| Project | Notes |
|---|---|
| Atlas-Modernization | App stack rewrite — core, protocol, SDK, plugins, command UI, simulations |
| DCS | Small personal Codex skills library — external review skills, architecture-map (0.14.0) |
| easymanet | Zero-touch OpenMANET provisioning and imaging |
| sidc-kit | Compact TypeScript toolkit for Symbol Identification Codes |
| Meshtastic-WIFI-bridge | Chunked, reliable Meshtastic transport with browser-over-mesh UI |
| Atlas-Mesh | Radio transport & mesh-routing lab under Atlas |
I work on systems where the map, the radio, and the backend have to agree — Go/TypeScript services, map consoles, mesh transports, and the small libraries that keep those pieces interoperable. Prefer things you can run locally, measure, and replay.
Main effort. Two repos: the application rewrite, and the radio/mesh lab underneath it.
flowchart TB
subgraph UI["Command surface"]
CI["Command interface<br/>map console · Cloudflare / Vite"]
SIM["Simulations<br/>local scenario workbench"]
end
subgraph APP["Application stack — Atlas-Modernization"]
SDK["SDK + asset runtime<br/>typed client · sync · telemetry"]
PROTO["Protocol<br/>schemas · contracts · validators"]
CORE["Core<br/>Go API · storage · object store"]
end
subgraph MESH["Transport lab — Atlas-Mesh"]
WEB["Replay viewer"]
MSIM["Deterministic simulation"]
MP["MeshProtocol"]
RAD["Radio"]
end
CI --> SDK
SIM --> SDK
SDK --> PROTO
PROTO --> CORE
CORE -. uses .-> MESH
WEB --> MSIM
MSIM --> MP
MP --> RAD
Atlas-Modernization — one workspace for the rewrite:
| Layer | Role |
|---|---|
| Core | Go HTTP API, durable storage, object store |
| Protocol | Schemas, generated contracts, validators |
| SDK & asset runtime | Typed client, sync, telemetry / command path |
| Atlas Link | Meshtastic Link service/SDK — Shared Picture, Tasks, radio transport |
| Plugins | Authoring runtime, first-party catalog, catalog releases |
| Core CLI | Default Ink action-list TUI for single-host atlas-core lifecycle and Plugin updates |
| Command interface | Map console on Cloudflare Pages / Vite |
| Simulations | Local scenario workbench + browser UI |
Recent focus: replaced Atlas Core release orchestration — tag reservation, one self-contained publication path, immutable sealed Release assets, and GHCR/npm reconcile without committing generated bundles to main. Core CLI hardened further: a shared active-work coordinator for interruption/cancellation and terminal-loss routing across lifecycle and Plugin flows, plus typed Plugin recovery outcomes (verified restoration, incomplete recovery/cleanup, unknown state) rendered the same in direct CLI and TUI. Helper consolidation across Meshtastic binary-io, CLI image validation, Plugin plan comparison, and Core coverage profiles. Prior window: shipped the action-list Ink TUI as the default atlas-core operator surface (lifecycle, diagnostics, Core/Plugin updates with recovery); verified-defect pass across Core, SDK, Meshtastic Link, simulations, and command interface — Task cancellation and Geofeature deletion, Argon2/MinIO object-restore hardening, Link queue-timer cleanup on settlement, map-source separated from MapLibre navigation. Continues independent Plugin releases (schema 4), Asset/Track movement history, and Meshtastic Link as the sole radio stack through Core v0.2.1; prior aggregate acceptance evidence still stands.
Atlas-Mesh — narrower question: how should bytes move between radios on an unreliable network?
Architecture is intentionally thin: Radio → MeshProtocol → Simulation → web replay viewer. Direct strategies (ack, stop-and-wait, …) sit beside routing experiments (gateway-tree, controlled flooding, on-demand, quality-tree). Simulator is seeded and deterministic, including a measured Heltec V3 LoRa airtime model.
Latest: consolidated radio verification lab.
Updated 18 Sep 2026