A living glossary published May 9, 2026, defines the surge of AI jargon — AGI, LLMs, agents, RAG, RLHF and chain — of-thought—and explains why those distinctions matter for builders, integrators and risk assessors.
An updated glossary published May 9, 2026, collects and defines the key terms shaping current AI work-from AGI and LLMs to RAG and RLHF-with the stated aim of demystifying a fast-growing vocabulary. The editors say five minutes of reading will expose readers to core concepts and to the technical distinctions that influence design and risk decisions across the ecosystem.
On artificial general intelligence (AGI), the glossary emphasizes the term’s vagueness and records differing definitions from major organizations. It quotes OpenAI CEO Sam Altman describing AGI as "the equivalent of a median human that you could hire as a co-worker," cites OpenAI's charter definition of AGI as "highly autonomous systems that outperform humans at most economically valuable work," and notes Google DeepMind's take that AGI is "AI that's at least as capable as humans at most cognitive tasks."
The entry for "AI agent" frames agents as tools that use AI technologies to perform multi — step tasks on behalf of users, extending beyond single — turn chatbots. Examples cited include filing expenses, booking travel or a restaurant table, and writing and maintaining code. The glossary cautions that "AI agent" can mean different things to different audiences and that much of the supporting infrastructure to realize the full agent vision is still being built.
For software builders, the guide likens API endpoints to the "buttons" on the back of a program that other software presses to make it act. It explains that developers rely on these interfaces for integrations and flags a key operational shift: as agents become more capable, they are increasingly able to discover and invoke API endpoints autonomously, opening powerful — and sometimes unexpected — automation opportunities.
On reasoning, the glossary presents chain — of-thought prompting as an approach that breaks problems into intermediate steps to improve final answers. It offers a concrete illustration: a farmer with chickens and cows totaling 40 heads and 120 legs yields 20 chickens and 20 cows when worked out stepwise. The guide also states that reasoning — focused models are being developed from traditional large language models and are optimized for chain — of-thought thinking using reinforcement learning.
The practical implication the glossary stresses for practitioners is to preserve technical distinctions that matter in implementation: differing AGI definitions, the autonomy and capabilities attributed to agents, how agents interact with API endpoints, and the emergence of reasoning — focused model variants. Those distinctions, it warns, affect architecture and integration choices as well as operational and safety risk assessments while the AI ecosystem matures.
Sources
Replies (0)
No replies in this topic yet.
