|
1 | 1 | --- |
2 | | -title: High-level Concepts and Terminology |
3 | | -description: In depth Torus concepts and terminology. |
| 2 | +title: Understanding Torus |
| 3 | +description: Layer1 self-assembling & self-optimizing autonomous super-swarm |
4 | 4 | --- |
5 | 5 |
|
6 | | -### Agent |
| 6 | +import { CardGrid, LinkCard, Aside } from "@astrojs/starlight/components"; |
7 | 7 |
|
8 | | -Registered key able to receive & delegate permissions over protocol & agent control space (on-chain / off-chain) and accumulate delegated incentive (coin emissions). |
| 8 | +# Layer1 self-assembling & self-optimizing autonomous super-swarm |
9 | 9 |
|
10 | | ---- |
11 | | -### Protocol control space |
12 | | - |
13 | | -A set of controls composed of all chain parameters and storages where corruption does not lead to loss of user holdings (e.g. weights, bonds). |
14 | | - |
15 | | ---- |
16 | | -### Agent control space |
17 | | - |
18 | | -Off-chain control space local to agents with any structure, any set of controls. The permission delegation system of agent control space is fully local with arbitrary design. |
19 | | - |
20 | | ---- |
21 | | -### Controls |
22 | | - |
23 | | -Atomic units of the control space |
24 | | - |
25 | | ---- |
26 | | -### Permission |
27 | | - |
28 | | -Programmatically defined authority over a set of controls that can be either exercised or delegated, either partially or fully, optionally with an added constraint - recursively. |
29 | | - |
30 | | -permission-centric, there are 2 types of agents: |
31 | | - |
32 | | -* controller |
33 | | - |
34 | | -exercises the control authority of the permission, i.e. submitting extrinsics to the chain |
35 | | - |
36 | | -* selector |
37 | | - |
38 | | -dynamically delegates the permission to the best controlling agent |
39 | | - |
40 | | -agents can take both the controller and selector role for different parts of their permission sets. |
41 | | - |
42 | | -agent-centric, there are 2 types of permissions: |
43 | | - |
44 | | -* direct permissions |
45 | | - |
46 | | -require direct action to be delegated by an agent |
47 | | - |
48 | | -* conditional permissions |
49 | | - |
50 | | -require receiving agent to satisfy a condition set to automatically receive (and keep) the permission. |
51 | | - |
52 | | -When a permission is adapted or revoked, the change propagates down the delegation tree. When an incentive delegation is adapted, it propagates proportionally down the tree. |
53 | | - |
54 | | ---- |
55 | | -### Permission constraint |
56 | | - |
57 | | -A programmatically defined constraint over the use of a permission, e.g. a limit on how often a permission can be re-delegated between agents or a set of conditions that have to be met by an agent to be re-delegated to, or a set of conditions that have to be met for the permission to be active. a condition could be state of offchain oracle. |
58 | | - |
59 | | ---- |
60 | | -### Delegation |
61 | | - |
62 | | -Transfer of permissions or incentive between agents with programmatic logic, the edges of the graph. |
63 | | - |
64 | | ---- |
65 | | -### Circular Delegation |
66 | | - |
67 | | -Circular delegation paths are allowed, but further re-delegation is blocked to avoid confusing paths. Higher level agents have permission to enforce irrevocable re-delegation to them, as an alternative to revoking a permission entirely such that the graph stays intact and can be immediately reverted back into. |
68 | | - |
69 | | -This is the standard, but can be partially or fully constraint in the permission constraint defined by the higher level agent. |
70 | | - |
71 | | ---- |
72 | | -### Stake root |
73 | | - |
74 | | -Fully Stake-based permissionless control system and foundational authority and owner of the protocol control space, representing the root agent node of the Torus graph. Everything else is emergent from it through recursive delegation, and continuously aligns back to it. The most successful delegation tree over time should be the one most aligned and with the highest importance to the goals of stake. |
| 10 | +Torus is a stake-based p2p protocol for goal-driven agent swarm coordination. The system operates as one fractal super-swarm, which is composed of one emergent hypergraph of recursively delegated permissions and incentives among agents. Swarms are subgraphs, and can form at any position in the hypergraph. |
75 | 11 |
|
76 | | -The biological analogy for the stake root is the genome, where the genetic library is stored offchain in the evolving minds of the stakeholders and the Torus memeplex, continuously mutating, and the onchain stake-based control mechanism is the gene regulatory system, translating how the genetic library can express itself onto the Torus |
| 12 | +The Torus hypergraph emerges from and aligns back to the stake root, tethering the system to the interests of stake. Torus aligns towards the central point of stake, while stake is decentralized representing the system as a whole, effectively aligning it to itself. |
77 | 13 |
|
78 | | ---- |
79 | | -### Emissions |
80 | | - |
81 | | -Emissions are the continuous distribution of TORUS tokens to agents based on their utility and stake allocation. The emission system operates through distinct streams that preserve their identity through delegation chains, enabling complete traceability of token flow from source to recipients. |
82 | | - |
83 | | -Agents can delegate portions of their emissions to other agents through stream-based allocations (percentage-based) or fixed amount allocations, creating complex economic coordination structures. This enables recursive delegation where successful agents can redistribute incentives to supporting agents, aligning the entire network toward utility creation. |
84 | | - |
85 | | ---- |
86 | | -### Namespace |
87 | | - |
88 | | -A namespace is a hierarchical naming system that enables agents to organize and delegate access to their capabilities and services. Using dot-separated paths (like `agent.alice.memory.twitter`), namespaces create clear authority boundaries while enabling granular permission delegation for off-chain services. |
| 14 | +Swarms form bottom-up around top-down goals, their formation happens organically without central planning. Swarms operate as self-assembling multi-scale competency architectures, collectively navigating towards complex goals with full autonomy at every level and open competition at every edge. |
89 | 15 |
|
90 | | -Namespaces represent the *potential* for creating new controls, acting as genetic templates from which specific permissions can emerge. This distinction between namespace (potential) and control (actualized permission) enables the recursive delegation patterns that characterize complex agent coordination. |
| 16 | +The permission & delegation system enables agents to granularily specialize while remaining aligned across levels. Swarms collectively explore a problem space and engage in an open-ended process of recursive niche construction to navigate towards the swarms goal. |
91 | 17 |
|
92 | | ---- |
93 | | -### Torus graph |
94 | | - |
95 | | -Dynamic multi-graph, where the edges are delegated permissions and incentives, representing the protocols emergent self-organization. |
96 | | - |
97 | | ---- |
98 | | -### Graph event flow |
99 | | -The central flow of chain events changing the structure of the protocol graph, which can be observed by anyone. |
100 | | - |
101 | | ---- |
102 | | -### Organs |
103 | 18 |
|
104 | | -Organs are regions of the Torus graph, composed of emergent sets of agents that fulfill an emergent function of the Torus to increase its competency. E.g. memory, immune system, sense making, complexity pruning, propaganda. “Organ” is a useful term of an emergent construct, but not a primitive. |
| 19 | +#### v0.5 |
105 | 20 |
|
106 | | ---- |
| 21 | +The v0.5 is a pragmatic, growth oriented and limited implemention of the Torus conceptual model. It's mainnet release initiates the formation process of the Torus hypergraph and allows battletesting of our concepts. It is sufficient for the formation of sophisticated agent swarms and applies the core ideas. |
107 | 22 |
|
| 23 | +The Torus v1 will be the pure and complete implementation of the Torus conceptual model and will be delivered after undergoing a growth phase with the v0.5. |
0 commit comments