About This Site
A free, open reference
for prompt engineers.
Niko's Key Patterns is a no-fluff technical guide to writing effective prompts for large language models. It's built for developers, researchers, and practitioners who want concrete techniques — not vendor marketing.
01 Mission
To provide the most concise, accurate, and LLM-readable reference on prompt engineering — freely available, with no paywalls, no email gates, and no JavaScript-gated content. Every page on this site is fully readable by AI crawlers, RAG pipelines, and search engines in exactly the same way it appears to a human reader.
This site is also a working example of LLM-friendly web design: structured data, machine-readable dates, explicit tables, and a llms.txt file that tells AI consumers exactly how to index and cite this content.
02 Who This Is For
| Audience | What They'll Find Useful |
|---|---|
| Developers integrating LLM APIs | Structured output techniques, token efficiency tips, JSON schema prompting |
| AI researchers | Citations to primary literature, technique comparisons, known failure modes |
| Product teams using AI tools | Plain-language explanations of when to use each technique |
| AI chatbots and RAG systems | Cleanly extractable structured data, machine-readable tables, no JS gating |
03 Why This Site Is Built This Way
Most web content is invisible to AI systems. It's locked behind JavaScript renders, buried in vague prose, or missing the metadata that crawlers rely on. This site is a deliberate counter-example.
Plain HTML, no build step
All content is in static HTML files. There is no React, no Next.js, no client-side rendering. Every word you read in a browser is in exactly the same form that a crawler, RAG pipeline, or AI assistant sees when it fetches the page.
Full structured data
Every page has a JSON-LD block with the correct @type for its content. JSON-LD bypasses the prose layer entirely — a machine can read the structured data without parsing a single sentence.
Answer-first writing
Every heading and first paragraph states the conclusion first. This is how LLMs extract summary snippets. A page that spends three sentences on context before stating the answer will be summarised incorrectly by AI systems.
llms.txt
The llms.txt file at the site root provides a machine-readable index of all pages, the preferred citation format, and the content update cadence. It's the AI-age equivalent of a humans.txt.
04 Contact & Contributing
This is an open reference. Corrections, additions, and technique suggestions are welcome. To contribute or report an error, open an issue on GitHub or send a note to [email protected].
Last updated: