# Pushkin

> Pushkin takes your rough or incomplete AI prompts and rewrites them into structured, model-aligned instructions that yield best AI answers. It works inside ChatGPT, Claude, Gemini, Perplexity, Grok, Lovable, and v0, and in a web studio.

## Overview
Pushkin operates directly inside AI chat interfaces (ChatGPT, Claude, Gemini, Perplexity, Grok, Lovable, v0) through a floating brand mark and keyboard shortcuts (`Alt+Shift+I` for Improve, `Alt+Shift+R` for Refine). It also provides a standalone web-based Prompt Studio for drafting, refining, and testing prompts across target models.

## Core Capabilities
- **Improve (one click)**: Rewrites a rough idea into a clear prompt: role, task, context, format, and what to avoid.
- **Refine (a few questions)**: Asks 2–4 short questions, with choices or a blank to fill in, then writes the prompt.
- **Model-Specific Tuning**:
  - **ChatGPT**: Employs stepwise instructions, clear success criteria, numbered sections, and explicit output shape.
  - **Claude**: Adds explicit reasoning boundaries, tone framing, XML-like/markdown semantic sections, and "do / do not" constraints.
  - **Gemini**: Prioritizes concise task framing followed by scannable bullet points for context, format, and boundaries.
  - **Perplexity**: Formats prompts for cited web research, source verification criteria, and recency constraints.
  - **Grok**: Produces direct, conversational prompts specifying tone, perspective, and candid evaluation.
  - **Lovable**: Structures specifications as UI/product briefs (pages, components, state, data contracts, interactive UX).
  - **v0**: Formats prompts for implementation-ready UI generation (React, Tailwind CSS, components, layout, state variants).
- **Placeholder System**: Automatically inserts `[bracketed placeholders]` for facts, dates, names, or credentials the user has not provided, preventing hallucination.
- **Language Preservation**: Preserves the user's input language (e.g., prompt entered in Spanish, French, or Japanese is rewritten in that exact language).

## Pricing & Usage Quotas
- **Free Tier ($0/month)**: 5 Improve or Refine actions per UTC week. Access to standard web studio and Chrome extension. Last 3 history items.
- **Pro Tier ($5/month or $40/year - 30% discount)**: 100 Improve or Refine actions per UTC week. Unlimited prompt history. Reset occurs every Monday at 00:00 UTC.
- **Quota Rules**: Improve and Refine share the same weekly quota pool. Starting Refine (generating questions) uses 1 action; finalizing the refined prompt does not consume an additional action.

## Key URLs & Resources
- [Pushkin Home](https://pushkin-prompt.app): Product overview, live demo showcase, and pricing.
- [Prompt Guide](https://pushkin-prompt.app/guide): How to write better prompts, how each AI differs, and how to stop invented facts.
- [Prompt Studio](https://pushkin-prompt.app/studio): Web-based prompt optimizer, history, and library.
- [Upgrade to Pro](https://pushkin-prompt.app/upgrade): Monthly and yearly subscription checkout.
- [Privacy Policy](https://pushkin-prompt.app/privacy): Minimal data collection principles (device ID, prompt text processed ephemeral via AI Gateway, no prompt storage on servers).
- [Terms of Service](https://pushkin-prompt.app/terms): Clear service terms, quotas, and cancellation policy.

## Optional & Deep Documentation
- [Full LLM Knowledge Base](https://pushkin-prompt.app/llms-full.txt): Comprehensive technical specification, prompt architecture templates, target model comparison matrix, and developer integration details.
