← 返回 T1ENS3 · Ground Your Wukong — Model + Knowledge
T1ENS3 · 课程资料

📝 S3 Homework 1 Guide (warm-up · not submitted · token cost reality check)

S3 Homework 1 (Warm-up · Not Submitted) — Do the Math: Monthly Fee vs Real Token Burn

This is the S3 warm-up homework — no submission required.

Purpose: turn S3's token economics from an abstract formula into a real number in your wallet. Most people who subscribe to ChatGPT Plus / Claude Pro have no idea how many tokens they actually burn each month, what the unit price is, or whether they're paying "sucker money". Spend 10 minutes doing the math and you complete the first cognitive step from "consumer" to "craftsman".


What you're doing (3 steps)

1. Pick the AI you use most

Pick the one you're paying monthly for or using most often:

Provider Monthly API price (input / output · per 1M tokens)
ChatGPT Plus $20 GPT-4o: $2.5 / $10
Claude Pro $20 Claude 3.5 Sonnet: $3 / $15
Cursor Pro $20 Mixed (avg ~$3 / $15)
Gemini Advanced $20 Gemini 1.5 Pro: $1.25 / $5
DeepSeek Free / top-up V3: $0.27 / $1.10 (cheapest)

2. Estimate how much you actually burned (rough is fine)

Think back over the past month:

Average chat turns per day :  ____ turns
Average input words per turn: ____ words
Average output words per turn:____ words
Chat days in the month     :  ____ days

Using the S3 conversion (1 English word ≈ 1.3 tokens, 1 Chinese character ≈ 1.5 tokens) and the API prices above, compute:

Monthly input tokens   = days × turns/day × input words × 1.3
Monthly output tokens  = days × turns/day × output words × 1.3

Monthly real cost (at API price)
   = (input tokens × input price + output tokens × output price) ÷ 1,000,000

3. Compare monthly fee vs actual burn

💰 Your monthly fee    : $______
🔥 Your actual burn    : $______
Gap = monthly fee − actual burn = $______

Three typical situations · which are you?

🌱 Type 1: You paid "sucker money"

Monthly fee $20, but you actually burned only $2-5 in API terms.

Meaning: the provider kept 75%-90% of your money. What you bought for convenience is a "psychological insurance policy".

Think about it: are you willing to pay this premium for convenience? Or should you call the API directly (DeepSeek or OpenAI API) so the money goes to actual compute?

🌳 Type 2: Your monthly fee "earned it back"

Monthly fee $20, but you actually burned $30+ in API terms.

Meaning: the provider gave you a 30% discount. You're a power user; the subscription pays for itself.

Think about it: you're already on the power-user path. Next question: are these tokens spent on YOU (consumer) or on the agent YOU'RE BUILDING (craftsman)?

🏆 Type 3: You're already running a business

You call the API directly, spending tens or hundreds a month, but your agent is helping others and users pay YOU.

Meaning: AI paid back its own cost. You're on the third layer (boss layer).

Think about it: what's the ratio (what you pay AI vs what users pay you)? How do you scale it 10×?


Don't forget what S3 said in class

"If you don't write the script, you're a character in someone else's."

This exercise isn't about "saving money" — it's about seeing clearly which layer your money is on. Once you see it, decide: keep paying the monthly fee / switch to direct API / start building your own Little Wukong.

No submission required, but strongly recommended — after the math, write two sentences to yourself:

Right now I'm on Layer ___.
Before S6 graduation, I want to reach Layer ___.

—— Stick these two sentences on your desktop for 6 weeks.