TechCrunch's AI Glossary: Right on Time, Not Too Soon

There is a specific date that matters here: May 9, 2026. That day, TechCrunch published a guide to artificial intelligence terminology aimed at those who have spent months, or years, hearing words like hallucination, fine-tuning, or context window and have politely nodded without fully understanding what they mean. It is not a technical article, nor is it written for engineers. It is, without ambiguity, educational content for non-specialist professionals working near AI without working in it.

That a flagship publication dedicates editorial space to this in 2026 says something about the actual state of public conversation around these systems. Adoption grew faster than shared vocabulary, and that gap has concrete consequences: poorly informed purchasing decisions, misaligned expectations, and debates about regulation where participants are not quite talking about the same thing when using the same terms.

What the glossary covers and what it doesn't

TechCrunch's piece targets the concepts generating the most noise in mixed conversations, between technical and business teams, between journalists and experts, between policymakers and developers. Terms like hallucination (when a model generates incorrect information with apparent confidence), prompt, fine-tuning, inference, or foundation model form the core of the glossary. These are useful entries precisely because they are frequently used imprecisely.

What it does not cover, at least not in depth, is the more operational layer of the current ecosystem: what distinguishes a subagent from a plugin, what role MCP (Model Context Protocol) plays as a standard for models calling external tools, or how hooks fit into a workflow with Claude Code. This is not a criticism of the article, its target audience does not need that level, but it does define its real utility well.

Who it helps, and who it doesn't

If you regularly work with APIs, configure MCP servers, or write skills for Claude, this glossary offers you nothing new. If instead you work in product, communications, legal, sales, or management and need to participate in conversations about AI without relying on someone translating for you in real time, the guide does its job.

There is a third profile that should pay attention to it: the technical professional who explains AI systems to non-technical audiences. Having a public reference, written thoughtfully and in a reputable publication, makes that translation work far easier. Rather than building explanations from scratch, you can lean on a source your audience recognizes.

The underlying problem the glossary reflects

A glossary of this type in 2026 is not an anomaly, it is a symptom. The pace at which new models, tools, and paradigms have been introduced over the last three years has outpaced the ability of mainstream media to cover them with precision. The result is a mix of hyperbole, misunderstandings, and terms used as synonyms when they are not, AGI and LLM for example, or agent and chatbot.

That carries a real cost. Companies making adoption decisions based on a fuzzy understanding of the vocabulary end up buying solutions that do not fit their problems, or dismissing useful tools because they confuse them with others that actually failed. Rigorous education, even if basic, reduces that cost.

A note on timing

Publishing this kind of resource in May 2026 is neither late nor early: it is ongoing. AI glossaries need constant updating because the field generates new terminology at a pace that makes any static publication obsolete within months. The entry on context window, for example, has a very different operational meaning today, with models handling up to a million tokens, than it did two years ago. A good living glossary should reflect that evolution.

---

From EP, we appreciate that publications like TechCrunch maintain these kinds of updated resources: they do not resolve technical complexity, but they do reduce noise in conversations where precision matters. The bar is not very high, but it is still worth setting.