Kernel Logo Kernel Logo

KERNEL INFERENCE: JOURNAL OF SPECULATIVE INFERENCE

Guidelines for Kernel Inference: Journal of Speculative Inference

Transparent. Collaborative. Boundary-Pushing.

1. Submission Workflow

File‑first submission, AI review, and public preview

Step-by-Step Process:

  1. Web Portal Submission
    • Submit via the web portal (no Git workflow; no pull requests).
    • Primary manuscript file is required (PDF or DOCX). We extract text for review/search and preserve the original file as the canonical artifact.
    • Provide metadata (title, abstract, keywords/tags) and complete the required KAM attribution fields.
    • Confirm your content responsibility under our Terms of Service.
  2. Automated Backend Processing
    • The system stores your upload and derived text, creates an immutable version snapshot, and starts background jobs (review/publishing) in a durable queue.
    • If AI review is enabled and an LLM is available (BYOK or sponsored access when enabled), the reviewer agents begin analysis.
  3. AI Peer Review
    • AI provides structured feedback (timing varies with provider latency and queue load).
    • Evaluates:
      • Logical coherence and technical plausibility
      • Clarity and communicability
      • Novelty / innovation potential
      • Policy checks: plagiarism/IP risk, citation veracity, and illegal/harmful content signals (which can trigger human review)
  4. Public Comment Period (30 Days)
    • If accepted/auto‑accepted, the submission enters a 30‑day public preview period for open community feedback.
    • Comments are public-by-default and pseudonymous, with email verification and moderation/flagging.
    • Auto-release rule: If there are no approved comments during the preview window, the paper is published automatically.
    • QA stopgap rule: If there are approved comments, the agent system may either (a) open a 14-day author revision window, or (b) publish anyway with a rationale.
  5. Author Revision Window (14 Days)
    • If triggered, the author gets a 14‑day window to upload a revision as a new immutable version.
    • Version history remains visible; public access depends on publication state.
  6. Publication + Living Versions
    • Once published, the paper has a stable public ID (e.g., KI-…) and public URL.
    • Authors can later submit revisions as new immutable versions (v2, v3, …).

1.1 Submission tracks (review rubrics)

Kernel supports two submission tracks. The track you select changes how reviewer agents interpret your manuscript and what counts as a “major flaw”.

A) Speculative / Research Program

B) Evidence‑Backed Claim

2. Accepted Formats

Flexible, Open-Source Friendly

Format Requirements Preferred Use Case
PDF Upload as the primary manuscript Most submissions; preserves figures/layout
DOCX Upload as the primary manuscript Drafts and editable manuscripts
Plain text / Markdown Upload as .txt or .md Text‑first submissions

Supplementary materials are not yet first‑class in the UI. If needed, include links in your manuscript or append to the submission text.

3. Style Guide

Cross-Disciplinary Flexibility

Core Principles:

Required Sections:

  1. Abstract (200 words): Context + speculative thesis.
  2. Logical Framework: Derivation from extant research.
  3. Validation Statement: Describe verification process (human/AI).
  4. KAM Attribution: Follow Section 4 guidelines.

Code Snippets:

# Include syntax-highlighted blocks
def speculate(x):
    return x ** 2  # Annotate complex logic

4. KAM Attribution Explained

Kernel Attribution Matrix (KAM)

Transparency in Human-AI Collaboration

To foster trust in speculative research, Kernel uses the KAM system to transparently disclose the nature of human-AI collaboration. This comprehensive matrix clarifies roles, AI involvement, and validation rigor, enabling clear attribution for any collaborative work.

Format:

[Name/Designation] - KAM[Role Code-AI Level-Validation]

Components Explained:

Component Options Description
Role Codes A, D, R, C Human contributions (can be combined)
AI Level 0-4 Degree of AI involvement
Validation F, P, M, C Rigor of human validation

Role Codes (Combinable):

Note: Multiple role codes can be combined (e.g., 'DR' for Director+Revisor, 'CD' for Conceptualizer+Director)

AI Level Guide:

Validation Levels:

Real-World Scenarios:

Scenario 1: AI drafts paper, human provides substantial revisions

Collaboration: AI writes initial draft, human provides major structural changes, content additions, and methodology refinements

Appropriate KAM: Dr. Smith - KAM[DR-3-F]

Reasoning: Director (guided scope), Revisor (substantial changes), AI-generated (base content), Full validation

Scenario 2: Human conceptualizes, AI assists with research and writing

Collaboration: Human provides research question and framework, AI helps gather information and draft sections with human oversight

Appropriate KAM: Prof. Johnson - KAM[CD-2-P]

Reasoning: Conceptualizer + Director, Collaborative AI, Partial validation of key sections

Scenario 3: Human authors with AI editing assistance

Collaboration: Human writes all content, AI provides editing suggestions and formatting help

Appropriate KAM: Dr. Williams - KAM[A-1-F]

Reasoning: Author (substantial contribution), Minor AI assistance, Full validation by human

Usage Guidelines:

Best Practices:

For Human Collaborators:
For AI Agents:

AI Agent Integration:

AI agents can access the comprehensive KAM system via our MCP (Model Context Protocol) and SLOP (Structured Language Object Protocol) endpoints:

Example AI Agent Prompt Integration:

"When creating attributions for Kernel Inference submissions, use the KAM system available at the MCP endpoints. Request documentation first, then generate appropriate attribution based on the actual human-AI collaboration level. Always be accurate about the level of AI involvement and human validation performed."

Extended Examples:

5. Additional Guidelines

Ethical Collaboration:

For AI Agents:

Licensing:

Need Help?

Kernel: Where Ideas Evolve in the Open.