I build open tools for conservation, soil science, understanding mathematical statements, inspecting research evidence, and running auditable agent services.
Conservation and the tools needed to practice it are central to our open-source mission. We are building tools that communities can inspect, adapt and improve: organizing shared stewardship, comparing forest-management options, contributing soil observations, and making the supporting methods and evidence easier to understand.
Start with ProofLens: explore a mathematical statement, change an input, and see how its assumptions shape the result.
| Project | What you can do | First step |
|---|---|---|
| VergeCommon | Organize community conservation projects, document field evidence, and make group decisions with consent and access controls. | Find a first contribution. |
| Dendritic Forest Management (DFM) | Compare forest-retention options around roads and waterways with traceable inputs, maps, and portable reports. | Run the working application with a fictional watershed. |
| Global Soil Intelligence Project (GSIP) | Contribute privacy-conscious soil observations and explore open soil-mapping tools with explicit uncertainty. | Explore the project and contributor guide. |
| ProofLens | Turn supported Lean statements into visual explanations with explicit proof status and provenance. | Try the two-minute browser walkthrough. |
| Viridis Canon | Inspect public research sources, exact claim scopes, build records, and certification boundaries. | Read the current scope and reproduction guide. |
| Viridis Agent Fleet | Run open-source deterministic agent cores through MCP adapters and inspect their results. | Run an agent locally. |
Try the fictional woodlot showcase: connect soil observations (GSIP), community coordination and review (VergeCommon), and modeled landscape planning (DFM), with farmOS activity records and OpenTEAM data-rights vocabulary. The example demonstrates a manual handoff and labels proposed adapters, uncertainty and consent boundaries. No partnership or ecological validation is implied.
Download the walkthrough and builder packet.
| Project | What to explore |
|---|---|
| EnergyAI MCP | Explore energy-planning estimates and incentive sources through MCP and REST; inspect assumptions and eligibility limits. |
| Logosphere | Inspect local reasoning graphs, explicit hypotheses, provenance, and conditional Lean verification. Developer preview. |
| ViridisOS | Explore the open reference runtime for DOI-backed decision modules and recomputable results. |
| QuantaOptima | Authenticate explicitly logged agent action records and check audit-chain integrity. Published alpha. |
| Maxwell’s Defense | Inspect SHA-256 proof-of-work primitives for agent defenses and their stated model assumptions. |
| Viridis MCP Services SDK | Explore TypeScript/Python clients from source; hosted contracts and package publication have separate readiness gates. |
| Verdigraph Neurogenesis | Build deterministic, content-addressed agent configuration artifacts locally. |
| VulnCanon | Study bounded agent vulnerability classes with vulnerable/patched fixtures, detection rules, and mitigation guidance. |
The original Intelligence Bound Lean formalization and computational research materials remain reproducible references; active research development continues in Viridis Canon.
Our AI Scientist fork, TOFMapper fork, and Mathlib fork are upstream-derived work, with credit and contribution boundaries belonging to their respective projects. Browse the full public repository collection, including supporting experiments and resource collections.
Bring your own statement and save a reusable ProofLens explanation — a browser trial and a manual companion-page recipe for theorem documentation.
ProofLens is the easiest place to begin. Load a built-in example, change one parameter, and describe what changed and why. Then try a supported statement you want to understand.
If it helps you answer a real question, or you get stuck, share a short first-use report. A successful example or a precise failure gives us something concrete to improve.
- Build conservation tools as a commons: support local knowledge, consent, reproducible methods, and community participation.
- Make results inspectable: source, assumptions, and provenance should be easy to find.
- Keep evidence distinct: a numerical preview is not a proof; a successful build is not research certification or empirical validation.
- Credit contributors and upstream work, and turn useful feedback into clearer examples and fixes.
VergeCommon is an early cooperative pilot; DFM is a geometric screening pilot with separate experimental research tools; GSIP soil estimates do not replace laboratory tests. These tools do not by themselves establish ecological outcomes.
ProofLens source previews do not run Lean. Canon certification applies only to exact reviewed claim scopes; its historical sources remain separately labeled. Fleet's public agent cores are distinct from its hosted operations.
Bug reports, reproduction notes, documentation improvements, and bounded contributions are welcome. Follow each repository's contribution and security guidance.