The_Plans_From_WithIn_10k

Creator: WithIn Us Ai (WithinUsAI)
Version: 1.0
Size: 10,000 unique professional examples
Format: JSONL
License: MIT

Description

The_Plans_From_WithIn_10k is a frontier-quality dataset designed to train advanced autonomous agentic language models on professional planning, decomposition, sequencing, dependency management, long-horizon reasoning, and executable workflow design.

Every example is a complete, realistic, professional planning task drawn from real-world domains including AI research, software engineering, machine learning, DevOps, cloud infrastructure, startups, science, medicine, finance, cybersecurity, robotics, logistics, and more.

Each example contains rigorous professional reasoning inside <think> tags, a clear goal, an ordered executable plan, identified risks, and measurable success criteria.

Purpose

Train LLMs to master:

• Complex task decomposition
• Long-horizon planning and sequencing
• Dependency identification and management
• Risk-aware execution planning
• Professional workflow design across technical and business domains

Training Benefits

This dataset improves:

• Planning ability and structure
• Reasoning depth and foresight
• Execution reliability
• Risk identification and mitigation
• Long-horizon task completion accuracy
• Professional output formatting for agentic systems

Dataset Structure

JSONL file where each line is a JSON object with the following schema:

{
  "instruction": "string - The high-level planning request",
  "input": "string - Additional context or constraints",
  "output": {
    "thinking": "<think>Detailed professional reasoning trace demonstrating decomposition, evaluation, prioritization, and decision making</think>",
    "goal": "string - Clear, measurable goal statement",
    "plan": ["string", "string", ...] - Ordered, executable steps,
    "risks": ["string", "string", ...] - Key risks and mitigations,
    "success_criteria": ["string", "string", ...] - Measurable success conditions
  }
}

Schema Documentation

• instruction: Natural language request for a complex plan (e.g., "Design a production ML model deployment pipeline for a fintech startup").
• input: Optional additional constraints, resources, or context.
• output.thinking: Professional reasoning inside tags showing how the planner decomposed the task, evaluated options, prioritized steps, identified dependencies, and made decisions.
• output.goal: Specific, measurable goal the plan achieves.
• output.plan: Ordered list of concrete, actionable steps that can be executed sequentially or in parallel where appropriate.
• output.risks: List of realistic risks with implied or explicit mitigation in the plan.
• output.success_criteria: Clear, verifiable conditions that indicate the plan has succeeded.

Example Records

Example 1: ML Model Deployment Pipeline

{
  "instruction": "Design a production-grade deployment pipeline for a credit risk prediction model at a fintech startup.",
  "input": "Team of 4 engineers, AWS infrastructure, strict compliance requirements (SOC2, GDPR), model must retrain weekly on new transaction data.",
  "output": {
    "thinking": "<think>The goal is reliable, auditable, and compliant model deployment with weekly retraining. Key domains: MLOps, compliance, startup resource constraints. First, decompose into data pipeline, training, validation, deployment, monitoring, and rollback. Prioritize compliance logging and reproducibility. Evaluate CI/CD vs manual steps — CI/CD wins for speed and auditability. Identify dependency on feature store and model registry. Risk of data drift and compliance violation is high, so include automated validation gates and human approval for production. Success requires zero-downtime deployments and full lineage tracking.</think>",
    "goal": "Deploy and maintain a credit risk model in production with weekly retraining, full auditability, and zero compliance violations.",
    "plan": [
      "Set up versioned feature store and model registry on AWS SageMaker",
      "Build reproducible training pipeline with DVC or MLflow for experiment tracking",
      "Implement automated data validation and drift detection on new weekly data",
      "Create CI/CD pipeline (GitHub Actions + SageMaker) with automated testing and security scanning",
      "Add human-in-the-loop approval gate before production promotion",
      "Implement real-time monitoring for model performance and data drift with alerts",
      "Create automated rollback procedure triggered by performance degradation or compliance flags",
      "Document full lineage, compliance reports, and runbooks for the 4-person team"
    ],
    "risks": [
      "Data drift causing model degradation between weekly retrains",
      "Compliance audit failure due to missing lineage or approval records",
      "Deployment failure during weekly update causing service downtime"
    ],
    "success_criteria": [
      "Model retrains and deploys successfully every week with full lineage recorded",
      "Zero compliance violations in SOC2/GDPR audits",
      "Model performance stays above baseline for 4 consecutive weeks",
      "Rollback completes in under 5 minutes when triggered"
    ]
  }
}

(Additional 9,999 unique professional examples follow the same rigorous standard across all listed domains.)

Domain Coverage

AI, Software Engineering, Machine Learning, Research, DevOps, Cloud, Linux, Databases, Startups, Business, Product Development, Science, Medicine, Education, Finance, Infrastructure, Cybersecurity Defense, Robotics, Logistics, Manufacturing, Data Science, Writing, Analysis, Operations.

Quality Standards

• Exactly 10,000 unique examples
• No duplicates, near-duplicates, placeholders, templates, or low-effort content
• Every example is realistic, executable, and provides genuine frontier-level training signal
• Thinking traces demonstrate professional senior-level reasoning
• Plans are ordered, dependency-aware, and risk-conscious
• Suitable for training state-of-the-art autonomous agentic models

Deduplication Process

All examples were generated with aggressive semantic and structural deduplication. No two examples share the same goal structure, plan sequence, or domain-specific phrasing beyond necessary professional terminology.

Validation Process

Every example was validated for:

• Executability of the plan
• Logical consistency between thinking, goal, plan, risks, and success criteria
• Professional tone and structure
• Coverage of the target domains
• Absence of any forbidden content types

Recommended Training Usage

Use for supervised fine-tuning of agentic models on planning tasks. Preserve the full <think> reasoning for Chain-of-Thought training. Ideal for improving long-horizon planning, risk-aware decomposition, and professional output formatting in autonomous systems.

Mix with verification datasets (The_Verifications_From_WithIn_10k) for full plan-and-verify agent training.

Citation

@misc{withinusai2026plansfromwithin,
  title={The_Plans_From_WithIn_10k: Frontier Professional Planning Dataset for Agentic LLMs},
  author={WithIn Us Ai},
  year={2026},
  howpublished={\url{https://huggingface.co/datasets/WithinUsAI/The_Plans_From_WithIn_10k}}
}

License

MIT License. Free for research and commercial use with attribution.

Acknowledgements

Created by WithIn Us Ai (WithinUsAI) with the goal of advancing reliable autonomous agentic systems.

---

WithinUsAI — Building the highest quality datasets for frontier agentic intelligence.