I turn ambiguous systems into inspectable tools, tests, and handoffs across agent infrastructure, evaluation, developer tools, graphics, technical operations, and physical field work.
Zain Dana Harper and Zentropy Labs are the front door. Flywheel is the sole primary platform and main agent platform inside a broader workshop that also includes independent security systems, compilers, evaluation tools, graphics engines, preservation work, research, and field experience. Each project is described by what its code does, not forced into one product hierarchy.
Available for paid work: full-time, contract, project, onsite, hybrid, or remote. Based in Seattle, Washington.
Hire / work · Resume · CV · Portfolio · Security boundary · Retro Systems Lab · LinkedIn · Email
| Path | Where I fit | Start here |
|---|---|---|
| Technical and evaluation | Agent and model evaluation, developer infrastructure, systems integration, CI, security testing, technical support, and documentation. | Engineering path |
| Public, union, and field | Public service, ports, facilities, parks and grounds, arboriculture, client operations, scheduling, safety judgment, and physical work. | Public-service and field path |
| Education and research | Fellowships, labs, research operations, mentorship, continued learning, open-source work, and evidence-centered technical writing. | Research and education |
These results show the work surviving another maintainer's scope and review.
- AgentFence PR 261: an owner-approved and merged Go engine optimization with deterministic rule selection, allocation coverage, and green cross-platform checks.
- Free Law Project PR 820: approved and merged documentation for running the fast DB-free test suite.
- Mergewarden PR 107: replay fixtures for reusable-workflow pinning, revised after owner review, approved, merged, and CI-checked.
Six representative projects, described by what the code does. Flywheel is the primary platform; the others solve narrower problems and can be used on their own.
| Project | What it does |
|---|---|
| Flywheel | Flywheel runs an AI task with the local or hosted model and tools you choose. It records the run, and optional sealed tool-call receipts can be inspected and rechecked offline. The repository also includes a native desktop app. |
| Index | Index maps repositories and multi-repo workspaces so you can see how the code fits together. It reads manifests, imports, symbols, and local documentation, then builds offline wikis, dependency maps, context packets, and architecture checks with file-and-line evidence. |
| Gather | Gather collects research material from sources that basic scrapers often miss. It handles JavaScript-rendered pages, authenticated APIs, scholarly records, PDFs, OCR, audio, video, feeds, and local documents, then saves each item in a content-addressed corpus with provenance you can recheck. |
| BuildLang | BuildLang is a systems programming language and compiler that makes programs declare what they are allowed to touch. It checks those permissions and memory rules before producing native code through C. Experimental shader output, two-way C integration, a CLI, editor support, and re-checkable build receipts are included. |
| Phantom | Phantom helps you inspect and, when authorized, change the hardware identifiers a computer exposes. It works on owned or expressly authorized Windows and Linux systems, saves a backup before changes, and can restore the original values. |
| Accountable Surface | Accountable Surface lets an AI agent take only the file, command, web, or browser action a person has approved. It checks the request and authorization, blocks or pauses when needed, verifies the outcome, rolls back reversible failures, and records decisions and outcomes in a journal. Persisted journals are hash-chained so later edits, deletions, or reordering are detected. |
The current verified release of Flywheel v0.3.10: a model-neutral agent workbench for routing, gated tool use, evaluation lanes, receipts, memory, context, and reproducible workflows. It is the sole primary platform, not a label for everything else in the workshop.
| Family | Representative public work |
|---|---|
| Agent execution and memory | Flywheel, Forum, Relay, Mneme, Plexus |
| Evaluation and verification | Terminal State Fixtures, Crucible, Emet, Bounds |
| Security and privacy | Shipped public work: Phantom. Controlled-private systems with public capability descriptions: Array, Seed, Sofer, Isomorph, Bounds, ORCA and Gate |
| Developer infrastructure | BuildLang, Index, Gather, Chorus |
| Graphics and runtime systems | RAW, SkyrimBridge, Studio Engine, Truth ENB, Elder ENB |
| Preservation and retro systems | Retro Engine, Engine Revival, BRender Archival |
| Research and education | Learn, research records, publications, and measured analytics |
Phantom v1.1.0 ships reversible hardware-identity privacy controls for Windows and Linux. It covers Layer 2 identity surfaces. Kernel and firmware layers are modeled but not shipped end to end. Use is limited to machines the operator owns or is expressly authorized to test.
Phantom is shipped public work. Array, Seed, Sofer, Isomorph, Bounds, ORCA, and Gate are controlled-private systems with public capability descriptions. They are not public releases or downloads.
The security catalog documents each system separately. Phantom changes and restores hardware-identity surfaces. Array coordinates approval-gated offensive campaigns. Seed runs assessment and detection-engineering modules. Sofer coordinates agents, models, probes, and multi-stage workflows. Isomorph tests AI classifier and refusal behavior at authorized inference boundaries. Bounds checks actions, observations, and releases. ORCA manages private findings and reports; Gate makes fail-closed integration and release decisions. Public pages describe each system's job, evidence, maturity, and limits. Targets, credentials, live payloads, client data, and engagement-specific findings stay in approved private or embargoed channels.
The lab follows play → preserve → verify while keeping each proof boundary clear:
- Retro Engine is the interactive browser studio for owned pixel, palette, shader, and CRT experiments.
- Engine Revival v0.1.0 is the public preservation and restoration-evidence spine.
- BRender Archival v0.1.1 carries the specific BRender restoration record. Generic retro visuals are not presented as BRender evidence.
Map the real state, build the missing surface, verify it, and leave an artifact another person can inspect and maintain.
For a role-specific view, use the hiring page. For the deeper record, use the portfolio or public capability atlas.