Repository navigation
Give a travelling body a floor to travel on - #29
Merged
Merged
Conversation
The ground grid was 6 m across — ±3 m from the origin — while one line travels 2.11 m on Kimodo and 3.89 m on SnapMoGen, the furthest ghost sibling reaches 4.77 m, and a whole baked poem's net displacement measured 3.63 m. So the body walked off its own floor, and once off it there was no ground left to read the travel against. It is 20 m now at the same 0.5 m cell: about twice the worst measured case, and centred on the origin, so only half of it lies in whichever direction the body happens to go. Day 4's headline 9.51 m is path length, not displacement — that body ran left and then right. Path length is the trails' and the floor path's problem, not the floor's, which only has to reach as far as the body gets from the start. The grid was also painted in --line (#232732): the hairline colour for 1px dividers between panels, chosen at desk distance. Measured against the backgrounds it actually sits on it reached 1.12:1 on the bench and 1.045:1 in performance, and only 1.34:1 even at full opacity, so no amount of alpha was ever going to carry it. It is #434b5c now, with #6b7689 centre lines so the origin is readable and travel reads as distance from somewhere. Both are a step along the house ramp between --line and --ink-faint, measuring 1.94/1.68:1 for the floor and 3.38/2.67:1 for the origin — deliberately under the 3:1 graphics floor, because the performance view exists to send the eye to the body and a floor that passed a text-contrast bar would compete with it. That second change measured better and was judged no better by eye. It is kept because the tuning stopped, not because it was shown to help. What this view needs first is an opinion formed in a room, which is the judgement roadmap.md track 3 says should not be made at a desk. The compare page is unchanged. Its panels pass grid: false and have always been floorless, so there was nothing there to recolour; whether they should share a floor — which would give the travel comparison a common scale instead of a footer number — is left open. roadmap.md records what the wider floor exposes: that nothing else in the performance view travels with the body, the spotlight least of all, since it is a radial gradient pinned to the centre of the viewport. A camera that followed the pelvis would fix that without touching the data, which is a different thing from pelvis-aligning the motion — but with the ghost-cloud on it cannot be done innocently, because the far siblings are where that ~75% of divergence lives. So what a follow camera should follow stays a research decision. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014yJSPiRjB5WvTH2x4LB8V6
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
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.
The ground grid was 6 m across — ±3 m from the origin. Against that:
docs/journal/2026-09-08-day-4.md:267)So the body walked off its own floor, and once off it there was no ground left to read the travel against. It is 20 m now at the same 0.5 m cell — about twice the worst measured case, and centred on the origin, so only half of it lies in whichever direction the body happens to go.
Day 4's headline 9.51 m is path length, not displacement — that body ran left and then right. Path length is the trails' and the floor path's problem, not the floor's, which only has to reach as far as the body gets from the start.
The colour, and an honest note about it
The grid was painted in
--line(#232732) — the hairline colour for 1px dividers between panels, chosen at desk distance. Measured against the backgrounds it actually sits on:rgb(23,26,33)rgb(13,15,20)rgb(35,39,50)That last row is why this is a colour change and not an opacity change:
#232732tops out at 1.34:1 even opaque. It is#434b5cnow with#6b7689centre lines, both a step along the house ramp between--lineand--ink-faintrather than an invented colour — 1.94/1.68:1 for the floor, 3.38/2.67:1 for the origin. Deliberately under the 3:1 graphics floor, because the performance view exists to send the eye to the body and a floor that passed a text-contrast bar would compete with it.The centre lines are new in effect:
GridHelpertakes a separate centre-line colour and both arguments used to be the same, so there was no origin cue at all. On 20 m of floor that is what lets travel read as distance from somewhere.This part measured better and was judged no better by eye. It is kept because the tuning stopped, not because it was shown to help. The likely reason is that 41 lines each way converging toward the horizon is a density problem rather than a contrast one, and brightening them made that worse. Reverting to
#232732while keeping the 20 m floor is a one-line change if that is the call.Not in this PR
grid: falseand have always been floorless, so there was nothing there to recolour. Whether they should share a floor — which would give the travel comparison a common scale instead of a footer number — is left open.controls.target, fit-to-trajectory bounds, dead-zone follow) and all wait on seeing this from the back of a room.What the change exposes, recorded in
roadmap.mdNothing else in the performance view travels with the body — the spotlight least of all, since
style.css:531is a radial gradient pinned at50% 44%of the viewport, so a body at 3.9 m stands outside its own light. A camera following the pelvis would fix that without touching the data, which is a different thing from pelvis-aligning the motion and does not reopen the "displacement stays" decision. But with the ghost-cloud on it cannot be done innocently: the far siblings are where that ~75% of divergence lives. So what a follow camera should follow stays a research decision.Verification
tscclean; 87 pass, 0 fail. No test is added — this is a scene constant, not a decision that can be wrong without looking wrong, which is what this repo's test convention covers.Fixture mode cannot exercise any of it.
slip-awayat 0.46 m is the only fixture that travels, so the change is invisible without the real workers.What has actually been done: the change was looked at in a browser with all three models running, and the colour was judged no better by eye. That is the extent of it.
Still owed, and named rather than implied — none of these has been run:
The last one is the only one that can settle the colour question, and it is the judgement
v4-proposal.mdtrack 3 says not to make at a desk.🤖 Generated with Claude Code
https://claude.ai/code/session_014yJSPiRjB5WvTH2x4LB8V6