Token Steward Teacher Guide
Phase 1 · Lesson 1 · Lovable AI
Becoming a Token Steward
Duration: 60 Mins
Target Grades: 6–12 (Beginner)
Guiding Mantras
"A credit is a decision."
Planning before prompting saves platform credits, electricity, and water.
Prerequisite
Vibe Coding: Built Game
Students need their existing Lovable project loaded.
Free Tier Limits
5 Credits / Day
Resets midnight UTC (up to 30/month; no rollover).
Privacy & Safety
COPPA Compliant
Never input student names, emails, or personal data.
Core Learning Targets
-
Differentiate: Distinguish tokens (compute units) from credits (platform billing).
-
Select Tools: Pick Visual Edits (free), Try Fix (free), Plan (1c), or Build.
-
Apply Protocol: Execute STOP-THINK-READ-USE before spending any credit.
-
Precision Rewrite: Transform vague prompts into high-efficiency instructions.
Essential Classroom Frameworks
THE S.T.O.P. PROTOCOL Pre-Flight Prompting Checklist
The Three Precision Rules
1. Name It Specify the exact component (e.g., "the start button", "score counter").
2. Say the Change Give explicit metrics & actions (e.g., "count down by 1 each second from 60").
3. Describe Done State observable outcome (e.g., "when it reaches 0, display 'Game Over'").
AI Literacy & Resource Stewardship · Lovable Edition Page 1 of 4
Instructional Sequence
Lesson Flow & Facilitation (60 Mins)
Pacing Guide
1 Hook: Where Did My Credits Go? 5 Min
Prompt students to open their existing game from the prerequisite lesson and reflect on prior runs.
Discussion Prompt: "Who ran out of credits last time? Did using more prompts make your game better?"
Real-World Case: Share the 2026 'tokenmaxxing' phenomenon—tech giants closed employee token leaderboards when developers burned tokens without shipping better code.
2 Tokens vs. Credits & The Arcade Analogy 8 Min
Bridge the technical gap: Tokens = raw computational fuel; Credits = Lovable's simplified meter.
Arcade Token = Credit
Visible price per play (e.g., 1 credit in Plan mode).
Machine Wattage = Token
Underlying compute cycles (hidden tokens processed).
Action: Have students click the three dots under any previous prompt in Lovable to view its exact cost.
3 Tool Selection: Right Tool for the Job 10 Min
| Tool Mode |
Cost |
Best Used For |
| Visual Edits |
Free |
Colors, typography, copy changes, simple margins. |
| Try to Fix |
Free |
Native build errors flagged in Lovable console. |
| Plan Mode |
~1 Credit |
Architecting ideas without modifying source files. |
| Build Mode |
Varies (~0.5–2+) |
Generating logic, state mechanics, and new features. |
4 The Planet Pays Too 5 Min
Data center awareness: Median prompt = ~0.24 Wh electricity & ~0.26 mL water (cooling). At scale across billions of queries, efficiency matters.
5 Plan Before You Prompt 7 Min
Model turning "make game harder" into a 3-rule precision prompt. Introduce the 5-Prompt Budget worksheet protocol.
6 Independent Practice: The Efficiency Mission (18 Min) & Debrief (7 Min) 25 Min Total
Student Goal: Ship 1 meaningful game upgrade using ≤ 2 Build mode prompts. Record planned vs. actual credit spend. Use Visual Edits for styling.
Showcase Debrief: Who achieved their feature with the lowest credit tally? What failed? Target: High Value, Low Spend
AI Literacy & Resource Stewardship · Lovable Edition Page 2 of 4
Evaluation & Misconceptions
Checking for Understanding & Rubric
Assessment Suite
Formative Check Quiz (5 Questions) Answer Key: 1-B, 2-C, 3-C, 4-B, 5-C
1. Which Lovable mode lets you brainstorm architecture without editing code?
A) Build Mode · B) Plan Mode (Correct) · C) Visual Edits · D) Try to Fix
2. Best way to change background color from white to navy?
A) Build prompt · B) Plan then Build · C) Visual Edits (Free - Correct) · D) Chat agent
3. Why do vague prompts like "make it cool" waste credits?
A) AI guesses and rewrites extra code · B) Rework costs more credits · C) Both A and B (Correct) · D) Long prompts cost more
4. Key difference between a token and a credit?
A) Same thing · B) Token = raw model text chunks; Credit = Lovable billing metric (Correct) · C) Tokens are only images
5. Which prompt satisfies all Three Precision Rules?
A) "Fix timer" · B) "Timer broken" · C) "Make game timer start at 30s, drop by 1/s, show Game Over at 0" (Correct)
Frequent Student Misconceptions
| Student Misconception |
Technical Reality |
| "Shorter prompts always cost less." |
Cost tracks the code changed. Vague short prompts force broad rewrites. |
| "Refreshing the browser resets my daily 5 credits." |
Credits bind to user accounts in the cloud; reset happens at midnight UTC. |
| "The AI will intuit what I intended." |
LLMs guess the most generic probability. Only explicit rules yield your vision. |
| "Every single click in Lovable spends credits." |
Visual Edits & Try to Fix are completely free of credit charges. |
Token Stewardship Rubric
Beginning
Spends credits impulsively; uses Build mode for colors and minor text tweaks; prompts lack target names.
Proficient
Separates styling (Visual Edits) from logic (Build); writes 3-rule precision prompts; logs actual credit spend.
Exceptional
Plans with Plan Mode first; achieves complex logic in ≤ 2 Build prompts; articulates physical compute & environmental costs.
AI Literacy & Resource Stewardship · Lovable Edition Page 3 of 4
Teacher Reference Appendix
Technical Deep Dive & Extension Activities
Appendix A–C
Appendix A: How LLMs, Tokens & Credits Actually Connect
Large language models process text in sub-word fragments called tokens (roughly 0.75 words per token). In standard chat, you only pay for what you type and read. But in an agentic code builder like Lovable:
1. Context Ingestion: Lovable feeds the model your prompt PLUS system instructions and relevant repository code files.
2. Iterative Generation: If instructions are unclear, the AI inspects multiple candidate files, racking up compute tokens behind the scenes.
3. Credit Calculation: Lovable translates these underlying token counts into simplified platform credits (e.g., 0.5 to 2.0+ credits).
Appendix B: Follow-On Student Challenges
The 1-Prompt Showdown (Speed Challenge)
Provide a strict feature requirement (e.g., "Add 3-heart life system, losing 1 heart when tapping an odd number"). Award honors to the student who accomplishes full implementation in exactly one Build prompt.
Precision Prompt Clinic (Worksheet Exercise)
Have students take 3 vague legacy prompts from peers ("Fix timer", "Make enemies spawn better", "Update theme") and rewrite them using Name It / Say Change / Describe Done format.
Environmental Benchmarks
- • Median Gemini query: ~0.24 Wh electricity (~1 sec microwave use).
- • Median cooling water: ~0.26 mL (~5 water drops) per query.
- • Code generation with agent steps consumes significantly more resources than simple chat queries.
Standard Alignment
- • ISTE: 1.1a, 1.1d (Empowered Learner), 1.2b (Digital Citizen), 1.4a-d (Designer).
- • CSTA: 2-AP-10, 2-AP-13, 3A-AP-16 (Algorithms), 2-IC-20, 3A-IC-24 (Impacts).
Documentation Note: Lovable periodically refines credit usage meters. The pricing benchmarks cited here reflect platform documentation as of November 2026. Always confirm live tier details on the Lovable dashboard prior to class.
AI Literacy & Resource Stewardship · Lovable Edition Page 4 of 4
