Living Systems — What's Running Now

Real-time status of the ecoPrimals sovereign mesh: active gates, deployed primals, capability routing, and live JupyterHub compute.

The Mesh Is Alive

This is not a description of future work. It is running.

NUCLEUS is LIVE on 3 gates. Provenance 7/7 validated on Linux and Windows. Sovereign CI automates push-to-deploy for 35 binaries across 3 platforms. sporeGate is 11/11 HEALTHY — the first clean gate health check. ZERO P0s. ZERO P1s. ZERO blocking P2s. gen4 is COMPLETE — gen5 begins: NUCLEUS as a platform serving real workloads.

Live gate mesh: sovereign compute nodes and their network connections
Live gate mesh: sovereign compute nodes and their network connections

Active Gates

GateStatusPlatformWhat’s Running
sporeGate11/11 HEALTHYLinuxSovereign CI LIVE, build authority, depot 35 binaries. biomeOS v4.51
eastGateOverwatchLinuxprimalSpring ( 1,312 tests), biomeOS + squirrel + petalTongue evolution
westGateNUCLEUSLinux13/13, 654 caps, 29 sockets, Provenance 7/7 COMPLETE. ZFS 25.4TB, 3,252 CAS objects
strandGateNUCLEUSLinux13/13, 1,742 caps, 674 IPC methods. RTX 3090, sub-ms GPU, 2,130 matmul/sec
blueGateNUCLEUSWindows13/13, Provenance 7/7 VALIDATED. 131.1 MB, TCP-only, DID key verified
ironGateOnlineLinux14TB+1TB+1TB+2TB. Takes esotericWebb from flockGate. Tower + HDD enclave
golgiBodyOnlineLinux (VPS)Depot (35 genomeBins), enrollment endpoint, Forgejo push mirror
northGateOnlineWindowsRTX 5090, AlphaFold source (~1TB)
grapheneGateOnlineAndroidTower LIVE (Pixel 8a), ADB mesh expansion
swiftGateHW ReadyWindowsAfter blueGate sub-builder stable
southGateHW ReadyLinuxOmada 10G, enrollment pending
flockGateDOWNLinuxRebooted, RustDesk locked out. esotericWebb → ironGate
fieldGateOfflineDead CMOS
biomeGateOfflineKernel recovery

Live Capabilities

When a gate starts Songbird, it announces its capabilities to the mesh. Other gates can then invoke any capability by name — songBird routes to the best available provider.

Currently Routed

CapabilityProviderPathUse
http.proxysporeGateLAN directHTTP routing to mesh services
peer.connectall meshed gatesbilateral TCPMesh peering, 0ms LAN
capability.callsporeGate → ironGateLAN directCross-gate compute dispatch
build.releasesporeGatelocalSovereign CI binary builds
cascade.syncgolgiWG15-min quorum cascade timer

Deploying

CapabilityProviderStatus
jupyter.executeironGateJupyterHub 5.4.5 LIVElab.primals.eco → 200
footprint.servesporeGateLIVEfootprint.primals.eco (200, 216ms)
esotericwebb.serveflockGateLIVEwebb.primals.eco (200)
ws.bridgesporeGateLIVE — petalTongue /ws JSON-RPC on :8080 (Wave 150g)
compute.gpuironGateRTX 5070 Ti ready, capability registration in progress
compute.cpustrandGateAwaiting hardware enrollment

JupyterHub — Live Compute

JupyterHub 5.4.5 is running on ironGate, serving at lab.primals.eco. The path is:

Browser → lab.primals.eco
    → bearDog :443 (ACME TLS)
    → songBird capability.call("jupyter")
    → ironGate :8000 (LAN direct, <1ms)
    → JupyterHub session

What makes this different from a cloud notebook: your computation runs on sovereign hardware in a private lab. No telemetry. No vendor. The mesh handles routing — if ironGate goes offline, songBird can route to strandGate (once enrolled) or any future compute node. The notebook doesn’t know which gate ran it.

Example Workloads Available

WorkloadHardwareSpring
16S metagenomics pipelineCPUwetSpring
GROMACS metadynamics (CAZyme FEL)RTX 5070 Ti GPUhotSpring
Salmon RNA-seq quantificationCPU + NVMewetSpring
STAR alignment (large genomes)64-core EPYC (strandGate)wetSpring
ET₀ irrigation modelingCPUairSpring
PK/PD compartmental modelingCPUhealthSpring

Each workload runs against the same infrastructure that produced the baseCamp results. Every run gets a provenance chain.

Sovereign CI Pipeline

All 15 primals are continuously built from source on sporeGate’s Sovereign CI. The pipeline:

Developer pushes to Forgejo (git.primals.eco)
    → golgi cascade timer (15-min quorum)
    → sporeGate pulls, builds x86_64-musl + aarch64-musl
    → BLAKE3 checksums computed
    → Binaries published to depot (membrane.primals.eco/depot/)
    → Gates cascade + pull from depot

Wave 133e result: 30/30 ecobins in pepti (15 x86_64 + 15 aarch64), all checksummed. 13/13 primals converged — zero CI workarounds, zero code debt. 4–5 binaries pending rebuild from latest source.

Mesh Health

The mesh is self-healing. If a gate goes offline:

  • Songbird detects peer loss via heartbeat timeout
  • Capability routing tables update across all remaining peers
  • Services that depended on the lost gate get routed to alternates
  • When the gate returns, peer.connect re-establishes bilateral trust

Key invariant: unplugging any single gate does not kill the network. The Flint edge router is the plasma membrane. Gates are ephemeral compute.

What’s Next

ItemWaveStatus
JupyterHub deploy132LIVE — JupyterHub 5.4.5, lab.primals.eco → 200
Pepti rebuild (5 stale binaries)134aNEXT — songBird, skunkBat, nestGate, coralReef, sweetGrass
WAN-DISPATCH-01 FULL PASS134aAfter pepti rebuild — songBird drawbridge committed
grapheneGate 13/13 from fresh pepti134aAfter pepti rebuild
bearDog CryptoProvider fix (UNIT-DIV-04)134bDONE — resolved, DNS live
DNS cutover: primals.eco → golgi (bearDog ACME)134bDONE — sovereign DNS live since Wave 100+
strandGate SSH enrollment134bPhysical access to House 2 needed
Live mesh visualization (petalTongue on golgi)134b+sporePrint host composition