An explainer that stops and asks

The HTML video

interactive/ is the folder export (index.html and assets/), playing right here. Click it, then press ? for the keys, or open it on its own.

Video: final.github.mp4 · 1920x1080 · 30 fps · 157.5 s · 9.7 MB · -14.0 LUFS / -1.6 dBTP · narrated, burned-in captions. The master, final.mp4 (177 MB), is the asset 25-nobel-physics-ice-telescope--final.mp4 of the media release.

HTML video that stops and asks: interactive/, a folder export (index.html and assets/, 4.1 MB) that plays offline in any browser, at its address on the site https://faviovazquez.github.io/showtime/media/25-nobel-physics-ice-telescope/interactive/. It stops at three questions, waits for an answer, says why, and goes on; 8 chapters; links to a moment (#t=1:05) or to a stretch that loops (#t=1:12-1:20); a link preview with the poster frame.

The request#

"Physics 2026 explainer for nobel-2026-lab: a telescope made of ice, ~2 min, three stop-and-ask questions, interactive export, 60 s vertical cut; publish today"

Mode: quick, quality review. Made on 6 October 2026 with showtime 0.4.0 on a 6-core Intel Mac, from the public lab repository nobel-2026-lab (MIT), whose own toy simulations give the film's numbers (kilometre-scale interaction odds, a toy telescope's aim against string spacing). This folder has the 16:9 film and its HTML export; the 60 s vertical cut is not part of the example.

What it shows#

• Questions that stop and ask (0.4.0). Three questions in showtime.json, each anchored to the narrator's line that asks it ("at": "shot-3"), so a re-voice moves the question with its line. The MP4 draws a pause-and-think beat with a 3 s ring; the HTML export stops there, shows the choices (A-C), marks the answer, says why, and plays on from the end of the beat:
  1. A high-energy neutrino (100 TeV) crosses a whole kilometre of ice. What are the odds it hits anything? About 1 in 100,000 is closest: in the toy model, about one in 65,000.
  2. Strings twice as far apart, a quarter as many: how much worse does the toy telescope aim? A few times worse: 5.6x in the toy model (trend only, not IceCube's real aim).
  3. IceCube records about 100,000 neutrinos a year. How many come from outer space? About a hundred, the Nobel Committee says.
• A storyboard table became the project (0.4.0). showtime new dom <dir> --from-storyboard storyboard.md wrote one scene per shot (14), the narration with each line pinned where its shot starts, and storyboard.json; the shots were then built by hand in the Tidewater look signature.
• A cold open, a roadmap and a tie-back. The film opens on the puzzle (how do you catch a particle that goes through everything?), names four steps on a roadmap bar that stays on screen with the current step lit, and ends back on the opening fingertip.
• Real data. The 6 PeV shower of 8 December 2016 is replayed from IceCube's public data release; the sensor layout is IceCube's published geometry.
• The findings gate (0.4.0). Every critic finding had to be fixed or waived before delivery; the review rounds below.

Commands, in order#

showtime is skills/showtime/bin/showtime; reconstructed from the job's ledger and the project's files.

showtime job init nobel-physics --platform x --request "Physics 2026 explainer for nobel-2026-lab: ..."
showtime new dom <job>/project --from-storyboard storyboard.md --look tidewater    # 14 scenes, narration.md
#   the shots built by hand: scenes/s-shot-*.js and .css, assets/ (geometry, the event), the roadmap
showtime voice script <job>/project/narration.md -o <job>/project/voice           # Kokoro am_michael
showtime retime <job>/project --from-voice <job>/project/voice/timeline.json --total 150
#   showtime.json: "questions" (three, each "at" a narration line), "chapters" (eight)
showtime check <job>/project
showtime render <job>/project --job <job>                  # final.mp4, then final-2.mp4 and final-3.mp4
showtime qa <job>
showtime review-pack <job>                                 # round 1: a critic on the pack
showtime review-pack <job> --against <the earlier final>   # round 2: two critics, old and new, both orders
showtime review-respond <job> --fixed r1-S1 "..."          # one line per finding
showtime deliver exports <job> --targets github,x,linkedin

# 2026-10-08, on the 0.4.1 build: the HTML export again, as a folder with a link preview (no re-render)
showtime export html project --folder -o interactive --share-url https://faviovazquez.github.io/showtime/media/25-nobel-physics-ice-telescope/interactive/

Timings#

QA summary#

• At delivery (0.4.0, final-3): PASS (0 fail, 0 warn), 157.50 s, -14.0 LUFS, -1.6 dBTP.
• Today (0.4.1, the same file): loudness, true peak, frame 0, black and frozen stretches all pass; qa now FAILs too_long against the job's platform, X, whose standard accounts take 140 s: the film is 2:37. Shorter cuts for X were part of the delivery and are not in this folder.
• The HTML export, checked in a headless browser from a server that answers range requests (as GitHub Pages does): the start screen, question 1 stopping at 37.2 s, an answer and its reply, Continue (on to 41.7 s), and the range link #t=1:12-1:20 playing 72-80 s; 55 requests, none off the site, no console errors. The embedded soundtrack is AAC 96k at -14.1 LUFS.
• The hearing pass of round 1 (review/round-1/audio.txt): every voice line 14.8-15.6 dB over the music, 135-200 words a minute.

What the critic found#

Two rounds, each critic a fresh agent given only the review pack (the harness refused their file writes, so the findings were saved by the maker, condensed: review/).

Round 1, on final.mp4: ship after fixes, 0 blockers, 10 should-fix, all fixed in final-2 (FINDINGS, RESPONSE):

Round 2, pairwise: final.mp4 against final-2.mp4, two critics, one per order, neither told which was newer (VERDICT): both preferred final-2, both would post it. The one should-fix both raised (captions 0-3 px under the counter's subline at 5.5 s) was fixed in final-3 by moving the counter block up 74 px (RESPONSE).

Files#

• final.github.mp4: the copy under 10 MB; final.mp4: the master, a release asset; poster.jpg: frame 0.
• interactive/: the HTML export (index.html, assets/vfs.js with the runtime, scripts and fonts, assets/media/_export/soundtrack.m4a, assets/poster.jpg, socratic.json with the three questions, .nojekyll).
• project/: storyboard.md and storyboard.json (the plan), narration.md, showtime.json (questions, chapters), index.html, scenes/ (one script and style per shot, the roadmap, the corner mark), assets/ (the sensor geometry, the toy telescope's geometry, the 2016 event), audio/mix.json, voice/ (timeline, word times, captions; the WAVs are not committed).
• review/: round 1 (FINDINGS.md, RESPONSE.md, audio.txt) and round 2 (VERDICT.md, the two orders' FINDINGS.md, RESPONSE.md).
• receipt.md: what the job took, from what showtime recorded; share.txt, credits.txt.

To rebuild: showtime voice script project/narration.md -o project/voice (the voice WAVs), then showtime render project; showtime export html project --folder for the web version.

Sources and license#

• The film: made with showtime for nobel-2026-lab (github.com/FavioVazquez/nobel-2026-lab, MIT): educational demos of the 2026 Nobel Prizes, not research. Unofficial: not affiliated with the Nobel Foundation or the IceCube Collaboration. Facts from nobelprize.org, with a claims ledger in the lab repository; the toy-model numbers come from the lab's own simulations.
• Event data: IceCube Collaboration, data release for the Glashow resonance event, DOI 10.21234/gr2021 (the event of 8 December 2016).
• Sensor positions: IceCube ppc geometry (geo-f2k), D. Chirkin, Zenodo, doi:10.5281/zenodo.10410725, CC BY 4.0.
• Music: "The space is big" by Komiku (CC0), from the album "Little Cat In The Big Wide Space".
• Voice: Kokoro (Apache-2.0), voice am_michael, made on the rendering machine.
• Fonts: Space Grotesk and Inter (the Tidewater look) and Geist Mono, SIL OFL 1.1, from the fonts showtime installs.

The project files on GitHub