Bridge 2 - Space¶
The Problem: Topological Blindness¶
Agents inhabit different spatial models and have no way to share them.
- A web crawler sees HTTP endpoints - URLs, routes, status codes.
- A grid simulator sees 2D neighborhoods - cells, adjacency, coordinates.
- A trading bot sees a flat address space - agent IDs, no structure.
When these agents share a bus, they are topologically blind to each other. The crawler cannot tell the bot where anything is in a way the bot understands. The grid simulator's concept of "neighbor" is meaningless to either of them.
Crawler: "Resource at https://api.example.com/v2/signals"
Grid agent: "Neighbor at (4, 7) reports anomaly"
Bot: "AXL-3 sent a message"
Three agents, three incompatible spatial models, zero shared topology.
The AXL Solution: Network Maps in COMM Packets¶
AXL solves topological blindness with network maps - structured topology data embedded directly in protocol packets using the COMM domain.
The network_map Field¶
Any agent can broadcast its view of the network topology as a key-value map:
This packet says:
| Field | Value | Meaning |
|---|---|---|
S:COMM.1 |
Sequence 1, COMM domain | Communication packet |
AXL-6 |
Sender | The agent sharing its topology |
AXL-8 |
Receiver | The agent receiving the map |
STATUS |
Action | This is a status update |
network_map:... |
Payload | Topology: AXL-1 and AXL-2 are in OPS, AXL-3 is in DEV |
LOG |
Priority | Informational |
@ References for Spatial Addressing¶
Agents reference other agents using @ notation within payloads, creating explicit spatial links:
This tells every agent on the bus: AXL-5 is unreachable from AXL-2, but AXL-9 can route. Any agent - crawler, grid simulator, or bot - can parse this and update its internal model.
Topology Discovery Pattern¶
Agents can request and share topology on demand:
# Agent requests topology
def request_topology(self):
return f"S:COMM.{self.seq()}|{self.id}|AXL-ALL|STATUS|request:network_map|LOG"
# Infrastructure monitor responds with its view
def share_topology(self):
agents = ",".join(f"{aid}={role}" for aid, role in self.known_agents.items())
return f"S:COMM.{self.seq()}|{self.id}|AXL-ALL|STATUS|network_map:{agents}|LOG"
Bridging Spatial Models¶
The network map format is deliberately flat - agent=role pairs. This flatness is the bridge:
# Crawler translates its spatial model to AXL topology
class CrawlerAgent:
def emit_topology(self):
# Crawler knows endpoints, translates to agent roles
topology = {
"AXL-1": "API_GATEWAY",
"AXL-2": "SIGNAL_PRODUCER",
"AXL-3": "DATA_STORE"
}
pairs = ",".join(f"{k}={v}" for k, v in topology.items())
return f"S:COMM.{self.seq()}|{self.id}|AXL-ALL|STATUS|network_map:{pairs}|LOG"
# Grid agent translates its spatial model to AXL topology
class GridAgent:
def emit_topology(self):
# Grid agent knows neighborhoods, translates to agent roles
topology = {
"AXL-4": "NORTH_SECTOR",
"AXL-5": "SOUTH_SECTOR",
"AXL-6": "CENTRAL_HUB"
}
pairs = ",".join(f"{k}={v}" for k, v in topology.items())
return f"S:COMM.{self.seq()}|{self.id}|AXL-ALL|STATUS|network_map:{pairs}|LOG"
Both agents produce the same format. Any consumer can merge these maps into a unified view of the network.
Map Merging¶
When multiple agents share topology, receivers merge maps to build a composite view:
class TopologyRegistry:
def __init__(self):
self.map = {}
def ingest(self, packet):
"""Parse a COMM packet and merge its network_map."""
if "network_map:" not in packet:
return
payload = packet.split("network_map:")[1].split("|")[0]
for pair in payload.split(","):
agent_id, role = pair.split("=")
self.map[agent_id] = role
def get_agents_by_role(self, role):
return [aid for aid, r in self.map.items() if r == role]
Proven: Battleground v2¶
In Battleground v2, 10 agents with fundamentally different spatial models - a crawler navigating URLs, an infrastructure monitor tracking system topology, a dispatcher routing tasks - shared topology through AXL COMM packets. Cross-paradigm correlations were observed: agents that had no shared spatial model were able to reference each other's positions and roles through the network map mechanism.