Jos De Roo

Eyelog and Eyeling: Recommended R&D Direction

Recommended direction: make Eyelog the primary R&D focus, while keeping Eyeling as a strategic Semantic Web and N3 compatibility layer.

A reasonable allocation would be approximately:

80% Eyelog
20% Eyeling

This allocation does not imply that Eyeling lacks value. Eyeling remains important for N3, RDF, Semantic Web compatibility, quoted formulas, graph-based reasoning, provenance, and continuity with the EYE tradition. The EYE ecosystem is centered on Notation3 and Semantic Web reasoning, including forward and backward chaining over N3 rules.

However, the strongest future opportunity appears to be broader than the Semantic Web. The main opportunity is symbolic reasoning as a compact, trustworthy component inside modern AI systems. In that setting, the reasoning engine should be easy to call, easy to inspect, easy to generate programs for, easy to embed, and easy to combine with subsymbolic systems such as large language models.

That favors Eyelog.

Eyelog is already positioned as a Prolog-style rule engine with an RDF bridge. That positioning is important: Eyelog can serve as the more general symbolic kernel, while RDF and N3 remain interoperability paths rather than constraints on every new research idea.

Core Distinction

Eyeling is Semantic-Web-native.
Eyelog is reasoning-kernel-native.

Eyeling starts from N3 and RDF concepts: triples, graphs, formulas, namespaces, Semantic Web syntax, and Web-style reasoning. This is powerful when the problem naturally lives in linked data, provenance, RDF vocabularies, or N3 rules.

Eyelog starts from ordinary symbolic logic-programming concepts: predicates, terms, facts, rules, queries, unification, recursion, lists, arithmetic, and finite search. This is often closer to how symbolic reasoning is needed inside AI systems, agent systems, validators, planners, scientific tools, medical-rule prototypes, legal-rule prototypes, and proof-producing assistants.

Strategic Assessment

The Semantic Web should be treated as an important interoperability ecosystem, but not as the main strategic bet. The original broad Semantic Web vision did not become the dominant public web architecture. RDF, SPARQL, ontologies, and knowledge graphs still have real value, especially in enterprise and research settings, but the wider AI opportunity is no longer dependent on convincing the world to adopt RDF or N3 as the main representation layer.

For neuro-symbolic AI, the more promising architecture is modular: neural systems handle perception, language, retrieval, approximation, and pattern recognition; symbolic systems handle exact reasoning, constraints, validation, explanation, proofs, and controlled search.

This makes Eyelog the better center of gravity for future work.

Recommended Architecture

A useful long-term architecture would be:

Eyelog as the core engine
The place for new reasoning algorithms, proof traces, search strategies, constraints, tabling, optimization, LLM tool integration, explainability, and neuro-symbolic experiments.

Eyeling as the N3/RDF layer
The place for N3 syntax, Semantic Web compatibility, RDF import and export, quoted graph reasoning, provenance-oriented examples, standards-facing work, and compatibility with the existing EYE lineage.

This split avoids two risks.

The first risk is making all new research depend on RDF and N3. That may slow adoption in AI settings where users want a small, direct, Prolog-like reasoning engine rather than a Semantic Web stack.

The second risk is abandoning the Semantic Web heritage too aggressively. That would lose a valuable niche, existing expertise, interoperability, and the distinctive EYE/N3 identity.

The balanced position is therefore:

Future research should primarily target Eyelog. Eyeling should be preserved and improved as a bridge, not used as the mandatory foundation for everything.

Suggested Eyelog Priorities

1. LLM-friendly symbolic reasoning

Eyelog should be easy for an LLM to generate, run, debug, and explain. Predicate-style rules are usually easier to synthesize than RDF/N3 triples.

Useful goals include:

• compact syntax;
• predictable error messages;
• explainable failure modes;
• examples designed for LLM tool use;
• automatic repair suggestions;
• stable JSON APIs for agent systems.

2. Proof traces and explanations

Every answer should ideally be accompanied by a compact proof object, dependency graph, or justification trail.

This would make Eyelog useful not only as an inference engine, but also as an explanation engine. In AI systems, this matters because neural systems can produce plausible answers without guarantees. Eyelog can provide a deterministic trace for the part of the task that needs exact reasoning.

3. Validation and guardrails

Eyelog can serve as a deterministic checker for LLM outputs.

Possible applications include:

• policy checks;
• medical-rule checks;
• legal-rule checks;
• data-consistency checks;
• workflow checks;
• mathematical sanity checks;
• authorization and access-control rules;
• compliance validation;
• structured-output verification.

In this role, Eyelog does not need to replace neural systems. It can constrain, verify, or explain selected outputs.

4. Neuro-symbolic tool use

Eyelog can become a small reasoning tool that neural systems call whenever exact inference is needed.

A typical pattern could be:

1. A neural model reads or retrieves information.
2. The neural model extracts candidate facts.
3. Eyelog checks rules, constraints, or consequences.
4. The result is returned with proof evidence.
5. The neural model explains the result in natural language.

This architecture gives each component a clear responsibility.

5. RDF/N3 bridge support

Eyeling should remain available for Semantic Web users, and translation between Eyelog and N3/RDF should become a major design goal.

The bridge should support:

• RDF import into Eyelog facts;
• Eyelog export to RDF or N3 when useful;
• N3 rule translation where possible;
• provenance-preserving transformations;
• compatibility examples showing the same reasoning task in both systems.

This allows the Semantic Web ecosystem to remain connected without forcing every new research path to use N3 syntax directly.

6. Embeddability

A small reasoning kernel should be easy to run in many environments:

• browsers;
• servers;
• notebooks;
• local-first applications;
• agent runtimes;
• command-line tools;
• CI pipelines;
• educational sandboxes;
• workflow engines.

This favors Eyelog as the primary kernel, because a Prolog-style language is easier to position as a general-purpose reasoning component.

7. Research clarity

Eyelog gives a cleaner space for studying reasoning itself without every experiment being shaped by RDF syntax and Semantic Web assumptions.

This is useful for research into:

• tabling;
• proof minimization;
• constraint solving;
• search control;
• finite model generation;
• aggregation;
• negation disciplines;
• probabilistic annotations;
• explanation formats;
• theorem-prover interoperability;
• LLM-guided rule generation.

Suggested Eyeling Priorities

Eyeling should continue to evolve, but with a clearer role.

Its strongest priorities should be:

• N3 conformance;
• RDF compatibility;
• Semantic Web examples;
• provenance and quoted formula reasoning;
• compatibility with existing EYE-style workflows;
• translation to and from Eyelog;
• documentation for users who need linked-data reasoning;
• standards-facing demonstrations.

Eyeling should not become a bottleneck for every new reasoning feature. Instead, features should be added to Eyeling when they strengthen N3/RDF use cases or when they expose Eyelog capabilities through a Semantic Web interface.

Positioning

The resulting positioning could be summarized as:

Eyelog should become the main platform for symbolic and neuro-symbolic reasoning research. Eyeling should remain the N3/RDF compatibility and Semantic Web interoperability layer.

This keeps the best of both worlds:

• Eyelog provides the broader AI-facing research path.
• Eyeling preserves Semantic Web strength, EYE continuity, and standards-facing credibility.

Final Recommendation

Future research and development should be concentrated mainly on Eyelog, because it is the better foundation for compact symbolic reasoning, LLM tool use, neuro-symbolic systems, validation, proof explanation, and embeddable AI infrastructure.

Eyeling should be preserved and improved as a high-value bridge to N3, RDF, Semantic Web standards, provenance, and the EYE tradition.

In practical terms:

Make Eyelog the kernel.
Make Eyeling the bridge.