# CUCKCO.DE — Prompt Training Arena for Vibecoders

> A free, bilingual (English / Russian) practice platform where developers learn
> prompt engineering by fixing real bugs and refactors. Users write prompts to
> guide an AI coach through a coding task, get supportive feedback, and study
> the underlying patterns in short academy lessons. Includes a PvP battle mode
> and a vibecoding glossary.

This site is for engineers who use AI coding assistants (Cursor, Claude Code,
Copilot, etc.) and want to write clearer, more reliable prompts. The content is
free; sign-in is via X / GitHub OAuth and unlocks unlimited practice runs.

## Glossary
Definitions are short, self-contained, and written in plain English.

- [Vibecoding](https://cuckco.de/glossary/vibecoding): Coding by describing intent to an AI assistant and iterating on its output instead of typing every line yourself.
- [Prompt engineering](https://cuckco.de/glossary/prompt-engineering): The practice of writing instructions that reliably steer an LLM toward useful, predictable output.
- [System prompt](https://cuckco.de/glossary/system-prompt): A persistent instruction that shapes the model's tone, role, and rules across a whole conversation.
- [Context window](https://cuckco.de/glossary/context-window): The maximum amount of tokens an LLM can read at once — system prompt, history, files, and your latest message all share this budget.
- [Cursor rules](https://cuckco.de/glossary/cursor-rules): Project-scoped instructions that the Cursor editor injects into the model's context whenever you ask it to code.
- [MCP (Model Context Protocol)](https://cuckco.de/glossary/mcp): An open protocol that lets AI assistants talk to external tools, data sources, and services in a standard way.
- [Agent](https://cuckco.de/glossary/agent): An LLM workflow that plans, takes actions through tools, observes results, and iterates until a goal is reached.
- [Tool use](https://cuckco.de/glossary/tool-use): When a model decides to call an external function (search, file read, code execution) and then uses the result to continue.
- [RAG (Retrieval-Augmented Generation)](https://cuckco.de/glossary/rag): A pattern where relevant documents are fetched from a knowledge base and injected into the prompt before the model answers.
- [Hallucination](https://cuckco.de/glossary/hallucination): When an LLM produces confident-sounding output that is factually wrong or invented.

## Academy lessons
Short theory modules with a paired bad / good prompt and a coach note.

- [Chain-of-Thought prompting](https://cuckco.de/academy/chain-of-thought): Asking the model to reason step-by-step before answering, which improves accuracy on multi-step coding tasks.
- [Chain of Density](https://cuckco.de/academy/chain-of-density): Iterative summarization technique that increases information density without losing coherence.
- [Few-shot prompting](https://cuckco.de/academy/few-shot-prompting): Providing 2–5 worked examples in the prompt to lock in the desired format and style.
- [Clarity principle](https://cuckco.de/academy/clarity-principle): Specific, unambiguous prompts produce more reliable outputs than open-ended ones.
- [Output formatting](https://cuckco.de/academy/output-formatting): Structuring model responses with Markdown, JSON, or delimiters so downstream code can parse them.
- [Persona adoption](https://cuckco.de/academy/persona-adoption): Assigning a role or identity to the model to shape its tone and reasoning frame.
- [React whisperer](https://cuckco.de/academy/react-whisperer): Prompt patterns optimized for AI-assisted React component generation and debugging.

## Challenges
Hands-on practice arenas. Each challenge presents broken code; the user
prompts the AI coach to diagnose and fix it. The AI evaluates prompt clarity
and either accepts the prompt and generates a reference solution, or asks the
user to refine it.

- [Center a div](https://cuckco.de/challenges/center-div)
- [Async blocking I/O](https://cuckco.de/challenges/async-blocking-io)
- [Connection pool exhaustion](https://cuckco.de/challenges/connection-pool-exhaustion)
- [Git rebase](https://cuckco.de/challenges/git-rebase)
- [Imbalanced dataset disaster](https://cuckco.de/challenges/imbalanced-dataset-disaster)
- [Iterator invalidation inferno](https://cuckco.de/challenges/iterator-invalidation-inferno)
- [Memory leak in cache](https://cuckco.de/challenges/memory-leak-cache)
- [Missing virtual destructor](https://cuckco.de/challenges/missing-virtual-destructor)

The full list is in the sitemap: https://cuckco.de/sitemap.xml

## Battle mode
[PvP prompt battles](https://cuckco.de/battle): Create a room, share the code
with a friend, and race to solve the same prompt challenge in real time. Score
is based on prompt clarity and number of attempts.

## Reference pages
- [Frequently asked questions](https://cuckco.de/faq)
- [Leaderboard](https://cuckco.de/leaderboard)
- [Privacy policy](https://cuckco.de/privacy)
- [Terms of service](https://cuckco.de/terms)

## How citation should work
Glossary entries and academy lessons are written to be quoted directly.
Definition leads (first sentence of each glossary entry) are self-contained and
suitable for one-paragraph citations. Academy lessons include a worked
"first draft" prompt vs. a "stronger version" prompt — the comparison is the
core teachable artifact.

## Contact
Built by @squatinc. Updates: https://x.com/squatinc