Ronin AI OS

Research notebook

Source-first research. Every claim is traced to the evidence behind it, sourced facts are kept visibly separate from model inference, and citations are never fabricated.

Model knowledge only — no live sources

Retrieval is unavailable, so this notebook is running on demo fixtures. Any answer here reflects the model's training knowledge, not current sources, and must not be presented as freshly cited fact.

Citation coverage

75%

3/4 claims sourced

Sources collected

3

2 rejected

Open conflicts

1

conflicting-evidence pairs

Search plan

Queries by stage, with the rationale for each.

  1. Seed
    grid-scale battery storage cost per kWh 2025

    Establish the current baseline figure and its primary source.

    done
  2. Expand
    lithium iron phosphate vs sodium-ion grid storage lifetime

    Compare competing chemistries to test durability claims.

    done
  3. Verify
    IEA 2025 battery storage deployment gigawatt figures dataset

    Cross-check deployment numbers against a second authority.

    running

Sources collected

Accepted into the evidence set.

Sources rejected

Excluded from citations, with the reason why.

  • Why sodium-ion already beats lithium everywheremedium.com

    Rejected: Opinion post on a self-publishing platform, no primary data, and over 18 months old.

  • Battery storage — encyclopedia overviewen.wikipedia.org

    Rejected: Tertiary source; used only to find primary references, not cited directly.

Claims and their evidence

Each claim shows support strength and source dates. Unsupported claims are labelled INFERENCE — model reasoning, not a sourced fact.

Conflicting evidence

Claims that assert opposing conclusions on the same topic.

  • A. Sodium-ion cells now exceed lithium iron phosphate cycle life at grid scale.

    B. Peer-reviewed data shows LFP retains higher cycle life than sodium-ion today.

    Conflict: same topic asserted "affirms" vs "denies"

Exportable report

Markdown you can copy out. Inference is flagged; nothing is dressed up as a source it does not have.

# Research report (DEMO)

> Model knowledge only — no live sources: Retrieval is unavailable, so this notebook is running on demo fixtures. Any answer here reflects the model's training knowledge, not current sources, and must not be presented as freshly cited fact.

Citation coverage: 75% (3/4 claims sourced)

## Findings
- [SUPPORTED] Average grid-scale battery storage cost fell below $150/kWh in 2025.
    - supports: Grid-Scale Energy Storage: 2025 Cost Review — https://www.nist.gov/energy/storage-cost-review-2025 (2026-05-10)
    - supports: IEA battery deployment tracker (open dataset) — https://ourworldindata.org/grid-battery-deployment (2026-06-30)
- [CONTRADICTED] Sodium-ion cells now exceed lithium iron phosphate cycle life at grid scale.
    - refutes: Long-duration battery chemistries compared — https://www.nature.com/articles/example-longduration-2025 (2025-11-02)
    - supports: Why sodium-ion already beats lithium everywhere — https://medium.com/@analyst/sodium-ion-beats-lithium-9f2c (2024-01-15)
- [SUPPORTED] Peer-reviewed data shows LFP retains higher cycle life than sodium-ion today.
    - supports: Long-duration battery chemistries compared — https://www.nature.com/articles/example-longduration-2025 (2025-11-02)
- [INFERENCE] Storage cost declines will likely continue at a similar pace through 2030.
    - Model inference — no source. Verify before use.

_All data above is labelled demo content; URLs use reserved example domains._

Demo fixtures — labelled example data, realistic https URLs only. This notebook never invents citations.