\n \n Every multi-agent coordination framework starts from the same assumption:\n agents need to talk to each other. Sync starts from a different premise.\n \n\n \n Agents don't need to talk to each other.
They can share reality.\n \n\n \n {[\n { label: \"Traditional\", accent: false, items: [\"Message passing\", \"Orchestration layers\", \"Rigid schemas\", \"Predefined workflows\"] },\n { label: \"Substrate\", accent: true, items: [\"Stigmergic traces\", \"Self-activation\", \"Open-world state\", \"Emergent coordination\"] },\n ].map(col => (\n \n {col.label}\n {col.items.map(item => (\n {item}\n ))}\n \n ))}\n \n\n \n The insight comes from an old lineage: Nii's blackboard architectures (1986), Selfridge's\n Pandemonium (1959), Linda tuple spaces. But those systems required rigid schemas and\n deterministic programs. Language models change the condition — they tolerate open worlds, interpret\n partially structured environments, and act meaningfully within them. Stigmergy becomes computationally viable.\n \n\n \n The system behaves less like software and more like an ecosystem.\n A reef rather than a workflow.\n \n\n \n Read the full argument: The Substrate Thesis · What Becomes True · Introducing Sync\n \n

\n );\n}\n\n// ── Section: Algebra ─────────────────────────────────────────────────────────\n\nfunction AlgebraSection() {\n const ops = [\n { name: \"read context\", endpoint: \"GET /context\", desc: \"Returns everything: state, views, agents, actions with schemas, messages. The world as perceived.\", color: \"var(--green)\" },\n { name: \"invoke action\", endpoint: \"POST /actions/:id/invoke\", desc: \"The only write path. Built-in and custom. Every invocation logged to _audit.\", color: \"var(--red)\" },\n ];\n\n const endpoints = [\n { path: \"POST /rooms\", purpose: \"Create room\" },\n { path: \"POST /agents\", purpose: \"Join room\" },\n { path: \"GET /context\", purpose: \"Read everything\" },\n { path: \"POST /invoke\", purpose: \"Write anything\" },\n { path: \"GET /wait\", purpose: \"Block until condition\" },\n { path: \"POST /eval\", purpose: \"Evaluate CEL\" },\n { path: \"GET /rooms\", purpose: \"List rooms\" },\n { path: \"GET /agents\", purpose: \"List agents\" },\n { path: \"GET /docs/*\", purpose: \"Dynamic docs\" },\n { path: \"GET /\", purpose: \"Dashboard\" },\n ];\n\n return (\n

\n \n Every multi-agent system, regardless of domain, reduces to two operations.\n Everything else is wiring.\n \n\n \n {ops.map(op => (\n \n \n {op.endpoint}\n {op.name}\n {op.desc}\n \n ))}\n \n\n 10 Total Endpoints\n \n {endpoints.map(ep => (\n \n {ep.path}\n {ep.purpose}\n \n ))}\n \n\n \n The kernel reduction (reference/kernel.ts) expresses\n this full algebra in ~300 lines of pure TypeScript. No dependencies. The proof that the surface area\n is genuinely minimal. See Architecture docs · Σ-calculus\n \n

\n );\n}\n\n// ── Section: Room ────────────────────────────────────────────────────────────\n\nfunction RoomSection() {\n const [active, setActive] = useState(null);\n\n const primitives = [\n { id: \"state\", icon: \"state\", name: \"State\", color: \"var(--green)\",\n desc: \"Scoped key-value entries. _shared is public; agent scopes are private. Absence is signal — a key that doesn't exist yet is not an error, it's information. State grows by accretion.\" },\n { id: \"actions\", icon: \"action\", name: \"Actions\", color: \"var(--red)\",\n desc: \"Delegated write capabilities. Registered by Alice, scoped to Alice, invokable by Bob — Bob's invocation writes to Alice's scope using Alice's authority. Capability delegation without an auth framework.\" },\n { id: \"views\", icon: \"view\", name: \"Views\", color: \"var(--yellow)\",\n desc: \"Declarative CEL expressions projecting private state into public meaning. Registered dynamically, scoped to registrar's authority, evaluated lazily at read time. Interpretation as a composable runtime layer.\" },\n { id: \"agents\", icon: \"agent\", name: \"Agents\", color: \"var(--accent)\",\n desc: \"Humans and AI are equivalent participants, distinguished not by kind but by observation modality. MCP clients are first-class agents with embodiment identity.\" },\n ];\n\n const activePrim = primitives.find(p => p.id === active);\n\n return (\n

\n \n A room is the fundamental unit. Shared state observed by self-activating participants.\n No controller. No router. The room's structure is the protocol.\n \n\n \n \n Room Primitives — {active ? \"tap to deselect\" : \"tap to explore\"}\n \n\n \n {primitives.map(p => (\n setActive(active === p.id ? null : p.id)}>\n \n {p.name}\n \n ))}\n \n\n \n {activePrim\n ? <>{activePrim.name} — {activePrim.desc}\n : Select a primitive to see how it works\n }\n \n \n\n \n {[\n { scope: \"_shared\", desc: \"Public truth. Visible to all.\", color: \"var(--green)\" },\n { scope: \"alice\", desc: \"Alice's private scope.\", color: \"var(--accent)\" },\n { scope: \"_audit\", desc: \"Every invocation. Immutable.\", color: \"var(--dim)\" },\n ].map(s => (\n \n {s.scope}\n {s.desc}\n \n ))}\n \n\n \n CEL (Common Expression Language) is the universal evaluation layer. Enabled expressions,\n wait conditions, view projections, action preconditions — all CEL. One expression language\n for the entire system.\n \n\n \n {\"// view: project meaning from raw state\"}
\n status = state.alice.health {\"<\"} 20 ? \"critical\" : \"stable\"
\n {\"// action: enabled only when relevant\"}
\n enabled = state._shared.phase == \"combat\" && state.alice.health {\">\"} 0\n \n\n \n See CEL reference · API reference · Views reference · Surfaces design principles\n \n

\n );\n}\n\n// ── Section: Convergence ─────────────────────────────────────────────────────\n\nfunction ConvergenceSection() {\n const projects = [\n { name: \"ctxl\", tagline: \"Intelligence inside a component\", color: \"var(--red)\",\n desc: \"The interface remains alive to reasoning. Components don't use intelligence — they are intelligent. useReasoning hooks, self-modifying React.\" },\n { name: \"sync\", tagline: \"Intelligence across shared state\", color: \"var(--accent)\",\n desc: \"Shared state is the execution environment. Agents observe and modify a common substrate. The reef emerges from local rules.\" },\n { name: \"playtest\", tagline: \"Experience emerging from observers\", color: \"var(--yellow)\",\n desc: \"Game mechanics as bounded engines. Organs resolving constraints, emitting stable facts back into shared reality. 192 BGG-sourced mechanics formalized.\" },\n ];\n\n const layers = [\n { layer: \"Participants\", sub: \"Humans / Agents / MCP Clients\", color: \"var(--accent)\" },\n { layer: \"Surfaces\", sub: \"Declarative observers, self-activating\", color: \"var(--yellow)\" },\n { layer: \"Derived Meaning\", sub: \"Views, CEL projections, interpretation\", color: \"var(--purple)\" },\n { layer: \"State\", sub: \"Scoped key-value substrate\", color: \"var(--green)\" },\n { layer: \"Organs\", sub: \"Localized state machines (Playtest)\", color: \"var(--red)\" },\n ];\n\n return (\n

\n \n Three projects. One architecture. The convergence became clear: they all describe the same\n structure at different scales.\n \n\n

\n {projects.map(p => (\n \n \n

\n {p.name}\n

\n {p.tagline}\n {p.desc}\n

\n \n ))}\n

\n\n \n Unified Architecture\n {layers.map(l => (\n \n {l.layer}\n {l.sub}\n \n ))}\n \n\n \n When these conditions hold: new surfaces can be added without modifying existing ones, new actions\n introduce new realities without rewrites, agents and humans collaborate without coordination logic.\n Design shifts from scripting behavior to shaping conditions under which behavior can arise.\n \n\n \n Read the essays: What Becomes True · The Substrate Thesis · Isn't This Just ReAct?\n \n

\n );\n}\n\n// ── Section: Salience ────────────────────────────────────────────────────────\n\nfunction SalienceSection() {\n const domains = [\n { name: \"Auftragstaktik\", focus: \"Intent-based delegation\", color: \"var(--red)\" },\n { name: \"Bjork/Vygotsky\", focus: \"Scaffolded challenge\", color: \"var(--yellow)\" },\n { name: \"Lakatos/Popper\", focus: \"Falsifiable design\", color: \"var(--purple)\" },\n { name: \"Gibson\", focus: \"Affordance perception\", color: \"var(--green)\" },\n { name: \"Dorigo\", focus: \"Swarm intelligence\", color: \"var(--accent)\" },\n { name: \"Argyris\", focus: \"Double-loop learning\", color: \"var(--red)\" },\n { name: \"RWML/RAGEN\", focus: \"RL world models\", color: \"var(--yellow)\" },\n ];\n\n return (\n

\n \n The deepest theoretical move: sync rooms are structurally isomorphic to RL world models.\n Adaptive salience fills the missing reward signal role.\n \n\n \n \n The gap identified as most important: aggregating signals across the vocabulary\n into a room-level assessment — not just surfacing individual problems, but synthesizing\n a coherent field of relevance that guides agent attention.\n \n \n room state → observation │ action invocation → policy │ salience → reward signal\n \n \n\n Seven Grounding Domains\n\n \n {domains.map(d => (\n \n \n

\n {d.name}\n {d.focus}\n

\n \n ))}\n \n\n \n The demo infrastructure tests this with a stigmergic ant colony scenario — genuinely\n independent subagents (narrator, scout, analyst, architect) coordinating through\n environmental traces, not messages. Each agent's action modifies the perceptual\n environment for others. Understanding accumulates socially.\n \n\n \n Implementation Stack\n \n TypeScript/Deno on Val.town ·{\" \"}\n SQLite persistence ·{\" \"}\n CEL expressions ·{\" \"}\n MCP native protocol ·{\" \"}\n SSR React dashboard
\n Codebase decomposed from monolith to 10+ domain modules. 51 tests.
\n Single domain: sync.parc.land\n \n \n\n \n Read more: The Pressure Field · Examples · Help reference\n \n

\n );\n}\n\n// ── Section registry ─────────────────────────────────────────────────────────\n\nconst SECTION_COMPONENTS: Record JSX.Element> = {\n thesis: ThesisSection,\n algebra: AlgebraSection,\n room: RoomSection,\n convergence: ConvergenceSection,\n salience: SalienceSection,\n};\n\n// ── Main export ──────────────────────────────────────────────────────────────\n\nexport function Overview() {\n const [activeSection, setActiveSection] = useState(\"thesis\");\n const ActiveComponent = SECTION_COMPONENTS[activeSection] ?? ThesisSection;\n const sec = SECTIONS.find(s => s.id === activeSection)!;\n\n return (\n \n

\n /sync / overview\n Vision & Architecture\n A room-based multi-agent coordination substrate\n

\n \n \n {SECTIONS.map(s => (\n setActiveSection(s.id)}>\n {s.label}\n {s.subtitle}\n \n ))}\n \n

\n {sec.title}\n {sec.subtitle}\n \n

\n \n \n \n );\n}\n","deleted","label",[179,180,161],"frontend","components","permalink","https://www.val.town/x/c15r/sync/code/frontend/components/Overview.tsx","extension","actionData","errors"]