Repository navigation
Conversation
…oses #222) Add brain summaries pipeline for Queen Lotus orchestration: - specs/queen/brain_summaries.t27: Spec for episode summarization - 6 tests for aggregation functions - 4 invariants for consistency checks - 2 benchmarks for performance - conformance/BRAIN_SUMMARIES_SCHEMA.json: JSON schema for summaries - .trinity/queen-brain/summaries/: Directory structure - .github/workflows/schema-validation.yml: Add brain summaries validation - .trinity/seals/BrainSummaries.json: Seal file Phase 4 progress: 50% (7/9 steps complete) φ² + 1/φ² = 3 | TRINITY Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add similarity search operations for VSA trit vectors: - specs/vsa/similarity_search.t27: Similarity search spec - Cosine similarity [-1, 1] - Hamming similarity [0, 1] - Jaccard index [0, 1] - Top-K search with multiple metrics - HNSW-like approximate search structure - 8 tests, 5 invariants, 2 benchmarks - .trinity/seals/VSASimilaritySearch.json: Seal file Spec growth: 81 → 82 specs (target: 100 by Ring 070) φ² + 1/φ² = 3 | TRINITY Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Phase 4 progress: 50% → 60% - Spec count: 81 → 82 - Add Rings 061 (brain summaries) and 062 (VSA similarity search) to completed rings - Update queen health to 0.95 - Update revision note φ² + 1/φ² = 3 | TRINITY
Add fundamental ternary logic gates: - specs/isa/ternary_gates.t27: Ternary logic gates spec - NOT gate (ternary inverter) - MIN gate (acts as AND) - MAX gate (acts as OR) - Consensus, majority, any, all gates - 13 tests, 7 invariants, 4 benchmarks - .trinity/seals/TernaryGates.json: Seal file Spec growth: 82 → 83 specs (target: 100 by Ring 070) φ² + 1/φ² = 3 | TRINITY Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Phase 4 progress: 60% → 70% - Spec count: 82 → 83 - Add Ring 063 to completed rings - Update revision note φ² + 1/φ² = 3 | TRINITY
…loses #228) Add balanced ternary arithmetic operations: - specs/isa/ternary_arithmetic.t27: Ternary arithmetic spec - Single trit addition with carry propagation - Single trit subtraction with borrow - Single trit and word-level operations - Ternary to decimal and decimal to ternary conversion - Comparison operations for ternary words - 8 tests, 6 invariants, 4 benchmarks - .trinity/seals/TernaryArithmetic.json: Seal file Spec growth: 83 → 84 specs (target: 100 by Ring 070) φ² + 1/φ² = 3 | TRINITY Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…oses #230) Add ternary encoding schemes: - specs/base/ternary_encoding.t27: Ternary encoding spec - Bit to trit and trit to bit conversion - Nibble to 3 trits mapping - Byte to trit array conversion - Balanced to unipolar ternary conversion - String/character encoding - 9 tests, 5 invariants, 3 benchmarks - .trinity/seals/TernaryEncoding.json: Seal file Spec growth: 84 → 85 specs (target: 100 by Ring 070) φ² + 1/φ² = 3 | TRINITY Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Phase 4 progress: 70% → 80% - Spec count: 83 → 85 - Add Rings 064 (ternary arithmetic), 065 (ternary encoding) to completed rings - Update queen health to 0.98 - Update revision note with ISA layer progress φ² + 1/φ² = 3 | TRINITY
Add ternary memory operations: - specs/base/ternary_memory.t27: Ternary memory spec - TritCell with metadata (state, access count, timestamp) - TernaryWord (27 trits) with checksum - TernaryMemoryBank for allocation and access - Locking and dirty state tracking - 8 tests, 6 invariants, 4 benchmarks - .trinity/seals/TernaryMemory.json: Seal file Spec growth: 85 → 86 specs (target: 100 by Ring 070) φ² + 1/φ² = 3 | TRINITY Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Phase 4 progress: 80% → 90% - Spec count: 85 → 86 - Add Ring 066 (ternary memory) to completed rings - Update queen health to 1.0 (excellent) - Update revision note with base layer progress φ² + 1/φ² = 3 | TRINITY
Add ternary shift and rotate operations: - specs/isa/ternary_shift.t27: Ternary shift spec - Left/right shift with zero fill - Left/right rotate (circular shift) - Arithmetic shift right (sign-preserving) - Trit extraction and insertion - 9 tests, 5 invariants, 4 benchmarks - .trinity/seals/TernaryShift.json: Seal file Spec growth: 86 → 87 specs (target: 100 by Ring 070) φ² + 1/φ² = 3 | TRINITY Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add ternary bitwise operations: - specs/isa/ternary_bitwise.t27: Ternary bitwise spec - Tritwise AND, OR, XOR, NOT operations - NAND, NOR, XNOR operations - Mask and merge operations - 9 tests, 5 invariants, 5 benchmarks - .trinity/seals/TernaryBitwise.json: Seal file Spec growth: 87 → 88 specs (target: 100 by Ring 070) φ² + 1/φ² = 3 | TRINITY Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add comparison and conditional operations for ternary values to ISA. - Single trit comparison: trit_compare, trit_equal, trit_not_equal - Word comparison: ternary_word_compare, ternary_word_equal - Sign detection: ternary_word_sign, is_zero, is_negative, is_positive - Conditional: ternary_select, ternary_conditional_move - Min/Max: ternary_word_min, ternary_word_max - 9 tests, 5 invariants, 5 benchmarks Verdict: clean Spec count: 89 → 90 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Owner
Author
|
Superseded by P0/P2 rewrite. The ternary ISA specs from these PRs are already in master through the P0 Core Rewrite and P2 de-Zigfication. These old Zig-based ISA PRs (Ring-064 through Ring-088) have been superseded by the new spec-first architecture where specs live in specs/ rather than Zig files. No conflicts — just clean cleanup of old PRs that no longer apply. |
gHashTag
added a commit
that referenced
this pull request
Jul 4, 2026
Honest split of the 83-format catalog into three SW conformance tiers: - 69 bit-exact (independent 2nd witness, abs_error=0) - 6 bit-exact self-consistent (single decode law, no 2nd witness -- promotable) - 8 structural (genuinely parametric families -- not promotable) gf14 correction (vs prior 68/7/8 draft): gf14 is bit-precise, NOT self-consistent. It has an independent second witness -- iverilog exhaustive 16384/16384 via gf_decode_param #(WIDTH=14,E=5,M=8,BIAS=15), abs_error=0, merged in trinity-fpga PR #239 (merge cb845f75f). That witness is now recorded verbatim in the gf14 pack witnesses[] field, flagged as an SW simulation witness (not an on-silicon Tier-E claim). Correct honest count is therefore 69/6/8, not 68/7/8. 6 promotable self-consistent rungs (need a 2nd oracle to reach bit-exact): gf48, gf96, gf128, gf512, gf1024 (Phase-B wide GF), gf256 (open bias). Cross-walk table (CROSSWALK_sw_hw.{md,json}) auto-derives SW tier from the INDEX and maps each format to decode-HW / compute-HW Tier-E status from the trinity-fpga #199 SSOT snapshot. HW ceiling on XC7A200T = 71/83 (terminal; takum32/64 = routing-failure). encoding != compute != FPGA; catalog = 83. README SHA-256 manifest refreshed (83/83). Generator idempotent.
This was referenced Jul 4, 2026
gHashTag
added a commit
that referenced
this pull request
Jul 4, 2026
Honest split of the 83-format catalog into three SW conformance tiers: - 69 bit-exact (independent 2nd witness, abs_error=0) - 6 bit-exact self-consistent (single decode law, no 2nd witness -- promotable) - 8 structural (genuinely parametric families -- not promotable) gf14 correction (vs prior 68/7/8 draft): gf14 is bit-precise, NOT self-consistent. It has an independent second witness -- iverilog exhaustive 16384/16384 via gf_decode_param #(WIDTH=14,E=5,M=8,BIAS=15), abs_error=0, merged in trinity-fpga PR #239 (merge cb845f75f). That witness is now recorded verbatim in the gf14 pack witnesses[] field, flagged as an SW simulation witness (not an on-silicon Tier-E claim). Correct honest count is therefore 69/6/8, not 68/7/8. 6 promotable self-consistent rungs (need a 2nd oracle to reach bit-exact): gf48, gf96, gf128, gf512, gf1024 (Phase-B wide GF), gf256 (open bias). Cross-walk table (CROSSWALK_sw_hw.{md,json}) auto-derives SW tier from the INDEX and maps each format to decode-HW / compute-HW Tier-E status from the trinity-fpga #199 SSOT snapshot. HW ceiling on XC7A200T = 71/83 (terminal; takum32/64 = routing-failure). encoding != compute != FPGA; catalog = 83. README SHA-256 manifest refreshed (83/83). Generator idempotent. Refs #1347
gHashTag
added a commit
that referenced
this pull request
Jul 4, 2026
* conformance: honest 69/6/8 pack split + SW/HW cross-walk Honest split of the 83-format catalog into three SW conformance tiers: - 69 bit-exact (independent 2nd witness, abs_error=0) - 6 bit-exact self-consistent (single decode law, no 2nd witness -- promotable) - 8 structural (genuinely parametric families -- not promotable) gf14 correction (vs prior 68/7/8 draft): gf14 is bit-precise, NOT self-consistent. It has an independent second witness -- iverilog exhaustive 16384/16384 via gf_decode_param #(WIDTH=14,E=5,M=8,BIAS=15), abs_error=0, merged in trinity-fpga PR #239 (merge cb845f75f). That witness is now recorded verbatim in the gf14 pack witnesses[] field, flagged as an SW simulation witness (not an on-silicon Tier-E claim). Correct honest count is therefore 69/6/8, not 68/7/8. 6 promotable self-consistent rungs (need a 2nd oracle to reach bit-exact): gf48, gf96, gf128, gf512, gf1024 (Phase-B wide GF), gf256 (open bias). Cross-walk table (CROSSWALK_sw_hw.{md,json}) auto-derives SW tier from the INDEX and maps each format to decode-HW / compute-HW Tier-E status from the trinity-fpga #199 SSOT snapshot. HW ceiling on XC7A200T = 71/83 (terminal; takum32/64 = routing-failure). encoding != compute != FPGA; catalog = 83. README SHA-256 manifest refreshed (83/83). Generator idempotent. Refs #1347 * conformance: process gates for #1345 — NOW.md 69/6/8 reclass entry (Closes #1347) Restore the docs/NOW.md sync entry that collapsed out of the diff during the rebase onto wave-loop-414, and add the L1 traceability issue link (#1347) so the now-sync-gate and check-linked-issue required gates pass. Content unchanged: the 69/6/8 conformance packs + gf14 witness fix + SW/HW cross-walk are already reviewed. This commit only adds the two process wrappers (issue-ref + NOW.md diff entry). --------- Co-authored-by: gHashTag <admin@t27.ai>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Ring 069: Ternary Comparison Operations
Closes #238
Changes
Add comparison and conditional operations for ternary values to the ISA specification.
Test Results
Verdict
clean
φ² + 1/φ² = 3 | TRINITY