Project summary

ChronoRealm is a local human–AI evidence workspace designed to make claims, corrections, and decision history traceable. I am seeking $10,000 for a defined validation and pilot milestone. The practical question is whether reviewers can detect unsupported claims and missing authorization more reliably using a structured evidence workflow.

Goals and method

Over eight weeks after funding and scope agreement, I will freeze a reproducible prototype and synthetic example tasks, reproduce available checks, seek external technical review within budget, and recruit 3–5 consenting evaluators. The pilot would compare review of comparable AI-assisted evidence packets with and without the structured workflow. Measures include unsupported-claim detection, citation-selection errors, authorization errors, review time, and reconstruction of corrections. This is exploratory research; no improvement is promised. Public outputs will include the evaluation protocol, synthetic examples, findings, limitations, and an itemized spending report. No private source code or confidential records will be published.

Use of funds

Engineering and reproducible packaging $4,000; external technical review $2,500; infrastructure and compute $1,500; human pilot and documentation $1,000; contingency and corrective work $1,000. These are proposed allocations, not vendor quotes. The minimum and target are both $10,000 to fund the described scope.

Team and track record

I am Austin Simpkins, an independent researcher and developer in the United States. I began ChronoRealm in early August 2026 and intend to work full-time on the funded pilot. The September 11 internal report records 39/39 suites passing and 2,596 tests counted, with Python/Rust agreement across 842 chain records and 70 grammar cases. Counts include skipped tests and overlapping checks; these results are not independent certification.

Risks and failure outcomes

The workflow may not improve reviewer accuracy or speed. Recruiting evaluators or securing an external reviewer within the proposed budget may fail. Native GUI/full live acceptance and formal projection soundness remain unresolved. I will report negative results and gaps rather than describe internal tests as independent validation. The pilot does not establish frontier-model safety or production readiness.

Funding

I have no secured project funding. I am applying for grants and in-kind API credits; no award is assumed. Any additional support will be reconciled against this scope to prevent duplicate funding of the same costs.

Public brief and test summaries: https://austinsimpkins1997-ship-it.github.io/ChronoRealmFunding/funding-brief.html

LinkedIn: https://www.linkedin.com/in/austin-simpkins-692a533b2/