AI's Vocabulary Keeps Growing: From Tokens to 'Claws,' a New Glossary Tries to Keep Up
A new CNET glossary of 54 essential AI terms captures how fast the industry's language is evolving, including the rise of locally run autonomous assistants known as 'claws.'
What matters
- CNET released a 54-term AI glossary on May 18, reflecting the speed at which AI vocabulary is expanding.
- 'Claws' are defined as locally run, open-source AI assistants that autonomously access calendars, email, coding tools, and browsers.
- The term has been popularized partly by AI researcher Andrej Karpathy and discussions around OpenClaw (also known as Clawdbot).
- The glossary highlights the growing mainstream need to understand agentic AI concepts beyond basic terms like tokens.
- No broad public backlash or hype cycle was evident in captured social data; discussion remains concentrated among AI insiders.
AI's Vocabulary Keeps Growing: From Tokens to 'Claws,' a New Glossary Tries to Keep Up
A new CNET glossary of 54 essential AI terms captures how fast the industry's language is evolving, including the rise of locally run autonomous assistants known as 'claws.'
What happened
On May 18, CNET published "Your AI Glossary: 54 Terms Everyone Should Know," a consumer-facing primer designed to decode the vocabulary of modern artificial intelligence. The glossary arrives as industry jargon accelerates beyond foundational concepts like tokens and transformers to newer, more specialized terms. Among them is "claws," a label for locally run, open-source AI personal assistants that can autonomously access calendars, email, coding tools, web browsers, and AI models to complete tasks on a user's behalf, often running on personal hardware such as a Mac Mini. According to Mashable, the term has circulated among AI researchers and online communities in recent weeks, with former OpenAI researcher Andrej Karpathy posting on X about "tinkering with claws." The concept is closely tied to OpenClaw—also known as Clawdbot—and the broader push toward agentic AI systems that act without constant human prompting.
Why it matters
The existence of a 54-term consumer glossary is itself a signal: AI is transitioning from research curiosity to everyday infrastructure, and the language gap between builders and users is widening. The specific inclusion of "claws" highlights a shift toward autonomous, local-first agents that operate outside traditional cloud APIs, raising immediate questions about security, privacy, and user control. If these assistants move from tinkerer setups to mainstream adoption, consumers will need to understand not just how a large language model consumes tokens, but how much unchecked access they are granting to software running locally on their own devices.
Public reaction
No strong public signal was available from Reddit or broad forum discussions in the captured inputs. Conversation about "claws" appears concentrated among AI researchers and early adopters on social platforms such as X.
What to watch
Watch whether major consumer publications continue to treat "claws" and agentic AI as standard vocabulary rather than niche developer slang. Monitor whether projects like OpenClaw release consumer-friendly installers, which would indicate a shift from hobbyist culture to mainstream utility. Finally, observe if glossary terms begin appearing in product marketing from Apple, Microsoft, or Google—an indicator that the lexicon has fully crossed into commercial reality.
Sources
- CNET: Your AI Glossary: 54 Terms Everyone Should Know
- Mashable: What are 'claws'? The next AI term you'll need to know
- ClawReviews: AI Glossary - 100+ Terms & Definitions
Why it matters
CNET has published a glossary of 54 AI terms aimed at general readers struggling to keep pace with the industry's rapidly expanding vocabulary. Among the entries is 'claws,' a term for open-source, local AI assistants that autonomously perform tasks across personal devices. The glossary underscores a broader challenge: as AI moves from cloud chatbots to agentic software running on private hardware, consumers need a new lexicon just to understand what they're using.
Public reaction
No significant Reddit or public forum discussion was captured for this story. Mentions of 'claws' appear concentrated on X and in AI researcher circles, suggesting the term has not yet broken into mainstream consumer debate.
What to watch
Watch for confirming reporting, product documentation, user-visible rollout details, and credible public discussion before treating this as settled.
Sources
Public reaction
No significant Reddit or public forum discussion was captured for this story. Mentions of 'claws' appear concentrated on X and in AI researcher circles, suggesting the term has not yet broken into mainstream consumer debate.
Signals
- Early adopter curiosity among AI researchers and developers
- Social media discussion concentrated on X
- No visible consumer backlash or confusion in captured data
Open questions
- Will 'claws' become a household term or remain developer slang?
- How will security and privacy concerns shape adoption of local autonomous agents?
What to do next
Developers
Experiment with local agent frameworks like OpenClaw to understand the API and permission models for autonomous assistants.
Early familiarity with local agent architectures will help you build safer, more efficient integrations as the ecosystem matures.
Founders
Evaluate whether your product roadmap needs an autonomous local-agent component or integration with 'claw'-style assistants.
Agentic AI is shifting from cloud-only to local-first; positioning your startup to interoperate with both models could be a competitive advantage.
PMs
Audit your product's terminology and onboarding to ensure users understand agentic features without requiring a 54-term glossary.
As AI vocabulary expands, user comprehension becomes a bottleneck; simplifying language can improve activation and trust.
Investors
Track GitHub activity and download metrics for OpenClaw and similar local-agent projects to gauge early traction.
Local autonomous agents represent a potential platform shift; quantitative signals can reveal which projects are converting hype to usage.
Operators
Review security policies for employee use of local AI assistants that may access email, calendars, or proprietary code.
Locally run agents bypass many cloud-security controls, creating new data-exfiltration and compliance risks.
Testing notes
Caveats
- This story is a terminology explainer and trend overview, not a product launch or API release.
- There is no single 'claw' product to test; OpenClaw and similar projects are experimental and may require advanced setup.
- Readers should treat glossaries as educational resources rather than hands-on tools.