full", or "ultra"
DefaultMode = "lite"

Behavior Levels

Level	Style	When to use
Lite	Normal sentences, filler removed. Answer first, then a short paragraph of context.	Everyday coding — saves tokens without changing communication style significantly
Full	Fragments and bullets. One idea per line. Cut greetings, sign-offs, self-narration.	High-volume sessions where every token counts
Ultra	Maximum compression. Only the essential facts survive.	Budget-constrained sessions or very long conversations

What Gets Cut

Greetings, sign-offs, self-narration (“I’ll now…”, “Let me…”)
Restating the request back to the user before answering
Preamble before answers and summaries that repeat what was just said
Generic transitions, hedging, apologies, praise, filler adjectives
Explanation nobody asked for

What Is Preserved (Never Compressed)

Code and code blocks
Command lines and file paths
Error text and stack traces
Test output and verification results
API signatures and URLs
Security warnings
Anything the user explicitly asked for detail about

A direct request for detail (e.g., “explain”, “walk me through”) overrides caveman mode for that reply. The AI will give the full explanation when you ask for it.

Per-Session Override

You can change caveman mode during a session without affecting the default:

/caveman lite     # switch to lite for this session
/caveman-finish   # turn off for this session

The session-level setting overrides the global default. When you start a new session, it reverts to the configured default.

Configuration Reference

Setting	Location	Values	Default
`Caveman.DefaultMode`	`.pando.toml`	`""`, `"lite"`, `"full"`, `"ultra"`	`""` (off)

Caveman mode is off by default. Existing installations see no behavior change until you explicitly enable it via slash command or configuration.

Auto HTTPS Certificates Session Compaction