Adam and Billy's Checklists

Live site: https://adamborecki.github.io/checklists/

Research-linked checklists for music teachers: preschool, K–12, college, private studio, TAs, and administrators (chairs, area directors, deans, provosts, presidents).

75 checklists in 18 categories, including health and wellness, English learners, autistic students, student safety, new faculty, and NASM accreditation. Filter by what you're teaching (private lessons, chamber and small ensembles, large ensembles, general music classes, lecture and academic classes) and who it's for.

Every checklist has:

• Checkable items (saved in your browser), many with a more button for detail
• Numbered citations that jump to the sources
• A Rationale section explaining why
• A Sources section with real, linked studies, laws, and guidelines
• An AI slop meter (in each checklist's Review section) showing how much has been checked, by AI and by people
• A History page with every prompt and change that made the site
• Emoji levels and celebrations (7 styles, or Surprise me) you control from the Display button
• Hearts for a quick favorites list
• A Your data page: everything is saved in this browser only, with export and import for backups

An AI (Claude) drafted all of this between 2026-10-22 and 2026-10-25. The first 47 checklists start at 85% slop, and the 28 newer ones at 93%, because the AI already verified their sources (links for 21 of them, claims for the other 7). See How the slop meter works.

Run it

It's a static site with no build step. Open index.html in a browser, or serve the folder:

python3 -m http.server 8417

Put it on GitHub Pages

1. Push this folder to a GitHub repo.
2. In the repo, go to Settings → Pages, set Source: Deploy from a branch, pick main and / (root), then save.
3. Your site appears at https://<you>.github.io/<repo>/.

.nojekyll is included so GitHub serves the files as-is.

Files

index.html                    page shell
css/style.css                 styles (light/dark, print, mobile)
js/app.js                     app: routing, filters, checklists, slop meter, storage
data/categories.js            categories, audiences ("Who it's for"), teaching settings ("What you're teaching"), evidence labels
data/emoji.js                 emoji levels, UI emoji, item keyword → emoji map, confetti glyphs
data/sources.js               the source library (every citation lives here once)
data/checklists-*.js          the checklists, grouped by area
data/history.js               site history: prompts (word for word), summaries, commits
tools/check-links.mjs         checks every source URL (node tools/check-links.mjs)

Emoji and celebrations

The Display button in the header has two settings, saved in this browser:

Emoji level	What you get
None whatsoever	No emoji anywhere. Emoji typed into checklist text are stripped too.
Barely any	One per checklist (on its page) and on category cards.
Tactful (default)	Checklist and category emoji in lists, one emoji in celebrations.
Plenty	Nav, headings, buttons, evidence labels, audiences, and items whose text matches a topic.
There's no such thing as too many	Every item, sparkles on titles, emoji confetti on every checkbox.

Emoji are decorative (aria-hidden), so screen readers skip them. Each checklist sets its own with emoji: "🎻". The keyword map that picks item emoji lives in data/emoji.js.

Celebrations fire on your first checked box, the halfway point, finishing a checklist, lowering a checklist's slop band, and reaching 0% slop. Pick a style:

Style	What it does (re-rolled every time)
Confetti (default)	Burst from where you clicked; big wins add side cannons. Random palette, shapes, tilt, spread.
Fireworks	Rockets launch and burst: sphere, ring, double, willow, or crossette.
Balloons	A bunch floats up with swaying strings.
Music notes	♪ ♫ ♬ drift up and sway (🎵 🎶 at higher emoji levels).
Sparkles	Twinkling starbursts in rings, plus stray sparkles for big wins.
Confetti shower	Paper and streamers fall from the top, blown by a random wind.
Bubbles	Bubbles wobble up and pop.
Surprise me	A different style each time, never the same twice in a row.
Quiet / Off	Just the message, or nothing.

Six palettes (theme, bright, pastel, gold, neon, marching band) are picked at random, and messages rotate. Previews play at half strength with a 3.5-second cooldown (the button shows a countdown). Real celebrations aren't throttled. With reduced motion turned on, only the message shows.

Events live in CELEBRATIONS in js/app.js (call celebrate("key", data, originElement)); styles live in EFFECTS.

Favorites and Your data

Tap the heart on any checklist to save it. The heart in the header opens your favorites (#/favorites).

There's no account. Checkmarks, favorites, reviews, source checks, and display settings are saved in this browser on this device only. The Your data page (#/data, linked in the footer and the Display menu) shows what's saved, exports a backup file, imports one (it merges, deleting nothing), and clears everything.

How the slop meter works

A cheap AI draft that checked nothing is 100% slop. Checks remove slop. People earn full points, and AI checks earn partial credit on the source checks only:

Check	AI can earn	Human earns
Read the whole checklist	0	−15
Fixed what was wrong or generic (or confirmed fine)	0	−20
Reviewed by someone who does this job	0	−20
Source links resolve to the right source	−7.5	−15 (split by source)
Sources support their claims (drafting AI matched)	−7.5	−30 (split by source)
…after an independent second AI double-check	−15 total

So: 100% = unchecked AI draft · 93% = AI verified links · 85% = AI verified links and matched claims to abstracts · ~78% = the most AI alone can reach · below that, only people.

Each checklist records its AI checks in an ai field:

ai: { links: true, claims: "drafter", note: "What the AI actually did, and when." }
// claims: "none" | "drafter" | "second-check"

A checklist with no ai field counts as an unchecked draft (100%). Only record an AI check that actually happened. To credit a second AI double-check, run an independent pass that re-reads each source and confirms the claim, then set claims: "second-check" and say so in note.

Reviews are saved in localStorage (this browser only). To make a review permanent for everyone:

1. On a checklist, open Review & AI slop meter and click Copy review for repo.
2. Paste the review: {...} block into that checklist's object in data/checklists-*.js.
3. For links you opened, add humanLinkCheck: "initials, YYYY-MM-DD" to the source in data/sources.js.
4. Commit. Everyone now sees the lower slop.

You can also Export all your reviews, checks, and favorites as JSON from the Your data page, and Import someone else's.

What the AI did and didn't verify

Sources added 2026-10-23 (the 21 newest checklists): Crossref and doi.org were unreachable from the drafting environment, so these were first checked by web search only. Titles, authors, journals, volumes, and pages matched search results. Later that day, all 147 sources were re-checked with tools/check-links.mjs from a machine that could reach Crossref. Every DOI resolved, including the three the cloud session couldn't confirm (boice1989, milton2012, paruthi2016). Claims in these 21 haven't been matched to their sources yet. NASM details (BM percentages, HEADS dates, the May 1 Code of Ethics rule) came from NASM pages and copies of past Handbooks, not a full read of the 2025–26 Handbook. Run node tools/check-links.mjs before trusting them.

Sources added 2026-10-25 (11 sources for the 7 lesson, chamber, and lecture-class checklists in data/checklists-settings.js): again checked by web search only from the cloud session. Claims were written to match abstracts and summaries found by search (claims: "drafter"), but links haven't been resolved by a live check (links: false), so these 7 also start at 93%. davidsonGood2002 links to the journal issue, because its DOI and page numbers weren't confirmed. Run node tools/check-links.mjs from a machine that can reach Crossref.

Sources from 2026-10-22:

• ✅ Resolved every DOI against the Crossref API, so titles, authors, journals, and years match real records.
• ✅ Opened or searched for every web URL. tools/check-links.mjs reports 0 broken. Nine sites (CDC, NAfME, ed.gov, NAEYC, Smithsonian Folkways, a stats blog) block automated checks and were confirmed by search instead.
• ✅ Read abstracts where available and wrote claims to match them.
• ❌ Did not read most full papers. Several claims come from abstracts or general knowledge of the literature.
• ❌ Did not check state laws, district policies, or institutional rules. Mandated reporting, FERPA, Title IX, Clery, English-learner, and copyright items are U.S.-centric.
• ⚠️ The 2015 federal English-learner guidance was rescinded in August 2025. The checklists say so and cite the underlying law (Title VI, Lau v. Nichols) instead.

This makes the "confirmed it supports how it's used" checks the most valuable thing a human reviewer can do.

Add or edit a checklist

Each checklist is a plain object:

{
  id: "unique-slug",
  title: "Doing the thing",
  emoji: "🎻",                         // shown depending on the viewer's emoji level
  fmt: ["small", "large"],             // settings from MTC_FORMATS; omit for role-only checklists
  cat: "rehearsal",                    // id from data/categories.js
  aud: ["secondary", "college"],       // audiences from data/categories.js
  tags: ["choir", "pacing"],
  evidence: "research",                // research | mixed | standard | practice
  summary: "One-sentence promise.",
  items: [
    { t: "Checkbox text.", more: "Optional detail.", src: ["duke2009"] }
  ],
  why: "Rationale. Separate paragraphs with a blank line.",
  sources: [ { k: "duke2009", use: "What this source is used for here." } ],
  ai: { links: true, claims: "drafter", note: "What the AI checked." },  // omit if nothing was checked
  review: { read: true, edited: true, expert: false, claims: { duke2009: true }, reviewer: "AB", date: "2026-11-01" }
}

Add new sources to data/sources.js first. The browser console warns about any missing source keys.

Site history

data/history.js holds one entry per prompt or change, oldest first: a quick summary, the prompt word for word (when we have it), clarifying Q&A, what happened, and commit links. Add an entry whenever a new prompt or edit changes the site. It renders at #/history, and each entry has a permalink like #/history/first-prompt.

Not advice

Nothing here is legal, medical, or mental-health advice. If a student is in danger, contact emergency services or, in the U.S., call or text 988.