A regularly updated, living AI glossary assembles multiple expert definitions and contrasts for contested terms — AGI, AI agents, API endpoints, chain‑of‑thought and coding agents — to help builders navigate inconsistent usage and infrastructure implications.
A living AI glossary has been updated to disambiguate core terms such as AGI, AI agents, API endpoints, chain‑of‑thought, and coding agents, assembling multiple, sometimes conflicting expert definitions for builders. That matters because clearer terminology affects how developers design, integrate and govern autonomous systems and the infrastructure that supports them.
On AGI, the glossary presents several competing framings rather than a single definition. It quotes Sam Altman describing AGI as the "equivalent of a median human that you could hire as a co‑worker," cites an OpenAI charter definition of "highly autonomous systems that outperform humans at most economically valuable work," and notes Google DeepMind’s framing of AGI as "AI that’s at least as capable as humans at most cognitive tasks." The entry acknowledges ongoing expert confusion and treats the document as a living resource rather than a final verdict.
The guide distinguishes AI agents from simpler chatbots by function and autonomy. An AI agent is defined as a tool that uses AI to perform a series of tasks on a user’s behalf — examples include filing expenses, booking tickets or a restaurant table, and writing or maintaining code. The entry flags that the label "AI agent" can mean different things to different actors and emphasizes that the infrastructure required to deliver reliable, multi‑step autonomous behavior is still being built across the ecosystem.
On integration and control, the glossary frames API endpoints as the hidden "buttons" developers press to make services behave. It explains how agents and integrations rely on those interfaces: developers use endpoints to pull data, connect apps, or allow an agent to operate third‑party services without manual intervention. The entry warns that as agents gain the ability to discover and use endpoints autonomously, they can open powerful but sometimes unexpected automation pathways that builders must plan for.
Chain‑of‑thought reasoning receives concrete pedagogy and an illustrative puzzle: given a mix of chickens and cows totaling 40 heads and 120 legs, solving requires intermediate steps that lead to 20 chickens and 20 cows. For language models, chain‑of‑thought means breaking a problem into intermediate steps to improve answer quality; it usually takes longer but yields more accurate results on logic and coding tasks. The guide also notes that specialized reasoning models are developed from LLMs and are often optimized for stepwise thinking, frequently via reinforcement learning.
A focused entry on coding agents describes a narrower category that operates across development workflows. Unlike a suggestion‑only assistant, a coding agent can write, test and debug code autonomously: it can run tests across a codebase, spot bugs and push fixes, and perform iterative trial‑and‑error work that typically consumes developers’ time.
Sources
Replies (0)
No replies in this topic yet.
