TechCrunch publishes AI glossary as jargon like 'opaque recurrence' enters the lexicon

The TechCrunch piece, published on Sept. 7, describes the list as a living document that will be updated as the field evolves. It is meant for anyone who encounters AI terminology in product meetings, pitches, or panels — whether they are building with the technology, investing in it, or simply trying to stay informed through news coverage.

Among the terms defined is artificial general intelligence, or AGI, which the glossary says is a nebulous concept with no single agreed-upon meaning. It quotes OpenAI CEO Sam Altman describing AGI as the "equivalent of a median human that you could hire as a co-worker," while OpenAI's charter calls it "highly autonomous systems that outperform humans at most economically valuable work." Google DeepMind, by contrast, sees AGI as "AI that's at least as capable as humans at most cognitive tasks." The glossary notes that even experts at the forefront of AI research are confused.

The entry for "AI agent" defines it as a tool that carries out a series of tasks on a user's behalf, going beyond what a basic chatbot can do, such as filing expenses, booking tickets, or writing and maintaining code. The glossary cautions that "AI agent" may mean different things to different people, and the infrastructure needed to deliver its full envisioned capabilities is still being built.

An "API endpoint" is described as a "button" on the back of software that other programs can press to make it do things. Developers use these interfaces to build integrations, and as AI agents become more capable, they are increasingly able to find and use endpoints on their own, opening up powerful — and sometimes unexpected — automation possibilities.

The glossary also covers "chain-of-thought reasoning," a technique large language models use to break a problem into smaller intermediate steps. Similar to a human using pen and paper to solve a multi-step math problem, this method takes longer but tends to yield more accurate answers, especially in logic and coding. Reasoning models, the glossary explains, are large language models optimized through reinforcement learning to reason in this way.

Another entry focuses on "coding agents," a specialized type of AI agent that can write, test, and debug code across an entire codebase with minimal human oversight. The glossary compares such an agent to a very fast intern who never sleeps, while noting that a human still needs to review the work.

"Compute" is defined as the computational power needed to train and run AI models. In practice, the glossary says, it often serves as shorthand for the hardware that supplies that power, including GPUs, CPUs, TPUs, and other infrastructure that forms the bedrock of the modern AI industry.