Skip to main content

How to read composite AI categories

How to read composite AI categories

DecAIHub assigns each project a composite category label such as AI/Agents | DeFi or AI/Compute | Infra. These labels look compact but carry two pieces of information at once: what kind of AI work the project does and where it applies that work. This page explains how to read them.

In short: the part before | is the AI subcategory (what the AI does); the part after | is the application domain (where it is used). Some projects have three parts when they span multiple domains.

Structure of a composite category

Every composite label follows the pattern:

{AI subcategory} | {Domain}

  • AI subcategory (before |) describes the type of AI work: agents, compute, inference, data, analytics, training, or other.
  • Domain (after |) describes the application context: DeFi, Infra, Consumer, Gaming, Meme, DeSci, and others.

When a project sits at an intersection of more than two areas, the label adds a third segment: AI/Agents | DeFi | Infra means agent technology applied to DeFi within an infrastructure context.

Parsing tips: read from left to right. The leftmost segment is always the primary AI subcategory. Additional AI segments (for example, both Inference and Compute) appear when the team’s core work genuinely spans those types; the rightmost segments usually narrow where that work sits (domain or infra context). If a label feels ambiguous, cross-check the AI component and Use case fields on the card — they describe the mechanism and product shape in more detail than the category line alone.

AI subcategories

Subcategory What it means
AI/Agents Autonomous or semi-autonomous AI systems that perceive, decide, and act
AI/Compute Decentralized GPU/compute infrastructure for AI workloads
AI/Inference Running trained models to produce outputs (predictions, generations)
AI/Data Data collection, labeling, or marketplace infrastructure for AI
AI/Analytics AI-powered data analysis, dashboards, and insights
AI/Training Model training, fine-tuning, or federated learning infrastructure
AI/Other AI work that does not fit the categories above

The AI subcategory determines what kind of AI work happens in the project and directly influences how DecAIHub evaluates the AI component field in the project card.

Domains

The domain segment tells you where the AI is applied:

  • DeFi — decentralized finance: trading, lending, liquidity, portfolio management.
  • Infra — infrastructure: protocols, platforms, middleware, developer tools.
  • Consumer — end-user applications: wallets, social, content, personal tools.
  • Gaming — games, metaverse, interactive entertainment.
  • Meme — meme-driven tokens and communities.
  • DeSci — decentralized science: research, biotech, open data.

Domains are coarse. They help browsing and comparison across hundreds of cards; they are not a full industry taxonomy. Two projects with | Consumer might serve different audiences, and that is expected — the finer grain lives in use case and AI component text.

Examples of composite categories

Composite category AI direction Applied to What it means in practice
AI/Agents | DeFi Agents Decentralized finance AI agents that trade, manage portfolios, or execute DeFi strategies
AI/Agents | Infra Agents Infrastructure Platforms and frameworks for building and hosting AI agents
AI/Agents | Meme Agents Meme narrative AI agents operating within a meme token context
AI/Agents | Consumer Agents End-user apps AI agents for personal use: assistants, content tools, social bots
AI/Compute | Infra Compute Infrastructure Decentralized GPU networks and compute marketplaces for AI
AI/Data | Infra Data Infrastructure Data pipelines, labeling platforms, or data marketplaces for AI
AI/Analytics | DeFi Analytics Decentralized finance AI-driven market analysis, risk scoring, or signal detection for DeFi
AI/Inference | AI/Compute | Infra Inference + Compute Infrastructure Combined inference serving and compute provisioning infrastructure

Use the examples as anchors, not limits. New combinations appear as the ecosystem experiments with agents, data, and tooling at different layers.

How composite categories appear in project cards

The Category field in every project card contains the composite label. When you filter or browse projects on DecAIHub, these labels let you quickly narrow down by both AI direction and domain.

The AI subcategory also guides which verification questions are most relevant. For example, an AI/Compute project is evaluated on whether it actually provides decentralized compute, while an AI/Agents project is evaluated on whether the agent genuinely perceives, decides, and acts.

Workflow suggestion: start from Category to pick a cluster, then read How to tell if a project really has AI with that cluster’s expectations in mind. A data-heavy label should eventually connect to datasets, pipelines, or labeling; an agents-heavy label should connect to tool use, memory, or action traces — as reflected in the card, not only in marketing.

Do not confuse with

  • AI component — the specific technology inside the project (e.g., "fine-tuned LLM," "reinforcement learning model"). The composite category is broader: it classifies the project's direction, not its internal architecture.
  • AI use case — what the AI is used for in the product (e.g., "portfolio rebalancing"). The domain segment of the composite category is related but more general.

Frequently asked questions

Can a project’s composite category change over time?

Yes. Teams pivot, add products, or refocus from one AI layer to another. DecAIHub updates labels when the dominant pattern of verified work shifts. Always treat the card date and evidence tables as the source of truth for “what the project is today.”

What if the label uses a domain I do not recognize?

Rare or emerging domain tags still mean “application context.” Use the AI subcategory links in the table above for the technical side, and read the project’s use case line for the product story.

Is the first segment always a single AI subcategory?

Usually. When two AI subcategories appear (for example, Inference and Compute), both describe types of AI work the project emphasizes. The domain segment(s) still describe where that work applies.

What composite categories alone do not prove

  • A detailed composite label does not prove the project has real AI. A project labeled AI/Agents | DeFi | Infra may still have ai_score = 0 if no Tier-1 evidence of working AI has been found.
  • Multiple AI segments in a label do not prove the project does more AI work than a single-segment project. Label complexity reflects categorization scope, not evidence depth.
  • A domain segment does not prove the project operates in that domain on-chain. A project labeled | DeFi might interact with DeFi protocols only in marketing, not in its actual product. Check the Onchain necessity flags for verification.

External references

See also

No comments to display

Back to top