← Editing fieldnotes

SOURCE-PAIRED STUDIES / REVISION R03

Let the brackets
arrive first.

One source moment, rebuilt as editable code. R03 gives the black letters their natural proportions and rebalances the punctuation around them.

Study revision on this page
1
New source ranges
0
Reviewed on this page
1
Independent wins
1

Independently preferred reconstruction. A fresh blind critic chose this remake in round 3, with moderate confidence, after the original won rounds 1 and 2. One source moment counts once; the alternate copy is not a second study. Read the verdict and evidence limits →

Watch the actual pair

3.1 seconds · 186 frames · 1920 × 1080 · 60 fps · silent

01 / Original excerptOpenAI · Build and share apps in Codex
02 / Code-only remakeDM Sans text + four SVG paths · revision r03

Local frame 0 aligns both files. Shared controls help compare; browser seeking is not a substitute for decoded-frame checks. Each native player remains usable without JavaScript.

TIMING UNCHANGED

Punctuation first.
Words follow.

Empty square brackets hold before the first word. A connector joins, then round brackets introduce the second word. The full phrase holds for 70 frames.

  1. 05First glyph after five blank frames.
  2. 29–37Square brackets hold empty.
  3. 66 / 69The connector reveals in two steps.
  4. 78–89Round brackets hold empty.
  5. 116–185The complete phrase holds without a transition.

Download the editable r03 build

Original excerpt ↓ · Reference remake ↓ · Alternate-copy remake ↓ · Provenance JSON · Counts JSON

What changed, and what remains open

DM Sans replaces the fitted Geist letterforms. At 100 px and weight 650, the measured cap height is 71 px; every word keeps scale X/Y at 1. The complete line is wider and recentered. Closing punctuation follows natural prefix advances. Reveal frames, parent-travel deltas and final-hold timing stay unchanged.

The licensed substitute's outlines, including its y, still differ from the original. Purple contours and spacing are authored decisions, not recovered OpenAI design settings. Geometry checks do not establish optical finish. Native Resolve execution remains untested; the blind preference is an agent editorial judgment from decoded progression and verified playback, not a human continuous-viewing certification.

Original footage remains copyrighted by its owner; no redistribution license is established here. It is separate comparative research material, not an all-hands presentation asset. Rights and reuse boundary →

What won the blind comparison

The remake's delimiter spacing stays more coherent as the words develop. Wider natural letterforms integrate better with the punctuation. Centering, reveal order and final hold did not provide a meaningful advantage.

At frame 112 the original closing parenthesis crowds the final n; at frame 99 it leaves excess trailing space. The critic preferred the remake's substring-dependent advance over equal per-character steps.

The code and download archives remain byte-identical to the clean-rendered pre-review package. Their embedded study record is a preserved pre-review snapshot; this dated review supplies the subsequent decision.

Successful playback execution, not continuous visual observation. Snapshots and recording failed before the preview host became unavailable.