# seale-semantic-agent-model-harness-ontology-data-2026-04-17

## Veille

Semantic agent: the model+harness and ontology+data symmetry, the collapse of agent frameworks, ontology as the only non-commodity asset

## Titre Article

There is a growing disconnect in the way people think about building AI agents

## Date

2026-04-17

## URL

https://www.linkedin.com/posts/tonyseale_there-is-a-growing-disconnect-in-the-way-share-7450647565982715904-kzDt/

## Keywords

semantic agent, agent harness, ontology, knowledge graph, business domain, Claude Code, Codex, agent frameworks, LangGraph, CrewAI, AutoGen, Semantic Kernel, scaffolding, shared domain model, typed linked data, commoditization, ontological moat, shared understanding, isolated vs combined intelligence, orchestration, post-training

## Authors

Tony Seale

## Ton

**Profile**: Perspective of a knowledge graph expert (The Knowledge Graph Guy), strategic and manifesto register, senior technical level with architectural and business scope. Tony writes like an evangelist who identifies a fundamental pattern and calls for an industry repositioning.

**Description**: The tone is punchy and pattern-focused. The article builds its thesis through symmetries and binary oppositions: model+harness on one side, ontology+data on the other, commodity vs moat, transient vs permanent. The LinkedIn style is short, dense, with sections marked by colored bullets (🔵) and a telegraphic rhythm (short sentences, lists of frameworks thrown out in rapid succession). The author leans on an Anthropic quote to reinforce their authority: "Every component in a harness encodes an assumption about what the model can't do on its own, and those assumptions can quickly go stale as models improve." The piece culminates in a proprietary maxim: "Everything else is scaffolding — useful for a while, but scaffolding comes down." The target audience is twofold: enterprise architects investing in frameworks, and data/AI leaders looking to identify the strategic asset to build within their organization. The tone is that of a consultant-philosopher looking to convert, not just inform.

## Pense-betes

- **Growing disconnect**: the industry invests heavily in orchestration frameworks (LangGraph, CrewAI, AutoGen, Semantic Kernel, OpenAI Agents SDK, AWS Bedrock, Google ADK) while leading-edge practitioners have moved to "powerful model in a powerful harness" (Claude Code, Codex, OpenClaw, Hermes)
- **The collapse of frameworks**: early frameworks were necessary when models couldn't handle multi-step tasks on their own. But as models improve, scaffolding built for limited models handicaps intelligent models. Scaffolding should decrease over time, not accumulate
- **Anthropic quote**: "Every component in a harness encodes an assumption about what the model can't do on its own, and those assumptions can quickly go stale as models improve"
- **Isolated Agents Are Not Enough**: access to the computer ≠ understanding. Give an agent 1000 documents and ask a question: it searches, hopes, guesses. Multiply that by 50 agents without a shared world model = intelligent in isolation, incoherent in combination
- **Need for a structured information environment**: at enterprise scale, agents need a shared domain model and humans in the loop
- **The fundamental symmetry**: apply the same simplification on the data side. The model sits in a harness that gives it access to the computer. The data sits in an ontology that gives it structure and meaning
- **Definition of ontology**: it defines what exists, its properties, its relationships. It's the interface through which agents understand data. Data — typed, linked, structured — forms a navigable graph of meaning
- **Two symmetric patterns**: (1) powerful model in a powerful harness, (2) powerful data in a powerful ontology
- **The Semantic Agent**: (Model + Harness) + (Ontology + Data). "It doesn't just generate. It starts to understand"
- **What is commodity**: everyone has access to the same frontier models. Anyone can build a harness. This thins out over time
- **What is NOT commodity**: the ontology, the domain model, the structured and linked knowledge that captures how YOUR organization understands the world
- **Models are rented, frameworks are transient**: the only thing left to build is knowledge — and it's yours
- **Tools cited**: Claude Code, Codex, OpenClaw, Hermes as examples of mature harnesses; LangGraph, CrewAI, AutoGen, Semantic Kernel, OpenAI Agents SDK, AWS Bedrock, Google ADK as previous-generation frameworks
- **External references**: Anthropic on Harness Design, Meaning IS Your Moat (earlier article by Seale)

## RésuméDe400mots

Tony Seale, The Knowledge Graph Guy, identifies a growing disconnect in the way the industry builds AI agents. On one hand, the industry is investing heavily in orchestration frameworks: LangGraph, CrewAI, AutoGen, Semantic Kernel, OpenAI Agents SDK, AWS Bedrock, Google ADK — each with its own orchestration graphs, state machines, and routing logic. On the other, leading-edge practitioners have moved to the "powerful model in a powerful harness" paradigm (Claude Code, Codex, OpenClaw, Hermes).

**The collapse of frameworks.** Early frameworks were necessary when models couldn't handle multi-step tasks on their own. Anthropic quote: "Every component in a harness encodes an assumption about what the model can't do on its own, and those assumptions can quickly go stale as models improve." Scaffolding built for a limited model handicaps an intelligent model. It should decrease over time, not accumulate. What remains is simple: a powerful model in a powerful harness. Many, interacting, collaborating. No framework required.

**Isolated agents are not enough.** Access to the computer ≠ understanding. Give an agent 1000 documents: it searches, hopes, guesses. Multiply that by 50 agents without a shared world model and you get intelligence that is isolated but incoherent in combination. At enterprise scale, the information environment needs structure — a shared domain model, with the human in the loop.

**The symmetry.** The answer is to apply the same simplification on the data side. The model sits in a harness that gives it access to the computer; the data sits in an ontology that gives it structure and meaning. The ontology defines what exists, its properties, its relationships — the interface through which agents understand data. Two symmetric patterns: (powerful model + powerful harness) and (powerful data + powerful ontology).

**The Semantic Agent.** Their combination produces the Semantic Agent: (Model + Harness) + (Ontology + Data). It doesn't just generate, it starts to understand. Everything else is scaffolding — useful for a while, but bound to come down.

**What you own.** Everyone has access to the same frontier models; anyone can build a harness. That's commodity, and it's thinning out every day. What is NOT commodity: your ontology, your domain model, the structured and linked knowledge that captures how your organization understands the world. Frameworks are a transitional phase. Models are rented. The only thing left to build — and that you own — is knowledge.

## GrapheDeConnaissance

- Tony Seale —publie→ Semantic Agent post LinkedIn (DOCUMENT, 0.99)
- Tony Seale —est_instance_de→ The Knowledge Graph Guy (persona) (CONCEPT, 0.98)
- Semantic Agent —est_basé_sur→ Model + Harness + Ontology + Data (CONCEPT, 0.97)
- Harnais d'agent —converge_avec→ Ontologie (CONCEPT, 0.92)
- Frameworks d'agents —est_instance_de→ Phase transitoire (CONCEPT, 0.9)
- LangGraph —fait_partie_de→ Frameworks d'agents (TECHNOLOGIE, 0.95)
- CrewAI —fait_partie_de→ Frameworks d'agents (TECHNOLOGIE, 0.95)
- AutoGen —fait_partie_de→ Frameworks d'agents (TECHNOLOGIE, 0.95)
- Claude Code —est_instance_de→ Pattern modèle+harnais (CONCEPT, 0.92)
- Codex —est_instance_de→ Pattern modèle+harnais (CONCEPT, 0.92)
- Ontologie —permet→ définition de ce qui existe, ses propriétés et relations (CONCEPT, 0.95)
- Modèles frontier —est_instance_de→ Commodité louée (CONCEPT, 0.88)
- Ontologie —est_instance_de→ Seul moat non-commoditisable (CONCEPT, 0.92)
- Anthropic —affirme_que→ les composants du harnais encodent des hypothèses qui vieillissent vite quand les modèles s'améliorent (AFFIRMATION, 0.95)
- Tony Seale —affirme_que→ des agents isolés sans modèle du monde partagé produisent de l'incohérence en combinaison (AFFIRMATION, 0.9)
- Agents à l'échelle entreprise —est_basé_sur→ Modèle de domaine partagé (CONCEPT, 0.93)
- Tony Seale —recommande→ faire diminuer le scaffolding à mesure que les modèles s'améliorent (AFFIRMATION, 0.9)

---
Canonical: https://www.thekb.eu/en/fiches/seale-semantic-agent-model-harness-ontology-data-2026-04-17/