Slopcraft

Not a rulebook. A way of seeing.

We argue about AI like it’s a single object. It isn’t — it’s a system, with upstream costs (electricity ⚡, cooling water 💧, hardware 🏭) and downstream effects (attention, culture, durability). Slopcraft gives kids, teachers, and parents a shared language for that system: hands-on activities, a Slop vs Craft heuristic, resource awareness, and co-creative play.

The question is not “use it or don’t.” The question is: does value rise faster than cost — and where do the costs land?

The X-Scale: not all outputs cost the same

Impact grows stepwise across mediums. These are order-of-magnitude intuition tools — directional, not accounting. One text prompt ≈ 1× (~0.3 Wh, ~0.3 mL water, per Google & OpenAI 2025 figures).

   RELATIVE ENERGY PER OUTPUT   (vs 1 text prompt = 1×)
   ────────────────────────────────────────────────────
   TEXT    1×       ▏                            ~0.3 Wh
   IMAGE   ~10×     ▊▊                           ~3 Wh
   MUSIC   ~50×     ▊▊▊▊▊                        ~30 Wh
   VIDEO   ~3,000×  ▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊  5s  ~1 kWh
   VIDEO   ~40,000× ▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊▊→ 60s ~12 kWh
   ────────────────────────────────────────────────────
   a 5-second AI video ≈ a microwave running ~1 hour
   (text-to-video energy ~doubles each time length doubles)

Sources: Google Gemini env. report (Aug 2025); OpenAI/Altman (Jun 2025); MIT Technology Review AI video analysis (2025).

One pair of jeans = one “budget token”

A shared reference unit: one pair of fast-fashion cotton jeans (worn ~1 year, ~3,800 L of water to make). What does that same footprint buy across AI mediums?

   👖 ONE PAIR OF FAST-FASHION JEANS  ≈  ONE BUDGET TOKEN
   ────────────────────────────────────────────────────
   ≈  decades of text-AI thinking          💧 sips
   ≈  thousands of AI images              🍵 cups
   ≈  hundreds–thousands of AI songs      🍺 pitchers
   ≈  only 3–10 minutes of AI video       🏊 pools
   ────────────────────────────────────────────────────
   FAST FASHION + DISPOSABLE AI share one sin:
   high throughput, low durability.  Moral hinge → longevity.

The Slopcraft Compass

What to ask — not what to shame. Match the tool to the stakes.

   🧭 THE SLOPCRAFT COMPASS
   ────────────────────────────────────────────────────
   🟢 LOCAL / LIGHT LIFT   quick summaries · brainstorming
   🟡 MID-TIER / DRAFTING  moodboards · lo-fi loops · edits
   🔴 FULL-COST / CRAFT    feature films · essays of meaning
                          · performances (human hands + soul)
   ────────────────────────────────────────────────────
   think WITH ai · sketch WITH ai · decide with humans
   publish with care · scale only what you would repair

The Wise Centaur: human judgment + machine reach, aligned.

Interactive tools

Four thinking tools — not calculators of record. The numbers are rough, directional, and proportional. They exist to provoke better questions about cost, medium, intent, and care. Play-test them, argue with them, break them.

The Slop Index updated 2026

Order-of-magnitude intuition, not accounting. Compute cost tends to climb text → image → music → video. Model names churn fast — the pattern is what matters.

Note: OpenAI retired the Sora 2 consumer app in April 2026. Video ≈ ~$0.10/sec of compute and thousands of × the energy of a text prompt. AI Output ≠ Ethical Input — always ask what it was trained on, who benefits, and at what cost.

The five levers

What to ask, not what to shame:

1. Displacement — does this replace something more resource-intensive?
2. Durability — will anyone care in a month? a year? a decade?
3. Provenance — what's powering it: renewables, fossil grid, on-site turbines?
4. Water path — evaporative cooling, dry cooling, reclaimed water?
5. Defaults — are platforms incentivizing volume (slop) or signal (craft)?

In the classroom

Slopcraft is modular and age-targeted — tactile play for the little ones, ethical experimentation for teens.

Core activities across all ages: Slop Towers · DIY AI Cards · Resource Budget Challenge · Systems Chain Game · AI Detective · Centaur Alignment · Local Model Lab.