Open source · NeoZorK Research

MIT · PyPI · MCP registry

Monte-Neo

The independent verifier for trading strategies written by AI agents and humans.

A coding agent can turn a trading idea into a backtest in minutes, and the headline number is usually wrong for a few repeatable reasons: look-ahead bias, missing costs, and selection among hundreds of variants. Monte-Neo checks the backtest rather than the idea. Given price data and strategy code or positions, it returns one of four verdicts, the checks behind it, concrete next steps and a reproducible, optionally signed certificate. It ships a command line, a GitHub Action and an MCP server so that AI agents can call the verifier directly.

First commit 23 January 2026
Python; MIT licence; PyPI
Verifies process, not profitability: no performance claims

Source · Documentation · PyPI · DOI (Zenodo)

Roadmap. In development, not yet released: a phantom-profit check for Monte-Neo, which tests whether an apparent backtest profit exists only inside the backtest. It is planned alongside the author's first paper, which is not yet published. This is a plan, not a shipped feature.

Apache-2.0 · PyPI · conda-forge

ClaimBound Evidence

Where is the evidence? Evidence cards for narrow public AI, ML and data claims.

ClaimBound turns one narrow public claim into a small evidence card: protocol, source boundary, hashes, exact result status, claim boundary and reproduction level. It is not a leaderboard, a hosted scoring service or a certification authority; it is a toolkit and an open registry that anyone can rerun locally. Negative, blocked and drifted outcomes are first-class evidence. The project is explicit about its own limits: a green status means one narrow claim passed under a frozen protocol, not that it was independently reproduced, and most public cards are still single-operator.

Started 29 April 2026 as OpenWave; renamed ClaimBound on 1 May 2026 (git history)
Python; Apache-2.0; PyPI and conda-forge
OpenSSF Best Practices badge; CITATION file included

Source · Pages · PyPI · conda-forge

Not affiliated with claimbound.com. ClaimBound Evidence is independent open-source work and has no relationship with claimbound.com or any commercial product or service that uses a similar name.

Honest context

Related and parallel work

EviBound

R. Chen, arXiv:2511.05524, submitted 28 October 2025

A governance framework for autonomous research agents: a pre-execution approval gate and a post-execution verification gate, with claims accepted only when backed by a queryable MLflow run, required artifacts and a finished status. Evaluated on eight tasks.

Difference. EviBound asks whether an agent finished with valid artifacts. ClaimBound asks what exact public claim is supported, by which frozen protocol and hashed sources, with which status, and where the claim stops. Different object (agent execution versus public claim), different substrate (MLflow runs versus source boundaries and hashes), different outcome space (verified or blocked versus passed, negative, blocked, insufficient, drift and reproduced). ClaimBound is not affiliated with, a fork of, or derived from EviBound; the paper was discovered after ClaimBound development had started.

Claim Boundaries for Execution Evidence

M. Sergeev, IETF Internet-Draft draft-sergeev-claim-boundaries-02, 15 October 2026

An individual draft arguing that the strength of a claim that an action was authorized or performed is limited by what the producing boundary actually observed and who controls it.

Difference. Adjacent in spirit, different in subject: it concerns execution and authorization evidence in agent protocols, while ClaimBound concerns reproducible checks of public AI, ML and data claims.

Not a novelty claim. Similar names and adjacent ideas exist; each is named, dated and distinguished above.

To avoid confusion

Names and affiliation

NeoZorK is a software-engineering and research handle. It is not affiliated with any video blogger, streamer or gaming channel that uses the same or a similar name.

ClaimBound Evidence is independent open-source work. It is not affiliated with claimbound.com or with any commercial product or service that uses a similar name.