Artificial intelligence is rewriting the world, and simultaneously inventing a whole new language to describe how it is doing it. Every product meeting, pitch, or panel is filled with acronyms like LLM, RAG, RLHF, and dozens of other terms that can make even very smart tech professionals feel insecure. To address this, TechCrunch has published an updated glossary for 2026, providing plain English definitions of the AI terms you are most likely to encounter, whether you are building with this technology, investing in it, or just trying to keep up.
AGI the holy grail of general artificial intelligence
Artificial General Intelligence, or AGI, remains a nebulous term. It generally refers to AI that is more capable than the average human at many, if not most, tasks. Sam Altman, CEO of OpenAI, once described AGI as the "equivalent of a median human that you could hire as a co-worker." OpenAI's charter defines AGI as "highly autonomous systems that outperform humans at most economically valuable work," while Google DeepMind views AGI as "AI that is at least as capable as humans at most cognitive tasks." Even experts disagree, making this one of the most debated concepts in the field.
Sponsored Protocol
AI agents autonomous assistants for complex tasks
An AI agent is a tool that uses AI technologies to perform a series of tasks on your behalf, going beyond what a basic chatbot can do. It can handle expenses, book tickets or a restaurant table, or even write and maintain code. However, as the original article explains, this space is still evolving and the infrastructure needed to fully deliver on its capabilities is being built. This concept also has implications for human labor: for instance, Amazon recently closed new customer access to Mechanical Turk, a platform that outsourced similar tasks to humans, which AI agents are beginning to automate.
Chain of thought step-by-step reasoning for language models
Chain-of-thought reasoning is a technique that allows large language models to break down a problem into intermediate steps, improving the quality of the final result. Just as a human uses pen and paper to solve a complex problem like "a farmer has chickens and cows, together they have 40 heads and 120 legs," an AI model with chain-of-thought produces more accurate answers, especially in logic and coding contexts. Reasoning models are optimized through reinforcement learning for this structured type of thinking.
Sponsored Protocol
Coding agents tireless software developers
A coding agent is a specialized version of AI agents applied to software development. Rather than simply suggesting code for a human to review and paste, a coding agent can write, test, and debug code autonomously, handling the iterative work that consumes a developer's day. It operates across entire codebases, spotting bugs, running tests, and pushing fixes with minimal human oversight. Think of it like hiring a very fast intern who never sleeps and never loses focus, though the work still needs human review.
Diffusion and distillation core of AI generation
Diffusion is the technology behind many image, music, and text generation models. Inspired by physics, diffusion systems slowly "destroy" data structure by adding noise, then learn a "reverse diffusion" process to reconstruct the original data. Distillation is a technique to extract knowledge from a large AI model (teacher) to train a smaller, more efficient model (student), as was done to develop GPT-4 Turbo from GPT-4. While all AI companies use distillation internally, some have used it to copy competitor models, violating terms of service.
Sponsored Protocol
For those who want to dive deeper, the original TechCrunch glossary includes terms like fine-tuning, GAN, hallucinations, and many more, providing a comprehensive picture of the AI lexicon. For broader context, you can refer to the Wikipedia page on artificial intelligence.
