# Simone Consumption Engine: implementation note

September 23, 2026. Author-reported source inspection by Simone Systems Research. Jonathan Simone leads development with AI assistance.

## Scope and availability

Simone Consumption Engine is in development. This note reports an inspection of selected prototype code for source preservation, retrieval and task records. The underlying prototypes are private. This public description is not a source release, a runnable demonstration or independent reproduction.

The existing software preserves research materials, organizes source records, retrieves claims with supporting passages and records how evidence is used in a task. These capabilities currently have separate interfaces. The complete product workflow, from preservation through recovery, remains to be demonstrated publicly.

## What the inspection supports

| Capability | Existing implementation | Current limitation |
|---|---|---|
| Preserve sources and provenance | The preservation prototype stores original documents, content hashes, source observations and extracted text. | Source identity and storage history do not establish claim accuracy or permission to redistribute a source. |
| Organize and retrieve | The catalog provides searchable research summaries and document passages. A separate retrieval component connects queries to recorded claims and their evidence. | The integration currently uses a fixed catalog sample. Connecting the full document collection remains in development. |
| Keep claims connected to evidence | The retrieval prototype carries source references, passages and recorded qualifications, with checks for changed or missing evidence. | Hash and passage checks establish source identity and confirm that the text appears in its source. Whether it supports a conclusion requires substantive review. |
| Retain uncertainty and corrections | Existing records can expose uncertainty, recorded conflicts and correction history. The retrieval prototype excludes claims marked as withdrawn or superseded. | Automatic contradiction detection and correction propagation through every downstream output are not established. |
| Recognize duplicate material | Source hashes and claim-level matching provide building blocks for duplicate handling. | Preventing dependent sources from inflating corroboration across the integrated product remains to be demonstrated. |
| Record use and checks | The task-record component links selected evidence, outputs, check reports and recorded review decisions to the specific files used. | It validates records. The task check and substantive review must be performed outside the validator. It does not authenticate reviewer identity. |

These are implementation observations, not measured accuracy, speed, reliability or learning benefits. No new experiment or complete product run was performed for this report. Source inspection does not establish that a workflow was executed successfully.

## Preservation and recovery

Existing recovery tooling and a retained same-host rehearsal provide a foundation for further work. That historical record does not qualify recovery on a clean machine or operation with network access disabled at the operating-system level. Local storage, private deployment, offline operation and recoverability require separate checks.

## Next demonstration

The next milestone is one reproducible workflow using an owned or synthetic corpus and a recurring question. It will cover source preservation, duplicates, evidence and uncertainty, retrieval, a checked task outcome, and recovery into a clean environment. It will include a correction and an insufficient-evidence case, with manual preparation identified explicitly.

The restored workflow must use its restored data rather than fall back to original source locations or workstation caches. Actual outputs, check results, failures and a non-authoring review must remain inspectable. General retrieval quality, durable learning and complete offline deployment remain unestablished by the current components.

## Supporting software

BigBoss is a separate public local approval tool with its own implementation and test evidence. Council contributes claim-review research; SEED contributes evaluation methods. These roles do not establish completed integration with Consumption Engine. BigBoss commands and test results apply to BigBoss.

- [Consumption Engine and current implementation](https://simoneresearch.com/product/consumption-engine/#current-implementation)
- [Evidence and supporting research](https://simoneresearch.com/evidence/)
- [Developer availability and BigBoss quickstart](https://simoneresearch.com/developers/)