causal-edge: exact truth codes for deduction, induction, abduction (D-NALP-4 harvest)

[AdaWorldAPI]

This makes deduction, induction and abduction exact in causal_edge::isa::truth, the one place NARS truth already lives. It is the harvest of draft #1544: its oracle and probe move into causal-edge, and the lane terminal it planned is not built.

Problem

A CE64 register holds f and c as u8 codes (x / 255). execute (and so forward and the named register methods) and syllogize decoded them to f32, applied the law with rounding at every step, and packed with (v · 255).round(). That is one confidence code off on 2,413 of 2³² inputs for deduction and 4,352 each for induction and abduction.

Change

• isa::truth::code::{deduction, induction, abduction}: u8 codes in and out, computed in integers and rounded once, half up.
• Opcode::truth_code: routes those three rules there. Revision and synthesis keep the f32 path, unchanged.
• execute and syllogize: pack truth_code's result.
• truth::pack: the one place (v.clamp(0, 1) · 255).round() is spelled.
• The f32 functions stay for real-valued callers.

Evidence

• Exhaustive probe (cargo run --release -p causal-edge --example truth_exact_probe), all 2³² inputs per rule:

  rule	truth::code ≠ oracle	old f32 path ≠ oracle
  deduction	0	2,413
  induction	0	4,352
  abduction	0	4,352

  The oracle (tests/common/truth_law.rs) is written independently of the crate.

• tests/ce64_truth_exact.rs, 8 tests:

  • hand-computed codes;
  • two half-up ties, the second of which rejects ties-to-even;
  • the induction/abduction mirror;
  • zero evidence;
  • OpenNARS nal1.1 at 4d1bdb0;
  • codes equal to the oracle on the every-5th grid;
  • execute and syllogize packing the oracle's codes on exactly the inputs where the f32 path was off: 12 deduction rows of the grid, and 17 induction and 17 abduction rows from an exhaustive scan of the three codes their confidence depends on. The grid holds none for those two.

• Disable runs, each red: truth_code back to f32; syllogize deduction back to f32; round down; ties-to-even.

• Unchanged: the legacy golden capture (126 words) and all other causal-edge tests, in both layouts.

• Dependents green:

  • planner lib (457 passed, 3 ignored), plus chain_confidence, d_rpf_table_1, moore_tenant_isa_equivalence and moore_nars16_probe;
  • cognitive-shader-driver lib (131 passed);
  • lance-graph-osint p5_syllogize (4 passed).

Board

• Entry: 2026-10-14-isa-truth-exact-codes.md.
• STATUS_BOARD: D-NALP-4 → partial, in PR.
• Plan: a MEASURED note in nal-polyfill-v1.md §4 C.

Not in this PR

• Exact revision and synthesis. Revision first needs a ruling on two operands at confidence 255, which today pack (0, 0) and are pinned as a known defect.
• Whether D-NALP-4/5 still need a lane terminal. This is the rescope of plan §4 B/C and stays open.
• Clippy. cargo clippy -D warnings on causal-edge already fails on six lints in code this PR does not touch; CI runs only cargo test for this crate. The additions here are clippy-clean.

🤖 Generated with Claude Code

https://claude.ai/code/session_011BMJnDBAfzcuQh5HeT39Rh

---

Generated by Claude Code

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

• Configuration used: Organization UI
• Review profile: CHILL
• Plan: Advanced
• Run ID: 8000def8-273d-4728-9a0e-e3073ba3a885

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Warning

Billing warning: we have not been able to collect payment for this subscription for more than 72 hours. Please update the payment method or pay any pending invoices in Billing to avoid service interruption.

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-14T05:58:01.906742Z	b0b3fb7	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[coderabbitai]

Add Carrot credits or activate Agent usage billing to use Autofix