Category 8 of 8 · 31 terms
AI and tooling
The fastest-moving category, and the one where words’ meaning can shift like sand in a storm. Agents, context windows, graph engineering, slop. There are huge opportunities for content here, but also holes to avoid.
All 31 terms in this category
- Agentic AIAI that takes actions and works through multi-step tasks towards a goal, rather than only answering a question.
- AI slopLow-value AI-generated material produced with little care, often at speed and in large quantities.
- APIApplication programming interface: the agreed way one piece of software asks another for something, like a waiter between the diner and the kitchen.
- Automated testingRepeatable checks that software runs automatically to flag defined problems in code or content.
- Context engineeringDesigning the instructions, examples, source material and surrounding information an AI system needs to do useful work.
- Context windowThe amount of information an AI system can consider at once when generating a response.
- Continuous integrationA practice where changes are regularly checked and merged so problems surface early.
- Decision modelsAn AI model built to decide rather than write: it picks from answers you've defined and says how likely each one is.
- DevopsShared practices that bring software development and operations together to deliver and maintain systems reliably.
- Feature flagA software-controlled switch that determines whether a feature or content variant is available, and to whom.
- GEOImproving the likelihood that content is selected, accurately represented and cited in AI-generated answers.
- GitA version control system for managing change, collaboration, review and release across files over time.
- Graph engineeringDesigning an AI system as an explicit map of steps and routes, so work moves between models, tools, agents and people in controlled ways.
- GuardrailsLimits and checks intended to keep AI-assisted work within defined bounds.
- HallucinationAn AI-generated claim that is fabricated, factually wrong or unsupported by the source it purports to describe.
- Knowledge graphA network of connected concepts, entities and relationships.
- LintingChecking text automatically against a set of written rules and flagging every place that breaks one.
- LLMLarge language model: an AI system trained to predict and generate language, used for tasks such as drafting, summarising and answering questions.
- Machine-readableRepresented in a form software can parse and process.
- MarkdownA lightweight way to write structured plain text using simple marks for headings, links, lists and emphasis.
- MCPAn open protocol for connecting AI applications to sources of information and tools.
- Merge conflictA clash between changes that the version-control system cannot combine automatically.
- PromptAn instruction or request given to an AI system.
- Pull requestA proposal to merge changes, keeping the differences, discussion and checks together for review.
- RAGAn approach in which an AI system retrieves relevant material and uses it as context when generating an answer.
- RefactoringRestructuring something without changing what it does or says, so the next change is easier.
- RollbackRestoring a selected earlier state to recover from a harmful change.
- Separation of concernsA design principle for dividing a system into parts that each deal with one kind of decision, so each can be understood and changed without disturbing the others.
- SkillA reusable AI capability or instruction set designed to help an assistant perform a specific kind of task consistently.
- StagingA production-like environment for reviewing and testing changes before release.
- System promptThe standing instructions an AI system receives before the user's request, shaping how it behaves and responds.