Economy Mode

TurboDev's Economy Mode reduces token consumption by instructing the AI to produce concise output. When active, the system prompt includes a directive that makes the LLM drop filler words, use sentence fragments, and keep code exact — same technical accuracy, ~50-70% fewer output tokens, direct cost savings.

Inspired by Caveman, implemented natively — no external skill or dependency.

How It Works

1. You activate a level with /economy eco or /economy ultra
2. TurboDev adds a conciseness directive to the system prompt
3. The LLM produces shorter responses — fragments instead of paragraphs
4. Code, commands, file paths, and error messages stay byte-exact
5. The level is persisted in ~/.turbodevrc and survives restarts

The economy directive costs ~60 tokens (negligible compared to the output savings).

Levels

Level	Description	Estimated Output Savings	Badge
off (default)	Normal output — full sentences, explanations	0%	—
eco	Direct and concise. No filler, no pleasantries.	~50%	ECO (yellow)
ultra	Telegraphic fragments. No explanations unless asked.	~70%	ULTRA (red)

Commands

Command	Description
/economy	Show current level and usage
/economy eco	Activate eco level (alias: /economy on)
/economy ultra	Activate ultra level
/economy off	Deactivate economy mode

Before / After

Normal (69 tokens)

"The reason your React component is re-rendering is likely because you're creating a new object reference on each render cycle. I'd recommend using useMemo to memoize the object."

Eco (~35 tokens)

"Component re-renders because inline object prop creates new ref each cycle. Wrap in useMemo."

Ultra (~19 tokens)

"Inline object prop = new ref each render = re-render. Use useMemo."

Same fix. Same accuracy. Fewer tokens.

What Stays Exact

Economy mode compresses prose style, not technical content:

• ✅ Code blocks — byte-exact
• ✅ Commands and CLI flags — exact
• ✅ File paths — exact
• ✅ Error messages — exact
• ✅ User's language — preserved (Italian input → concise Italian output)
• ❌ Filler words ("let me...", "sure!", "I'll now...") — dropped
• ❌ Restating the question — dropped
• ❌ Transitions and pleasantries — dropped

Persistence

The level is saved in ~/.turbodevrc:

{
  "economy": {
    "level": "eco"
  }
}

On startup, TurboDev loads the saved level and activates economy mode automatically. Toggle it anytime with /economy.

StatusBar

When economy mode is active, the StatusBar shows a colored badge:

• ECO (yellow) — eco level active
• ULTRA (red) — ultra level active

The badge sits next to the cost indicator so you can see both savings and spending at a glance.

When to Use

• eco — daily work. Direct responses, still readable. Best balance of clarity and savings.
• ultra — power sessions. Maximum density, minimal prose. You know what you're doing and just want the answer.
• off — learning/exploring. When you want full explanations and context.

Inspiration

Economy Mode is inspired by Caveman by Julius Brussee — a Claude Code skill that proved a simple prompt instruction can cut 65% of output tokens without losing technical accuracy. TurboDev implements its own native conciseness prompt; it does not bundle or install the Caveman skill.