Deterministic reasoning trace
Deterministic reasoning for any text input
Type natural-language text. RIC turns it into a fixed sequence of steps (parse → structure → rule → final), hashes each step, and anchors the chain into a graphHash. Same input, same steps, same graph on every machine.
Quick start
1. Use the Helix contradiction example.
2. Run reasoning and note the graph hash.
3. Remove “not” and rerun — the graph stays the same while legality flips.
Try the deterministic reasoning engine
Paste a claim, observation, or short paragraph. For a visible effect, start with the Helix contradiction shown below.
Tip: keep the text short and concrete so the steps are easy to read.
Last reasoning run
Run the demo to see a fixed reasoning trace. For identical input, the run id, graph hash, and step sequence are identical across machines and time.
What you type
Free text: claims, observations, short paragraphs. No special syntax. The engine reads what the text actually says.
What RIC does
Builds a fixed sequence of reasoning steps (parse → structure → rule → final), hashes each step, and anchors them into a deterministic graph.
What you get
A proof bundle: steps, legality, hashes, and graphHash. You can store it, replay it, and gate actions on it.
Where this fits
- • Safety or compliance gates in front of agents.
- • Infra control planes that must be replayable.
- • Workflows that need full reasoning provenance.
Next surfaces to try
- • Legality demo — deterministic PASS / HALT over proposals.
- • RIC-STEM — deterministic ODE and algebra on the same core.
- • Home — overview and API access form.