New Word Glossary

Compressed pronunciation of "understand-well-exact-clean-sense-bearing." Describes judging situations well and handling oneself appropriately.

Compressed from "wanting perfection stresses me out." Describes the exhaustion and stress caused by pursuing perfection.

Compressed from "spoiler burns joy." Describes enjoying stories like dramas, movies, web novels even after knowing the ending.

A Korean slang combining "god" (신) with "life/living" (생). Describes a perfectly organized, productive, and admirable lifestyle.

Slang meaning "angry like a king," or "seriously frustrated." Expresses extreme frustration or annoyance, amplified from casual anger.

Online fandom that criticizes and ridicules celebrity behavior on SNS. Etymology possibly blends "억울한" (aggrieved) with "audience" (청중).

A derogatory Korean internet slang term for an old person, from '틀니 딱딱' (the clacking of dentures). Used to mock elderly people seen as outdated or stubborn. It is a clear ageist slur, so it should be used with caution.

Designing and optimizing the "harness" around an LLM or AI agent — its tools, prompts, action space, observation formatting, verification loops, and permissions — to get higher task-completion rates from the same model.

Designing and operating AI agents so they iterate in autonomous loops to plan, execute, and verify tasks without direct human intervention.

Building software by describing your intent to an AI in natural language and accepting the output by "feel" rather than scrutinizing every line of code.

An autonomous AI system that, given a goal, plans actions, executes them, evaluates results, and iterates to solve problems without direct human intervention.

Systematically designing and refining input queries (prompts) to maximize the performance and accuracy of Large Language Models.

A technique where an LLM retrieves relevant information from an external database before generating responses, ensuring accuracy and currency of output.

The maximum number of tokens (words) an LLM can process in a single interaction. Larger context windows enable deeper understanding of longer documents and more accurate responses.