Preserved public research papers, mathematical specifications and reproducibility artifacts by Shafaet Brady Hussain and identified contributors. GitHub is the durable publication location; former hosted services are not required to read this repository.
This repo is designed for clean, citeable public work:
- source-ledger first
- claim/evidence graph before synthesis
- careful academic wording
- no private policy services, prompts, server credentials, or key material
- generated working papers ready for human review
- OpenZero: independent self-hosted AI runtime, local improvement and training workspaces, and optional operator-configured peer exchange.
- ZeroThink Local 1.0: new Apache-2.0 open-source research engine and standalone CLI, with offline evidence maps, nine research processes, selected-source ledgers and optional bounded draft, critique and revision passes. Download v1.0.0; no company endpoint, account or hosted database is required.
- ZMath Local 1.1: standalone file/message encryption, the dual-key notes and attachment vault, preserved signed device-envelope research and public QuantumEncryption1 website material. Download v1.1.0. Newly released author-owned code uses PolyForm Noncommercial 1.0.0; research already published under CC BY 4.0 retains that license.
- ZERO ONE Desktop 8.0: the local ZeroThink integration is published in source and as direct Windows x64, macOS Apple Silicon and Linux x64 downloads, with release and submission evidence. Microsoft completed certification and publication: Partner Center confirmed the latest product is available, and live Submission3 offers version8.0.0.0 on 30 September 2026. Microsoft Store listing. A Store-signed installation on this PC has not been verified.
- Preservation catalog: project locations, source boundaries and archived-service status.
- Published release audit, 30 September 2026: verified Store versions, current GitHub downloads, cross-platform repair and free static-site update evidence.
- Formatted paper archive: 11 previously public manuscripts preserved as original PDF and DOCX pairs, with individual rights and SHA-256 provenance.
- Download the offline public preservation kit: five reviewed public source snapshots, 612 committed files, a start guide, source/license records and local verification checks. Keep your own copy; original licenses and paper disclosures remain intact. Kit checksum.
- Benchmark Results: preserved public benchmark snapshots.
- TalkToAI: static public project hub; no central runtime is required.
Historical papers may contain dated references to CallChat, ZeroThink, ResearchForumOnline or QuantumEncryption1 hosted services. Those references are part of the original evidence record, not a statement that the services are currently available.
The new ZeroThink Local source release is distinct from the historical ZeroThink service papers below. Source-ID checking and repeated model revision do not establish factual truth, scientific novelty or peer review. The dual-key vault uses standard AES-256-GCM layers; two layers do not constitute AES-512, and its small visual-pattern space does not replace a long unique passphrase. Preserved experimental signed-envelope code is not a live mail service or an independently certified protocol.
Boundary Completion and Cycle Memory in Zero Boundary Algebra by Shafaet Brady Hussain. The index links the original 17-page PDF, covering guarded phase transitions, endpoint algebra and cycle memory, validator equivariance, and auditable histories. The supplied PDF is preserved without modification; it identifies itself as working research draft 0.1 and contains an AI-assistance disclosure. Claims and scope limits are described in the paper.
Reset Compatibility in Zero Boundary Algebra — Six-Element Transformation Monoids and Conservative Validator Synthesis, by Shafaet Brady Hussain. Complete proofs classify the reset–mirror monoid, refine the validator contract, and quantify conservative repair. Includes LaTeX manuscript and reproducibility artifacts: 496 small validator cases and 248 feasible synthesis inputs passed. Working paper with AI assistance disclosed; no global-priority or external-peer-review claim.
Five author-led engineering working papers by Shafaet Brady Hussain examine distinct, testable parts of a local and self-hosted coding agent. They are source-grounded preprints with explicit limits, not peer-reviewed claims of AGI or superiority over other assistants. The frozen evidence ledger links the public source commit, file hashes, release and cross-platform workflow.
| Paper | Research question |
|---|---|
| Outcome-First Task Navigation | Can bounded Desktop discovery and privacy-aware public-doc preflight reduce setup work without leaking request text? |
| Bounded Long-Session Agents | How should an agent distinguish continued activity from observed progress and verified completion? |
| Tool-Protocol Reliability for Local Models | Where do model-to-tool interactions fail, and how should protocol fidelity be measured? |
| Reversible Recursive Source Improvement | When does an isolated source candidate qualify as a checked or measured improvement? |
| Release Evidence Boundaries | What evidence supports source, installer, platform, runtime and task-outcome claims separately? |
The papers disclose AI assistance and the author's association with the evaluated application. Each states what the current release does and what a future controlled study would need to establish.
| Paper | Focus |
|---|---|
| Zero Boundary Algebra 1.1: Formal Specification | Expanded typed algebra, executable verifier, property results, encryption-profile audit, privacy/governance analysis, and falsifiable evaluation protocol |
| Zero Boundary Algebra 1.0: Formal Specification | Preserved first formal release |
| ZeroThink 1.0: A Reproducible Architecture for Evidence-Gated AI Agent Services | Consolidated architecture, dated positive and negative evidence, executable release-invariant tests, and falsifiable baselines |
| ZMath Shield and ZME1 1.0: Authenticated Evidence Containers | Standard-primitive envelope specification, frozen source hashes, regression evidence, threat model, and structural negative vectors |
| Zmail and CallChat 1.0: Layered Protected Communications | Evidence-bounded email, Matrix messaging, object, device, and call-media protection with explicit downgrade rules |
| OpenZero 1.0: A Local-First Agent Runtime | Frozen runtime architecture, serving-path failures, offline boundaries, and explicit tool-authority risks |
| OpenZero Model and Dataset Methodology 1.0 | Deterministic private-corpus curation, prompt-group isolation, QLoRA experiment gates, and rights-bounded release controls |
| ZeroMint AIOS Engineering Evaluation 1.0 | Live distribution metadata, split-image and torrent integrity checks, installer boundaries, and supply-chain gaps |
| Cymatics Zero Platform Evaluation 1.0 | Synthetic-corpus and browser-audio audit, live count inconsistency, and generator reproducibility failure |
| Quantum-Ready Evidence Graphs and QPU-Factor Evaluation 1.0 | Exact classical claim-graph baseline, greedy comparison, and synthetic-fixture QPU-factor contract boundary |
| Probability of Goodness Decision Routing 1.0 | Frozen substring-score audit, eight synthetic counterexamples, and hard-rule-first routing specification |
| Bio-Inspired Recursive Computation 1.0 | Frozen genetic-formula audit, bounded recurrence, and explicit metaphor-to-biology evidence boundaries |
| Research-Corpus Curation and Claim-Risk Benchmark 1.0 | Text-free metadata audit, title-signal review queue, and synthetic rule-specification benchmark |
| Zero Boundary Algebra as a Provenance Workflow | Public mathematical/workflow framing |
| ZeroThink as a Sovereign Reasoning Layer | Research writing and audit workflow |
| OpenZero and Local-First AI Nodes | CPU-friendly sovereign agents |
| ZMath Shield and Portable Evidence Containers | Public behaviour spec, no private encryption code |
| Shafire and OpenZero Local Benchmark Snapshot | First CPU/API benchmark snapshot for Shafire models and OpenZero |
| Spectra8, TalkToAiQ, and SpectraMind OpenZero Benchmark | Requested benchmark pass for Spectra/TalkToAiQ/SpectraMind artifacts |
| Separating Model Artifact Performance from Serving-Path Failure | Methods-and-results preprint on local-route template contamination and reproducible exact-answer testing |
| ZeroThink Public System Benchmark | Public live-route, API-guard, CLI-device-flow, and OpenZero integration benchmark |
| Boundary-Oriented Evaluation of a Deployed AI Agent Service | Methods-and-results preprint built from the reproducible public-surface benchmark; includes limits and conflict disclosure |
| ZeroThink Authenticated Intelligence Benchmark | Account-routed exact-answer comparison across ZeroThink/OpenZero and saved provider lanes |
| Botanical Formula Research Roadmap | Safety-first research plan |
The generator crawls public forum pages, indexes first-post text, filters likely spam/noise, marks third-party posts separately, then applies the public ZeroThink Paper Creator protocol.
Run:
python tools/build_research_release.pyRead PUBLIC_RELEASE_BOUNDARY.md before adding material. This repo publishes research and reproducibility artifacts. The separately reviewed ZMath source release is linked above; private prompts, credentials, customer data and server policy internals remain excluded. Historical papers retain their original licenses and claims boundaries.