Begin here
TS is a research programme, not one magical repository.
The current work separates human language, verifier authority, learned mechanisms, adaptive compute, and deterministic systems so each claim can be tested at the correct boundary.
Reasoning and language
ts-chat-language → TS-Reasoner
Can human language become explicit semantic state, then pass through typed verification without a chatbot bypass?
Model science
Ten-SON-LM ← TensionLM
Does a tension signal causally improve revision and routing, or is it only an ordinary learned gate with a dramatic name?
Adaptive inference
TSQ
Can tension and verifier failures allocate precision more efficiently than always-low or always-high execution?
Verified computing
bogbin + TensionForge
Can storage, execution, training, persistence, and rollback expose exact state transitions and receipts?
What TS is
A verifier-first engineering programme
A collection of falsifiable model and runtime experiments
A way to expose semantic, graph, compute, and state transitions
A public archive that keeps negative results and superseded ideas visible
What TS is not
Not a finished AGI
Not a theory-of-everything proof
Not a general theorem prover
Not a production operating system
Not proof that tension beats standard methods
Not permission to treat confidence as evidence
The intended route
A usable turn should pass through explicit boundaries rather than hiding everything inside one model response.
1. Compile
Human text becomes semantic frames and a MeaningGraph.
2. Verify
Typed checks decide what has support.
3. Update
Accepted state changes; rejected candidates do not contaminate it.
4. Plan
The system selects a response act from verified state.
5. Render
Language expresses the plan without inventing new authority.
Where to go next
Pick the route matching the kind of understanding you want.
Understand the architecture
Read the current projects as four separate programmes with explicit interfaces.
Open routeActive project nodes
These carry the current code. Historical repositories live in the lineage rather than competing on this page.
TS-Reasoner
Typed verifier authority, containment, repair, memory, risk gates, replay, and receipts.
Open nodeTS Chat Language
Human language to inspectable semantic state without using an external LLM as the language substrate.
Open nodeTen-SON-LM
Tests whether tension is a useful causal control signal inside a recurrent semantic workspace.
Open nodeTensionForge
Custom RX 480 OpenCL training with forward, backward, optimizer, and recurrent parity evidence.
Open nodeTSQ
Adaptive precision routing with cognitive and compute receipts.
Open nodebogbin / BogOS
Verified storage, execution, workspace roots, persistence, and rollback receipts.
Open nodeTS-Benchmarks
Shared experiments designed to expose failure and neutral results, not manufacture wins.
Open nodeTensionLM
Previous-generation tension-attention experiments and public model artifacts.
Open node