EpochLex · Dictionary of the AI Era
How EpochLex builds an entry
EpochLex is a living reference, but that does not mean anything can simply be added. Each entry is meant to represent real language, explain it clearly, and distinguish what the evidence supports from what remains uncertain.
1. Notice a candidate term
Terms can surface in research papers, technical documentation, product interfaces, workplaces, standards, news, online communities, or everyday conversations about AI.
2. Confirm that people actually use it
EpochLex is not a collection of phrases an AI model can invent on demand. A candidate should have documented real-world usage. The evidence may look different for technical vocabulary, emerging slang, or governance language, but there should be a reason to believe the term exists outside the dictionary.
3. Decide whether it deserves its own entry
Not every phrase near AI needs a page. EpochLex asks whether the concept is distinct enough to look up, whether it is better treated as an alias, and whether adding it will genuinely help a reader understand AI-era language.
4. Explain it in plain English
Definitions should be accurate without requiring the reader to already know a second layer of jargon. Entries also include a natural example, pronunciation, category information, aliases when useful, and a status that helps show whether the language is technical, established, emerging, slang, contested, or otherwise notable.
5. Research the provenance
Knowing what a term means is not the same as knowing where it came from. A source can explain a concept well without proving who coined its name or when the phrase first appeared.
EpochLex therefore researches meaning and history separately. When the evidence supports it, an entry may document a naming event, an early use, a person or organization associated with the term, important moments in its development, and sources that support those claims.
6. Use careful language when history is uncertain
Some terms have clean origin stories. Many do not. EpochLex would rather leave an earliest-use field unresolved than invent precision. Words such as coined, introduced, formalized, and popularized are not interchangeable, so the wording should match the evidence.
7. Connect it to related ideas
Related terms help readers continue exploring. EpochLex can surface direct and reciprocal connections, but it does not invent relationship labels that the underlying research does not support.
8. Keep a human in the review loop
AI may assist with discovering terms, finding possible sources, organizing research, drafting language, and maintaining the software. AI output is not treated as a source, and AI does not independently decide what gets published.
Human review is responsible for inclusion, source judgment, final wording, classifications, provenance claims, and publication.
9. Publish it as part of a living dictionary
A published term receives a dedicated page and becomes discoverable through search, A-Z browsing, categories, and related-term navigation. Publication is not the end of the entry's life. Meanings, statuses, sources, and historical understanding can change as language and evidence change.
What “researched” means
EpochLex currently has 100 published terms, each with a provenance record that has completed an initial human-reviewed sourcing pass. “Researched” does not mean permanently settled. It means the current entry has been reviewed against sources and can be revised if better evidence appears.
Living Dictionary and annual editions
The website is the continuously updated EpochLex Living Dictionary. Annual editions are frozen snapshots intended to preserve how the dictionary looked at a particular editorial cutoff.
Want the detailed standards?
This page explains the process for readers. The open-source repository contains the deeper contributor standards for provenance, architecture, AI transparency, design, editions, and contribution workflows.