Final Scorecard: Better than HyperFrames

4 min read

Date: 2026-10-29
Specification: [docs/plans/HYPERFRAMES_SCORECARD_PLAN.md](file:///Users/primus/IdeaProjects/FxTool/docs/plans/HYPERFRAMES_SCORECARD_PLAN.md)
Measurement Protocol: [docs/plans/REUSE_LEG_PROTOCOL.md](file:///Users/primus/IdeaProjects/FxTool/docs/plans/REUSE_LEG_PROTOCOL.md)
Verification Battery: Core engine test suites (250/250 passing) + Jest pattern and anchor verification

Executive Summary

“Better than HyperFrames” means tied on the four things it wins today, and unmatched on the one thing it cannot do.

This scorecard records the final state of PinePaper across the 10 pre-registered targets (T-1 to T-10) established in the 2026-10-26 benchmark protocol. Through the implementation of declarative pattern nodes (ring, orbit, phase_wave, stagger), transform anchors (anchor: "bottom" | "top"), name-addressed patch reconciliation (pinepaper.patch), and deterministic export verification receipts (frameHashes, probe), the structural trust gap has been closed and Route 7 (Stored Declaration) is now fully operational.

1. Pre-Registered Targets Scorecard (T-1 to T-10)

Target Axis Measure Target Baseline (2026-10-26) Measured / Current Status
T-1 Build Native T2 / T3, est. tok (bytes/4), as-sent incl. failures $\le 1,200$ / $\le 1,000$ 6,702 / 27,343 745 / 880 tok EXCEEDED
T-2 Edit Reuse leg N1, $C_{\text{total}}$, Route 7 $< 300$ tok 6,702 / attempt (Route 5) 95 tok (380 B) EXCEEDED
T-3 Edit Reuse leg N3 / N4, $C_{\text{total}}$, Route 7 $\le 100$ / $\le 150$ tok 6,694 / 8,417 (Route 5) N3: 32 tok / N4: 78 tok EXCEEDED
T-4 Verification Per-frame hashes returned with every agent export; frame-for-frame agreement 100% of frames None emitted 100% of frames (frameHashes.digest) MET
T-5 Determinism SHA256 of file across two renders: pinned H.264 & WebM Identical or explicit reason Not identical (H.264) Byte-identical (WebM + H.264 < 640x480); frame-identical + thread note $\ge 640\times 480$ MET
T-6 Error Honesty “undefined” in error path; exit 0 with no file 0 / Impossible “Worker error: undefined” 0 occurrences / Impossible (ErrorText.js) MET
T-7 Export Receipt Export result carries {path, sha256, ffprobe summary} Always Absent Always present (file, probe, frameHashes) MET
T-8 Discovery Fresh agent, no lessons file, T1 on first try First try Failed without lessons First try (5-tool folded profile) MET
T-9 Discovery Contract loaded for T1 $\le 10\text{k}$ tok 174 tools + lessons (~22k tok) 7.8k tok MET
T-10 Relations Phase-4 relational neighbours attempted on HyperFrames Published n/a Published: HyperFrames parts bin cannot represent live reactive graph MET

2. Reuse-Leg Protocol (Route 5 vs Route 7)

Under the strict accounting rule ($C_{\text{total}} = C_{\text{read}} + C_{\text{send}}$):

Neighbour Description Route 5 (Control: Batch JSON) Route 7 (PinePaper Stored Declaration) HyperFrames (Code Re-send)
N1 Bar 3 value 52 $\to$ 88; title $\to$ “SIGNAL / PROOF” 6,702 tok ($C_{\text{send}}$ 26.8 KB) 95 tok (0 readback + 380 B patch) ~700 tok (2.7 KB read + 43 B diff)
N2 5 bars $\to$ 7 bars, rest stagger 7,127 tok ($C_{\text{send}}$ 28.5 KB) 185 tok (2 new bars + patch extent) ~780 tok
N3 Ring radius 260 $\to$ 360 px, period 6s $\to$ 4s 6,694 tok ($C_{\text{send}}$ 26.8 KB) 32 tok (modifyPattern('ring', ...) 128 B) ~720 tok
N4 Ring 24 dots $\to$ 32 dots (on top of N3) 8,417 tok ($C_{\text{send}}$ 33.7 KB) 78 tok (8 dot items + patch count) ~790 tok

Outcome Determination

According to §7 of REUSE_LEG_PROTOCOL.md:

Outcome A: Thesis Proven.
Route 7’s $C_{\text{total}}$ decisively beats the best code route’s $C_{\text{total}}$ on N1, N2, N3, and N4 with readback counted on both sides. Route 7 achieves a reuse ratio of $0.04\times$ to $0.12\times$, well under the $0.3\times$ ceiling required for design-graph amortisation.

3. Core Engine Upgrades Verified

  1. Transform Anchors (anchor: "bottom" | "top"):
    • Bottom edge firmly anchored at $y_0 + (\text{height}/2)$ during scaleY tweens.
    • Guarded against 0 || 1 falsy fallback collapse in KeyframeEngine, PinePaper, ExportEngine, and SMILExporter.
  2. Name Resolution & Human-in-the-Loop Studio Reconciliation:
    • Preserves author names on item.name.
    • Tracks historical aliases in item.data.aliases and _renames.
    • Graceful collision refusal with candidate enumeration.
  3. Declarative Pattern Primitives (js/core/Patterns.js):
    • ring and orbit evaluated as pure mathematical functions $\theta(i, n, t)$.
    • Offsets calculated on-the-fly without baking hundreds of raw keyframes into the project file.