Course reading draft · For review. This material is not a complete or graded course.

Understanding AI usage

Review draft for the TaiYlored learning plan. Including AI usage education is DECIDED. The public lesson title and final presentation are PROVISIONAL. This lesson is not published.

Your AI has an allowance

Think of AI usage like the data allowance on a cell phone plan. You can use the service, but some activities draw on the allowance differently. A paid plan can still have limits for particular models or tools. Check the allowance that applies to the thing you want to do. OpenAI usage guidance, Google usage guidance, Perplexity plan guidance

Taylor's remembered magic-meter analogy from The Santa Clause 2 gives us another way to picture it: an energy gauge goes down as you use it. Our illustration is an original teaching diagram, with no film imagery or branded watch design.

Illustrative energy meter: before a sample task, 8 of 10 example units remain; afterward, 3 of 10 remain. All numbers are invented for teaching. This is not your account balance or a provider's rate.

Text version: the sample task uses 5 imaginary units, leaving 3 of the original 8. These numbers do not predict any real task. A low meter says nothing about your intelligence, learning ability, or the quality of an answer.

There can be several meters

Usage allowance: how much of a model or service you can use during its measurement window. More demanding tasks may use more; one message need not equal one fixed unit
Tool limits: research, images, video, uploads, or other features may have their own limits. Some features share an allowance, so another mode is not necessarily a fresh pool. OpenAI, Perplexity
Context capacity: how much material the model can work with together. Text is processed in tokens, which can be words, word pieces, or punctuation. Context capacity is not your remaining subscription balance. A short answer can involve more processing than its length suggests. Tokens explained
Money or credits: what a particular product charges. An allowance reset is not a refund or automatic refill of every paid balance. For example, ChatGPT subscription billing and OpenAI API billing are separate. Billing explained

When does it come back

Read the provider's current usage notice. Some limits use moving time windows, where older activity eventually stops counting. Others refresh on a stated schedule. Different features can have different reset times, which need not match your subscription payment date. Starting a new chat does not itself reset your account allowance. OpenAI usage, research limits, billing dates

If you hit a limit, read which feature is affected and when it refreshes. You may be able to wait or use another available mode. Review any purchase or upgrade separately. Never assume a guessed percentage is your real balance.

Before you spend usage

Name the result. What do you need, and what would count as done?
Check the meter. Which tool or model, which allowance, and which reset time? If the app does not show it, record “unknown”
Choose the mode. Try ordinary chat for a small rewrite or quick follow-up. Consider deeper reasoning or research when complexity or multiple sources justify it. More thinking can use more allowance. Google thinking levels
Send a useful brief. Include the goal, essential facts, constraints, and desired output. Remove private details and duplicate material. Reuse checked facts instead of asking every tool to rediscover them
Check and stop. Open original sources, check important claims, and request only the missing correction. Save the useful answer. Stop when the result meets your criteria

Try this prompt: “Help me achieve [result]. Use these approved facts: [brief]. Return [format]. Identify anything you cannot verify. Before suggesting deep research or another tool, explain what it would add.”

Use this with the GPS method

A GPS request prepares prompts for Gemini, Perplexity, and Spark. You choose which are worth running and bring their responses back for comparison. Every Perplexity prompt states a mode and a reason: ordinary chat for a narrow check; Deep Research for a genuinely multi-source question. Check the current account's available labels and limits.

Track time, runs, rework, and any verified cost. Different providers' units and percentages are not directly comparable. Using several tools may add useful perspectives, but it can also increase total usage. Savings are something to measure, not promise.

Human Check: before your next AI task, explain which meter matters and why your chosen mode fits. If you cannot tell, check the provider's notice before proceeding.

Next: Plan clearly and hand work over covers efficient prompts, a separate planning/work conversation, dot and other helpers, optional Grok review, and when personalization helps or gets in the way.

Teaching and access notes

The diagram also ships as editable SVG with a title and full text description. The lesson's text communicates the same information without relying on colour or vision. No account connection, live usage data, subscription changes, telemetry, external assets, or automated cross-provider sharing is included. Provider documentation was checked on 1 October 2026; exact quotas, prices, model names, and feature availability should be checked again when used.