AI that helps you learn it, not a replacement for your brain.
Seventy-four templates for the part of college that is actually yours: the lecture notes, the reading, the practice, the seminar. You paste what you have and say what the test is on. Your AI builds the study guide, the practice set with the answer key that explains why, or the tutor that only asks questions. About five minutes, notes to study guide.
Runs on ChatGPT · Claude · Gemini · whichever AI is already open on your laptop
10 free prompts a month, no card
If you typed one of these into Google the night before, this is the page.
- how to make a study guide from my lecture notesStudy Notes Maker
- practice test with an answer key that explains whyPractice Worksheet Generator
- how to prep for a philosophy seminar without getting taken apartSocratic Method Framework
- turn my notes into Anki flashcardsFlashcard / Spaced-Repetition Deck Builder
- how to summarize a research paper for classResearch Paper Summary
- is my research question too broadResearch Question Refiner
- how to outline an argumentative essay before I write itEssay Structure Builder
- study schedule for a final in ten daysQuick Study Plan
- explain this concept like the test is tomorrowConcept Explainer
Nine of seventy-four. Study guides, practice sets, flashcards, research, essays, logic, focus, and a college budget. Every one of them starts from your material.
What you paste in from your notes. What comes back before the test.
Made the way you would make them: real notes and real sticking points in the blanks, run once, not cleaned up after. The student and the course are invented. The study guide, the practice set, and the drill are not.
A study guide built from one student's own BIO 201 lecture notes
Key Terms That Earn Their Place
| Term | Student-friendly definition | Versus note |
|---|---|---|
| Fermentation | Anaerobic pathway that regenerates NAD+ from NADH so glycolysis can continue; yields lactate or ethanol + CO2, only 2 ATP total. | vs. Anaerobic respiration (uses an alternate terminal electron acceptor in an ETC). |
| NADH (about 2.5 ATP) | High-energy electron carrier from glycolysis, pyruvate oxidation and Krebs; feeds electrons at Complex I. | vs. FADH2 (about 1.5 ATP), enters at Complex II, fewer protons pumped. |
| CO2 source in exhaled air | Produced only during pyruvate oxidation and the Krebs cycle; the ETC does not generate CO2. |
Main Ideas, as a Logic Chain
- When O2 is absent: ETC stalls, NADH cannot be oxidized, glycolysis halts unless NAD+ is regenerated by fermentation. Fermentation yields only the original 2 ATP.
- Blockage consequences: Complex IV inhibition (cyanide) = electron backlog, no H+ pumping, gradient collapses, ATP stops. Uncoupling (DNP) lets protons flow back without making ATP: heat, wasted fuel.
The 20-Minute Study Plan
0 to 5 min: Cover the one-screen summary. Write it out from memory on scrap paper. Check against the sheet. 18 to 20 min, the self-test rule: if you can answer all 12 key terms, three location questions, and two application questions without looking, you are ready. If any slip, repeat that block until fluent.
Plus the one-screen summary, the full 12-term table, eight likely test questions with model answers, and memory hooks for the three hardest items (NADH rides the double-decker bus; fermentation is home karaoke with no power grid). Six sections.
A 12-problem t-test practice set, with the answer key that says why
6. One-sample, the "claim in H0" trap
A clinical trial states: "The new drug does not increase mean recovery time beyond 10 days." Sample (n=15) yields t = 2.48, critical value (one-tailed, a=.05) = 1.75. Give the correct hypotheses, tail, df, decision, and conclusion.
11. One-sample, "reject vs fail to reject"
A coffee shop claims the average caffeine content is at least 95 mg per cup. Sample (n=13) gives t = 0.47, critical value (one-tailed, a=.05) = 1.77. State hypotheses, tail, df, decision (pay attention to "t smaller than critical"), and plain-English conclusion.
Answer Key and Common-Mistake Notes
| # | H0 / Ha | Decision | Common-mistake note |
|---|---|---|---|
| 6 | H0: mu is at most 10. Ha: mu is greater than 10. One-tailed (right) | df 14. Reject (t=2.48 is greater than 1.75). The drug does increase average recovery time beyond 10 days. | Putting "at most 10" in Ha is the classic reversal error. |
| 11 | H0: mu is at least 95. Ha: mu is less than 95. One-tailed (left) | df 12. Fail to reject (t=0.47 is greater than -1.77). Not enough evidence that caffeine is below 95 mg. | Students often think "smaller t means reject"; for a left tail, reject when t is at or below the negative critical value. |
General reminder on wording: Never "accept" H0. Say fail to reject or do not have sufficient evidence. The alternative hypothesis always contains the claim.
Plus the warm-up problems, two find-the-error problems written the way you make the mistake, an explain-your-thinking paragraph, the full 12-row key, a helper's tip for whoever is checking it, and a pointer for what tomorrow's practice should target.
A Socratic rehearsal for Thursday's free-will seminar
(1) Define the concept clearly
What exactly are we claiming when we say "free will is real"? Does "could have done otherwise" refer to metaphysical possibility (a real branch of the world) or merely epistemic ignorance (we don't know what will happen)? Write down your working definition in one sentence. Anything you later deny must contradict this sentence.
(3) Challenge the assumptions underneath
- Assumption C: moral responsibility requires alternatives. You claim "if free will didn't exist nobody could be blamed, and that is wrong." Why must blame require alternatives? Could we hold people responsible on the basis of desert or reactive attitudes even if actions were determined?
(4) The counterexample
Frankfurt's case: you intend to vote for candidate A. Unbeknownst to you, a device would intervene and make you vote for B if you tried to vote otherwise. You vote for A; the device never activates. If you accept Frankfurt's conclusion, responsibility does not require alternatives, so the link between free will and moral blame is weaker than you thought. Either deny Frankfurt or concede. Either move forces you to revise your definition.
Open question for Thursday
If an agent's action is fully determined by prior facts and the agent nevertheless endorses the reasons for that action, can we coherently say the agent "could have done otherwise" in any meaningful sense?
Plus the edge cases (habit, coercion, compulsion), the refined definition, what the concept still cannot explain (Libet-style timing experiments, moral luck, the compatibilist reinterpretation), and the stronger thesis to open with. Seven steps, and it never answers for you.
Three steps. Notes to study guide.
Pick the study tool
Study guide, practice set, flashcard deck, seminar drill, research question check, essay outline. Search it the way you'd text a friend in the class.
Paste what you already have
Your lecture notes, the chapter, your draft, the claim you have to defend. Then what the test is on and where you're stuck. Five to seven blanks, chips for the common answers, your own words for the rest.
Send it to your AI
Ultra Prompt builds the prompt and hands it to ChatGPT, Claude, or Gemini. Back comes the study guide, the practice set with the key that explains why, or the questions you have to answer yourself. Your work stays yours.
Templates handle tonight. A study agent handles the semester.
The syllabus, your notes, and the way you actually learn can live inside an AI agent that knows the course, so every study session starts from your material instead of a blank chat. Three study agents worth building, and none of them writes the assignment:
The Course Tutor That Only Asks Questions
Holds the syllabus, this week's readings, and a rule it will not break: it never hands you the answer. Paste a concept or a claim you want to defend; it asks the next question until you can explain it back without notes.
Runs in Claude Projects, a Custom GPT, or a local modelThe Flashcard Drill
Knows your notes from every lecture so far. Say the chapter; it drills you, tracks what you missed, and comes back to it tomorrow before it comes back to what you know.
Runs in your phone, through the AI you already useThe Citation and Outline Checker
Holds your assignment prompt, your citation style, and the sources you already found. Paste your outline; it flags claims with no source behind them, sources you cited but never used, and the gap between your thesis and your evidence. Your words stay your words.
Runs in Gemini, ChatGPT, or ClaudeTheo, the study desk, a tutor that asks before it answers, is being built.
Theo is in the build lane now and goes through the same certification every showroom agent passes before the door opens. Meanwhile, the live agents are open to everyone.
See the showroomBuild the course tutor in an afternoon.
Agent Studio is in every Pro and Business seat. Give it the syllabus, the readings so far, the professor's phrasing, and one rule: it asks, it does not answer. Then, the night before seminar, paste in the claim you have to defend; back comes the question you did not see coming, then the next one, until you can hold the argument on your own. Export it as a Claude Project, a Custom GPT, a Gemini Gem, Copilot, or a local model on your laptop. It runs on the AI you already have. Nothing to host, nothing to install in the library.
Open Agent StudioOr have it built around your courses.
Want it built for you? A custom build is one conversation about what you do every week, then an agent brief written for you, exported to your AI, with a walkthrough. From $450, one time.
Ask about a custom buildFiles are yours to keep. The library, updates, and support ride the subscription.
Two ways in. Both honest.
Pro is $5 a month: unlimited generates, the full 1,000+ library (every study template here, plus the rest of life), and Agent Studio. The free tier is 10 finished prompts a month, no card.
Start with ProFree to start: 10 finished prompts a month, no card. Cancel anytime.
Build the tutor that only asks questions, the flashcard drill on your own notes, or the citation checker for your major, and run it on the AI you already have. Included with Pro.
See Agent StudioThe questions students ask before they trust it.
Is this cheating?
It does not do the assignment for you. It helps you understand it: a study guide from your notes, a practice set with the reasoning, a tutor that asks questions. Check your school's AI policy, and where you use it in submitted work, cite it the way your professor asks.
Do I need ChatGPT or Claude to use this?
Yes, one of them, and the free tier of any of them works. Ultra Prompt builds the prompt and sends you there with it. Most students already have one.
Will it make up facts or sources?
It works from what you paste. These templates ask for your notes, your reading, your sources, and tell the AI to say plainly when something is not provided rather than invent a study, a page number, or a statistic. You still check anything you plan to cite, and you still check an answer key against your own working; the AI is a study partner, not the professor.
How is this different from just asking ChatGPT to explain it?
You can ask ChatGPT to explain cellular respiration and get the textbook again. These templates were written by people who know that rereading feels like learning and is not: they ask what the exam emphasizes, where you are stuck, and what you already have, and they build the self-test in. The difference is that they make you do the recall.
Does it write my essay?
No, and you would not want it to. It outlines with you, checks your draft against your thesis, and explains a grammar rule you keep breaking. The sentences are yours; the feedback is on them.
What does it cost a student?
Free: 10 finished prompts a month, no card. Pro: $5 a month, unlimited generates, the whole library, Agent Studio. Cancel anytime.