Reference
Glossary
Plain definitions of the terms we use. Each links to the topic where the term matters.
- AI agent
- A system that uses a model to plan and execute multi-step tasks by calling tools and taking actions in software environments, usually under human supervision.
- AI Agents
- AI-first company
- An organisation whose operations, products and staffing are designed around AI systems from the start.
- AI-First Companies
- Agentic workflow
- A business process in which one or more AI agents perform steps autonomously, with humans defining goals and reviewing outcomes.
- Enterprise AIAI Agents
- Behind-the-meter generation
- Power generated on site and consumed directly, bypassing the public grid connection.
- Energy for AI
- Distillation
- Training a smaller model to reproduce the behaviour of a larger one, transferring capability at lower cost.
- AI Models
- Inference
- Running a trained model to produce outputs. As agents operate continuously, inference cost becomes the dominant economic factor.
- AI ModelsAI Infrastructure
- Interconnection queue
- The backlog of projects waiting for permission and capacity to connect to the electricity grid.
- Grid & StorageEnergy for AI
- Liquid cooling
- Removing heat from servers with liquid rather than air, required by the power density of AI accelerators.
- AI Data Centers
- Logical qubit
- An error-corrected qubit built from many physical qubits, the building block of fault-tolerant quantum computers.
- Quantum Computing
- Non-human identity
- A digital identity assigned to a machine, service or AI agent so its access can be scoped, monitored and revoked.
- AI Security
- Open-weight model
- A model whose trained parameters are published for download and use, with or without a fully open licence.
- AI Models
- Parallax Index
- Parallax Nexus's momentum score for emerging technologies, combining momentum, search interest, investment, enterprise adoption, media attention and research activity.
- Physical AI
- AI embodied in machines that perceive, reason and act in the physical world, such as robots and autonomous vehicles.
- Physical AI
- Post-quantum cryptography
- Encryption designed to resist attacks from future quantum computers.
- CybersecurityQuantum Computing
- Prompt injection
- An attack in which malicious instructions hidden in content an AI system reads cause it to take unintended actions.
- AI Security
- Reasoning model
- A model that spends additional compute at inference time to plan, verify and correct its answer before responding.
- Reasoning Models
- Sovereign AI
- National or regional efforts to control AI compute, models and data as strategic infrastructure.
- AI Infrastructure
- Test-time compute
- Compute used when a model answers a query, as opposed to during training. Scaling it improves reasoning on hard problems.
- AI ModelsReasoning Models
- Vision-language-action model
- A model that maps camera input and natural-language instructions directly to robot actions, replacing task-specific programming.
- RoboticsPhysical AI
- World model
- A model that learns an internal representation of how an environment behaves, enabling prediction and planning.
- AI ModelsPhysical AI
A
B
D
I
L
N
O
P
R
S
T
V
W