A browser sandbox for strength and conditioning education. Students plan three 4-week training blocks for a rhino, with a baseline test and a retest after each block, and watch the rhino’s performance change over a 12-week run. The rhino has three tasks: push a boulder up a mound, pull a stump out of the ground, and toss a tire over a river. Each task needs a different kind of training, so the right program depends on the goal.
Anchor principle: specificity. Training must match the quality being built. Running theme: Goldilocks. Too little stimulus gives no adaptation. Too much causes breakdown. The skill is finding the zone and moving it as the animal adapts.
Built with Claude (Anthropic). The model is directional, not exact, and it may contain errors. It is not a clinical tool.
| Version | State |
|---|---|
| 3 (idealized base model) | Final. This is the current build |
| 4 | Planned: audit fixes, About page, learning objectives, sandbox and guided modes, end-of-debrief questions, PDF report |
| 5 | Planned: realistic model (budget, outside stress, missed sessions, rhino traits, noise, group roles) |
| 1 and 2 | Earlier builds, kept for reference |
The v3 audit is in RHINO-v3-Audit-Report.docx (and .md).
Live page: https://lentinz.github.io/S-C_Animal_Sandbox/rhino-v3-base.html (GitHub Pages, from the main branch)
Repository: https://github.com/lentinz/S-C_Animal_Sandbox
Or download the repo and open rhino-v3-base.html (or dist/rhino-v3-base.html, the same file) in any browser. The page is one self-contained file with no server and no network use, so it also works from a course site or an LMS upload. Run history is kept in the browser’s local storage (key rhino-v3-base-runs).
If the rhino is injured or overtrained, the student decides to continue, modify the plan, or rest 1 to 4 weeks.
The full design is in Version-2-Build-Spec.md (sections 12 to 21 cover version 3) and Animal-Training-Sandbox-Design-Outline.md.
| Path | Purpose |
|---|---|
rhino-v3-base.html |
Copy of the standalone v3 page at the repo root, so GitHub Pages serves it at the address above. Generated by the build |
dist/ |
Build output folder: rhino-v3-base.html (standalone page) and rhino-v3-artifact.html (fragment) |
src/engine-v3.js |
The model: every rule and number |
src/v3/page.html, app.js, debrief.js |
Page, planner and animation, debrief |
build_v3.py |
Inlines the engine and scripts into the pages in dist/ and the root rhino-v3-base.html |
tests/v3/ |
Tests and search tools (see tests/README.md) |
Version-2-Build-Spec.md, Animal-Training-Sandbox-Design-Outline.md |
Design documents (.docx copies alongside) |
RHINO-v3-Audit-Report.md |
The v3 audit: findings, evidence table, v4 plan |
src/engine-v2.js, src/v2/, build_v2.py, dist/rhino-v2-*.html, tests/*.js |
Version 2 |
src/model.js, src/page.html, src/try.js, dist/rhino-prototype.html, dist/artifact.html |
Version 1 prototype |
Run from the project folder. Node 18 or newer and Python 3 are needed.
python3 build_v3.py # rebuild rhino-v3-base.html and dist/ after editing src/
node tests/v3/acceptance.js # 80 engine checks, about 30 seconds
node tests/v3/baseline-snapshot.js check # confirms the engine still matches the frozen v3 results
node tests/v3/playthrough.js normal # headless run of the page; also: injury, phone, dark
The playthrough needs Playwright (npm install playwright, then npx playwright install chromium). It saves screenshots to tests/v3/shots/, which is not tracked in git.
Change the engine in src/engine-v3.js, then rebuild. Do not edit the root rhino-v3-base.html or the files in dist/; the build overwrites them.
Every number lives in CFG at the top of src/engine-v3.js. After changing a rate or threshold, re-run node tests/v3/frontier.js and the acceptance checks, because the goal margins and score bands are set from them. Reference values:
| Item | Value |
|---|---|
| Sound balanced plans score | 84 to 92 out of 100 |
| Lopsided but safe plans score | 70 to 81 |
| Best plan found by search | 97 |
| Deload effect (test-sum change), moderate plan | block 1: -0.9, block 2: +0.2, block 3: +1.6 |
| Deload effect, hard plan | block 1: +16.5, block 2: +30.8, block 3: +31.4 |
tests/v3/baseline-v3.json freezes the v3 results. The realistic version must match it exactly with realism switched off.
| Animal | Quality | Possible course module | Signature obstacle | Status |
|---|---|---|---|---|
| Rhino | Maximal strength, pull strength and explosive power | Muscle performance | Plateau, taught through progressive overload and periodization | Built (v3) |
| Tiger | Muscle growth (hypertrophy) | Muscle performance | Overuse injury from too much volume | Planned |
| Sled dog | Endurance | Aerobic training | Overreaching or burnout, or a stall from one steady pace | Planned |
| Gazelle | Jump, speed and agility | Integration modules | Skill practice sets how much of the physical ceiling is used | Planned |
| Stretchable monkey | Mobility and flexibility | Flexibility and mobility | Strain from too much stretch without control | Possible |
| Mountain goat | Balance | Balance | Falls as the ledge terrain gets harder | Possible |
Posture does not fit the trainable-quality model and stays a regular module.
When a second animal exists, split the shared structure (inputs, weekly loop, meters, tests, scoring, findings) from the per-character content (configuration and art).