Agent systems that leave receipts — govern, prove, remember.
| Repository | Role | Purpose |
|---|---|---|
Agnet |
PROVE | Portable accountability infrastructure for agent identity, signed tasks, durable Swarms, artifacts, receipts, and independent verification. |
Aetherion |
GOVERN | Local-first agent harness kernel for policy, memory, event ledgers, capabilities, and governed tool execution. |
Anamnesis |
REMEMBER | Proof-aligned PyTorch research scaffold for budgeted long-context memory across recurrence, depth routing, and hashed Engrams. |
CodeNexus |
TEACH | Multi-tenant educational online judge with sandboxed execution, classroom workflows, contests, and AI-gated analysis. |
NEW-Finding-Shortest-Paths-Algorithm |
OPTIMIZE | C++ implementation of BMSSP Algorithms 1–3 and Lemma 3.3, benchmarked and correctness-checked against Dijkstra. |
Agent systems · 2 modules
| Module | Purpose |
|---|---|
Agnet |
Agent identity, signed work, durable Swarms, evidence closure, and independent verification. |
Aetherion |
Human-governed local runtime for agent permissions, memory, capabilities, and tool execution. |
Research and education · 3 modules
| Module | Purpose |
|---|---|
Anamnesis |
Budgeted long-context memory research across recurrent, depth, and sparse lookup paths. |
CodeNexus |
Rust and React online judge for secure educational code execution and classroom workflows. |
NEW-Finding-Shortest-Paths-Algorithm |
BMSSP paper implementation with Dijkstra baselines, strict correctness checks, and benchmarks. |
GitHub · Blog · Research DOI · Email


