Checking graphicsroot —
Current developmental stage
Infant Kernel
The kernel is ready. Add knowledge, begin learning, and its memory structure will start to mature.
2% dense · 98% sparseArchitect sleeping
Living system
A brain that grows from what you give it.
Teach it with your own files and text. OneBrain stores its learned memory locally, adapts its routing, and makes every boundary inspectable.
Live capacity
Observed kernel activity, not estimated cloud capacity.
Memories
0
persistent records
Knowledge tokens
0
locally indexed
Experiences
0
learning events
Synaptic activity
0
events per second
Maturity score
0.00
Autonomous architect
Continuously balances reach, focus, memory pressure, and proofability.
Policygrow
Routing2 / 98
Corpus pressure
0%
Orthospace pressure
15%
Consolidation
5%
Learning signals
Strongest active mechanisms.
Recent activity
Latest meaningful events.
Conversation
Ask what it knows. Teach what it does not.
Answers are grounded in OneBrain’s locally stored memory. Retrieved context is shown beside every response.
OneBrain dialogue
Local retrieval, synthesis, and memory formation.
Ctrl / ⌘ + Enter to send
Knowledge ingress
Give OneBrain a world to learn from.
Import text-based files, paste source material, or fetch a CORS-permitted URL. Processing is local and incremental.
Import knowledge
TXT, Markdown, JSONL, CSV, XML, HTML, and source code.
Drop knowledge files here
Files are chunked, vectorized, projected, and stored locally.
Nothing is uploaded by this page.
0 characters
Knowledge sources
Most recently imported material.
Corpus routing map
Reference adapters only. OneBrain never silently downloads external corpora.
Mathematical fabric
Structure is part of the computation.
OneBrain uses deterministic cut-and-project coordinates to route activity, place memories, open dense basins, and preserve sparse global reach.
m ∈ ℤᴰ → z = π∥(m), ω = π⊥(m)z governs physical/eigenspace routing. ω governs Orthospace acceptance, phason pressure, pruning, and persistence placement.
TriFabric lanes
Each lane has a distinct operational responsibility.
01 · 8-FOLD
Ammann–Beenker
Retrieval and ingestion routing. Optimized for broad associative reach and stable shard placement.
Live acceptance 0.00
02 · 10-FOLD
Penrose
Active compute and dense local microtiles. Opens around hot, coherent reasoning basins.
Dense allocation 0.0%
03 · 12-FOLD
Dodecagonal
Memory persistence and replay. Prioritizes durable placement and consolidation continuity.
Consolidation 0.0%
Live invariants
Observed geometric health.
Window accept
0.00
Basis error
0e+0
Shard skew
0.00
Pruned routes
0
Architect tuning
Bound the density ceiling without changing the global virtual address space.
42%
Neurological capability
Neurological visualizations, benchmarks, and test evidence.
Show the system thinking, not just talking. These views expose synthetic neural dynamics, signal bands, connectivity structure, and repeatable local benchmark tests so OneBrain can genuinely demonstrate capability.
Neurological activity views
Synthetic but state-grounded local views of activation, oscillation, and functional structure.
Cortical band fielddelta · theta · alpha · beta · gamma
Spike rasterpopulation burst timing
Connectivity matrixfunctional coupling estimate
Delta0.00
Theta0.00
Alpha0.00
Beta0.00
Gamma0.00
Coherence0.00
What this proves
OneBrain can surface internal state as interpretable visual telemetry rather than a black box chat shell.
Boundary
These are synthetic local neurological views driven by OneBrain state variables. They are not claims of biological measurement or clinical neuroimaging.
Benchmark summary
Repeatable device-local capability checks.
Retrieve latency
—
Median query roundtrip over local memory.
Projection rate
—
Cut-and-project operations per second.
Vector throughput
—
Tokenize + vectorize workload per second.
Visual refresh
—
Observed local animation cadence.
Benchmark details
Latest results with notes and context.
Run the suite to populate measured results.
Capability tests
Operational checks that verify the system is behaving as designed.
Retrieval grounding
Checks that the latest answer is based on locally retrieved memory rather than ungrounded generation.
Geometry determinism
Checks repeated cut-and-project runs for stable acceptance and routing statistics.
Persistence continuity
Confirms memories, checkpoints, proofs, and settings survive local lifecycle operations.
Visual telemetry
Confirms the lab canvases continue to reflect changing kernel state as activity and memory pressure evolve.
Distributed WebGPU fabric
Scale compute coverage across workers, tabs, and the service-worker scope.
The service worker is the resilient control plane: it owns coverage, caching, discovery, and short event-bounded GPU tasks. Sustained compute runs in recoverable dedicated WebGPU workers across every participating OneBrain tab.
Coverage control plane
Scale deliberately. One service-worker scope coordinates many recoverable GPU execution lanes.
Secure contextCheckingHTTPS or localhost is required.
Service workerRegisteringScope not established.
SW WebGPUUnknownCapability probe pending.
Dedicated WebGPU0 / 0Long-running recoverable compute lanes.
Controlled tabs0Pages inside this service-worker scope.
Logical processors—Browser-exposed scheduling capacity.
Four-layer coverage architecture
Each layer has a different lifecycle and responsibility. Scale comes from composition, not pretending every worker is permanent.
01 · CONTROL
Service-worker scope
Offline shell, version control, client discovery, message routing, and bounded WebGPU jobs where supported.
StateWaiting
02 · CLIENTS
Participating tabs
Every controlled OneBrain page becomes a recoverable compute host and advertises live capacity to the mesh.
Coverage0 tabs
03 · COMPUTE
Dedicated WebGPU workers
Actual sustained GPU work is sharded here with heartbeat, device-loss detection, CPU fallback, and backpressure.
Ready0 workers
04 · CONTINUITY
Persistent local state
IndexedDB checkpoints and proof roots let the fabric restart without treating worker memory as durable.
Root00000000
Compute-lane telemetry
Measured backend, throughput, workload, heartbeat, completed tasks, and rolling root per worker.
Aggregate fabric capability
Live local totals across all active lanes.
0%
coverage readiness
Aggregate throughput
0
measured operations / second
Completed tasks
0
shards returned successfully
GPU lanes
0
workers with active WebGPU devices
Scheduler pressure
0%
adaptive intensity after frame protection
Scope coverage map
Current tab, peer tabs, and service-worker-discovered clients.
Fabric event stream
Lifecycle, GPU, scheduler, coverage, and recovery events.
Audit and recovery
Every claim has a boundary. Every state has a root.
Export proofs, preserve checkpoints, inspect operating logs, and recover the local brain without relying on a server.
Integrity roots
Rolling digests over activity, memory, and corpus state.
Brain root
—Memory root
—Corpus root
—Export and recovery
Portable local artifacts.
Scientific claim boundary
What is implemented, and what is not implied.
| Capability | Status | Boundary |
|---|---|---|
| Local lifelong memory | Real | IndexedDB in this browser profile |
| 26B virtual address space | Real | Sparse deterministic addressing, not 26B materialized neurons |
| Dense local compute | Real | Opened selectively around active basins |
| Service-worker compute control | Real | Coverage, caching, discovery, routing, and event-bounded GPU tests |
| Scalable WebGPU execution | Real | Dedicated workers across opt-in controlled tabs with CPU fallback |
| Full human-corpus model | Boundary | Requires licensed corpora and external-scale compute |
| General oracle behavior | Boundary | Answers are limited to learned local memory |
Architect log
Newest events first.