The Afterlife of Light

I watched the film first, as asked: one frame per second (203 frames, viewed in order as contact sheets) plus full-resolution checks of the margin credits. I read no other files in the folder. I could measure the audio but not hear it.

VERDICT: FAIL

STRONGEST: The drawn inserts carry the film, and they are clear about what they show.

• Earthshine (0:24–0:31): a crescent Moon with the dark part faintly lit.
• Huainanzi dew vessel (0:58–1:04): silhouettes under the Moon. The slip "The fire is real. The dew comes from a clear night sky. The Moon gets the credit" is the best one in the film.
• Clouds (1:07–1:16): done in red-figure vase style.
• Herschel's thermometer past the red (1:33–1:37).
• The magic mirror (1:45–1:53): a plain disc throws a seated figure onto the wall. This is the film's best visual idea.
• Kyoto's Daimonji fires (2:36–2:42): the 大 character builds stroke by stroke.

The look is consistent: typed paper slips, a vertical PHOTOGRAPHED credit in the margin, an INVENTED tag on drawn inserts. The opening sun returns at the end, which closes the loop.

WEAKEST: It plays as a list.

• Same beat throughout: eleven chapters in about 190 seconds. Each one is a 2–3 s title card, then one or two slips of 3–5 s. Nothing in the film speeds up or slows down.
• Audio does not build: integrated loudness is -20.5 LUFS with a loudness range of only 4.3 LU. The level stays at about -19 dB in every 5-second window from 0:00 to 3:15, then drops to about -25 dB over the credits.
• Chapter V "What the light holds" (1:21–1:41) is rushed: it packs in Ibn al-Haytham, Newton, Herschel and Melloni in about 17 s. Melloni gets about 4 s. Al-Haytham's slip says nothing about what he found.
• Stray insert at 1:17–1:20: "A record is one thing. What people remember is another," over a tractor snapshot. It does not connect to the debt story before it.
• Chapter VI (1:42–1:53) never names the makyō, nor where or when it comes from.
• The through-line keeps changing:
  • The opening says "What it becomes depends on what it meets."
  • The title says "what light leaves behind."
  • The question at 2:55 asks "What survives when the source is gone?"
  • The last line is "No object has only one role."
• The ending does not land: the central question is asked only in the last 30 seconds. It is then answered with three cases the film never set up: "Warmth in a stone" (2:59), "A warning on the next hill" (3:02) and "A keepsake" (3:05, over a woman in a sickbed). The answer is gestured at rather than earned.

HONESTY: Several archive shots imply something false.

• 0:36: under "Sun → Moon ⇄ Earth," the footage cuts to a red planet with a moon. It looks like Mars, from "Ticket to Planetary Entry."
• 0:54–0:57: "a bronze mirror takes fire from the Sun" plays over a bonfire from an amateur film.
• 1:24–1:27: "Cairo, 11th century. Ibn al-Haytham" plays over a 1940s room with a lamp and curtains, from "Door to Heaven."
• 1:38–1:41: "Melloni finds that rock salt passes heat" plays over glass jewels or coins from the same film. This most directly implies the wrong substance.
• 2:07–2:10: "On the Yatsushiro Sea… shiranui" plays over a murky, foggy woodland from "The Green Promise," not a sea.
• Drawn insert, 1:28–1:32: the Newton slip shows a hand putting a thermometer into the spectrum. That is Herschel's 1800 experiment, not Newton's.
• Margin credit at 0:32–0:36: the possessive apostrophe after "NASA" renders as a broken glyph.

The end card lists Kojiki and "Ichihashi et al., UIST 2024 and TEI 2025," but I saw neither used on screen. It also says every caption is INVENTED, which undercuts the factual claims those captions make.

FACTS: Most claims hold up.

• Correct: the Moon returning about a tenth of the sunlight it receives; Leonardo on earthshine around 1510; obsidian mirrors at Çatalhöyük around 6000 BCE, some found in burials; Huainanzi 6 on the burning mirror and dew vessel (2nd c. BCE); Clouds 423 BCE and the burning stone; Herschel 1800; Melloni and rock salt in the 1830s; fireflies' cold light; the 囊螢映雪 saying; shiranui; Kyoto on 16 August.
• Minor error: for SN 1987A, the light left about 168,000 years earlier; "about 160,000" undercounts.
• Debatable: "Newton 1666" is the conventional date, but his prism work ran into the following years.
• Interpretive gloss: "The writing melts. The debt stays" does not happen in the play.

Slip readability:

• Most slips stay up 3–5 s, which is enough for one line.
• The Huainanzi slips at 0:58–1:04 compete on screen with a quoted subtitle and a vertical Chinese banner.
• The typewriter slip text will be small on a phone.

REPAIR: The three biggest changes:

1. Say the question at the start: "What survives when the source is gone?" Then let each chapter answer it, so the ending pays something off.
2. Fix the false matches: replace or cut the Mars, bonfire, al-Haytham living-room, Melloni jewels and shiranui forest shots with drawn inserts or with footage of the real thing. Redraw the Newton insert so it shows only the prism.
3. Change the pace:
   • Drop the stray tractor-photo insert.
   • Split or trim chapter V.
   • Give two or three chapters (the magic mirror, the dew, Daimonji) twice the time and some silence, and let the sound rise and fall with them.
   • Set up the stone, beacon and keepsake earlier, or cut them, so the ending goes back over cases the film has already shown.

Smaller fixes: correct SN 1987A to about 168,000 years, name the makyō, fix the NASA apostrophe, and make the end card list only the sources the film actually uses.

Everything is in /tmp/claude-0/-home-claude/21362850-8176-5f25-9feb-97722a7915b9/scratchpad/f. The frames are o001.jpg to o203.jpg (frame n is at about second n-1), the contact sheets are sheet00.jpg to sheet16.jpg, and the audio levels are in rms.txt.

VERDICT: PASS

OLD PROBLEMS: All four are fixed.

1. The question "What survives when the source is gone?" now appears as its own card at 0:17–0:20, and each of the 11 numbered sections ends on a "What survives: ..." slip. The thread is there from the start.
2. Every image the earlier critic flagged has been replaced by a drawn insert tagged INVENTED: earthshine, the burning mirror (yangsui), the makyō, the fangzhu dew vessel and the shiranui lights. The Cairo and rock-salt material is gone.
3. The Newton slip now runs over a drawn prism (1:46–1:50), and Herschel gets his own thermometer drawing (1:51–1:57).
4. The ending (3:09–3:30) repeats one short slip per section in order, then returns to the opening sun footage. It introduces nothing new.

STRONGEST: The bookend works: it opens on the clouded sun under "Every story here begins with the same light" and closes on the same shot under "Something always survives. / Rarely the thing that gets the credit." The best sections are the Huainanzi sequence (concave mirror lighting tinder, then "The dew comes from a clear night sky. The Moon gets the credit."), the shiranui lights multiplying along the horizon, and Supernova 1987A cutting to two figures looking at a star. In those, the drawn image does the explaining and the slip only names it. Keeping archive footage to general claims (a mirror, sweat, skin, a runner on a beach) is the right discipline.

WEAKEST: The refrain has made the film regular but has not made it build. Eleven sections run on one template (title card, case slip, "What survives:" slip), each 13–18 s long, so it still feels list-like. Several sections don't answer the question as posed:

• Newton, Herschel and "The body": the source, the Sun, is never gone.
• Fireflies, Kyoto fires and SN 1987A are not "the same light" the opening promises.

The closing line about credit is set up by only one case, the dew and the Moon. So the ending lands as a clever line rather than a conclusion the film earned.

HONESTY AND FACTS:

• The slip facts I checked are correct: lunar albedo of about a tenth, Leonardo and earthshine c. 1510, Çatalhöyük obsidian mirrors c. 6000 BCE, Huainanzi (2nd c. BCE) yangsui/fangzhu with 方諸取露於月, Clouds 423 BCE ll. 749–772, Newton 1666–72, Herschel 1800, 囊螢映雪, shiranui as refraction over the Yatsushiro Sea, Kyoto's Gozan no Okuribi on 16 August (drawn as 大), and SN 1987A at about 168,000 years.
• New image problem, the same kind the last critic failed (smaller, but still a fix): at 1:09–1:12 the slip "China, 2nd century BCE. The Huainanzi pairs two bronze instruments" sits over archive footage of a timber barn's rafters with sunbeams. The margin credit reads "Anatomy Of An Ancient Barn," but on screen the barn stands in for Han China.
• At 2:55–2:59, "Starlight can arrive after its source has changed" sits over "The Eclipse," which looks like a staged early-cinema starfield with painted shooting stars. Tagged only PHOTOGRAPHED, it can read as a real sky.
• "Melting the letters would not cancel the debt" is the film's own comment, not something in Aristophanes; it should be phrased so it doesn't read as the play's verdict.
• Small points: SN 1987A was visible only from the southern hemisphere, which the drawn village doesn't show. The archive title "Sun Peeks Outs" carries its source's typo.

Readability and sound:

• The slips are readable at 720p, but the 2-line makyō slip gets about 3 s.
• In the dew section, the slip, the banner and a Huainanzi quote are on screen at once, and the quote repeats the slip.
• The vertical margin credits and the grey end-credit sources are too small and dim to read at playback speed.
• Sound: integrated loudness is -22 LUFS (quiet for web playback, where about -16 is typical), loudness range is 6.3 LU, and true peak is -1.6 dBFS (no clipping). Momentary loudness sits around -21 and jumps to -14 to -17 at roughly every section change (0:20, 0:52, 1:06, 1:29, 1:58, 2:10, 2:22, 2:37, 3:08). That is a consistent accent, not a fault. The audio drops to -30 to -36 in the makyō section (0:54–1:05) and the Kyoto section (2:41–2:51).

REPAIR:

• Replace the barn shot under the Huainanzi slip with a drawn or neutral sunlight image, or move the slip onto the drawn yangsui frame.
• Either tag "The Eclipse" as a staged film or put the starlight slip over the drawn SN 1987A image.
• Change the opening line or the question so the Sun-present cases (Newton, Herschel, body) and the non-solar cases (fireflies, Kyoto, SN 1987A) belong. One option is "What does light leave behind?" Otherwise cut or merge the cases that don't fit, which would also break the template rhythm.
• Seed the credit idea earlier, for example at the Moon in section I, so the last line has been earned.
• Drop the duplicate dew quote, give 2-line slips at least 4 s, raise the end-credit source text, and bring loudness up to about -16 LUFS.

Contact sheets and crops I made for review are in /tmp/claude-0/-home-claude/21362850-8176-5f25-9feb-97722a7915b9/scratchpad/sh/ and /tmp/claude-0/-home-claude/21362850-8176-5f25-9feb-97722a7915b9/scratchpad/cr/.