# OpenAI Study Plans
Four day-by-day study plans for the OpenAI tracks — a 1-week Foundations sprint, a 3-week Foundations-pathway plan, a 6-week developer plan for the API and Codex, and a 2-week leadership plan — each with reading, hands-on tasks, mock-exam checkpoints, a weekly review ritual and readiness gates.
import { Tabs, TabItem } from '@prosefly/astro-components';
All four plans use the same rules: study follows the mock's domain weights, you do the work in a real ChatGPT, API or Codex environment rather than reading about it, and you only move on after a mock exam clears the readiness gate. These plans prepare you for the **OpenAI Academy** assessments and pathway completion; the Academy assessments themselves are 10–20 questions from a 50-item bank at **80% to pass**, and Academy badges and pathway certificates are **not certifications**. Our mocks are independent practice, deliberately longer, with the same 80% line called out.
:::tip[Readiness gate]
Move on only when a timed mock scores **80% or higher overall** with **no single domain under 70%**. Under 70% overall means keep learning; 70–79% means drill your weak domains before advancing; 90%+ means you have a comfortable margin on a short randomised assessment.
:::
For someone who needs the *AI Foundations* badge fast. Mirrors the Academy AI Foundations course (60–75 min) in the Foundations pathway. About 10–12 hours over seven days.
| Day | Reading | Hands-on | Checkpoint |
| --- | --- | --- | --- |
| 1 | Foundations track overview; [credential landscape](/openai/credentials/) | Sign in to ChatGPT; write and refine three prompts using the [prompting cookbook](/appendix/openai/prompting-cookbook/) role and success-criteria patterns | — |
| 2 | Foundations D1–D2 (AI/LLM fundamentals, ChatGPT surfaces) | Compare a small and a large model on the same task; note the difference | Answer every in-page question in D1–D2 |
| 3 | Foundations D3–D4 (prompting, context) | Attach a file / connector and ground an answer; deliberately elicit a hallucination and detect it | In-page questions D3–D4 |
| 4 | Foundations D5–D6 (verification, responsible use) | Verify a Claude-style "confident" answer against a source; draft a one-line acceptable-use rule for a task | In-page questions D5–D6 |
| 5 | Foundations D7 (whatever the track's final domain is) + re-read weakest domain | Redo the hands-on for your weakest domain | [Mock exam 1](/openai/foundations/practice-exam/), untimed, as a diagnostic |
| 6 | Re-read the two weakest domains from mock 1 | Redo missed in-page questions aloud, explaining each distractor | Retry-incorrect-only on mock 1 |
| 7 | Light review, no new material | — | [Mock exam 2](/openai/foundations/practice-exam-2/), timed; take the Academy assessment if ≥ 80% |
Covers the whole Foundations pathway — AI Foundations + Applied AI Foundations + Agents and Workflows — toward the pathway certificate of completion. About 14–18 hours.
| Day | Reading | Hands-on | Checkpoint |
| --- | --- | --- | --- |
| **Week 1 — AI Foundations** | | | |
| 1–2 | Foundations track, D1–D4 | Prompt patterns; ground an answer; elicit and catch a hallucination | In-page questions |
| 3–4 | Foundations D5–D7 | Verify outputs; write an acceptable-use note | In-page questions |
| 5 | Review week 1 | Redo weakest hands-on | Foundations mock 1, untimed |
| **Week 2 — Applied AI Foundations** | | | |
| 6–7 | [Applied AI track](/openai/applied-ai/) D1–D3 (decomposition, input/output contracts) | Turn a recurring task into a documented workflow with a review point | In-page questions |
| 8–9 | Applied AI D4–D6 (capability selection, review points) | Add a validation gate to your workflow; choose the model per step | In-page questions |
| 10 | Review week 2 | Refine the workflow | Applied AI mock 1, untimed |
| **Week 3 — Agents and Workflows** | | | |
| 11–12 | [Agents track](/openai/agents/) D1–D3 (objectives, context, boundaries) | Write a delegation brief; run a bounded agent task with oversight | In-page questions |
| 13–14 | Agents D4–D6 (verification, reliability) | Add a done-condition and a failure path to the brief | In-page questions |
| 15 | Re-read weakest domain across all three | Redo missed items aloud | Foundations mock 2 (timed); take Academy assessments if ≥ 80% each |
Covers the API pathway (five courses) and the Codex pathway (three courses). Assumes coding fluency. About 22–28 hours. Do everything in the Responses API and Codex, not on paper.
| Week | Reading | Hands-on | Checkpoint |
| --- | --- | --- | --- |
| 1 | [API track](/openai/api/) D1–D2 (scoping, Responses API, model selection); [prompting cookbook](/appendix/openai/prompting-cookbook/) | A Responses-API script with structured output and validation-retry; pick a model per task by cost | API in-page questions D1–D2 |
| 2 | API evals domain; [evals cookbook](/appendix/openai/evals-cookbook/) | Build a 30-case golden set from real inputs; add exact-match and code graders | API mock 1, untimed |
| 3 | API agentic-systems domain; the three agent runtimes | Build a tool-using assistant; compare Responses+tools vs the Agents SDK vs the Agents API | API in-page questions |
| 4 | API RAG domain; [RAG cookbook](/appendix/openai/rag-cookbook/) | Build a small RAG pipeline: chunk, embed, hybrid retrieve, rerank; measure recall@k and faithfulness | API in-page questions |
| 5 | API performance domain; [governance & security](/appendix/openai/governance-and-security/) | Add prompt caching; set spend limits; wire an eval into CI | API mock 2 (timed); ≥ 80% before moving on |
| 6 | [Codex track](/openai/codex/) all domains; Codex models and reasoning ladder | Write a code-change brief; run Codex on a real repo with sandboxing and auto-review; use `codex exec` non-interactively | Codex mock 1 then mock 2 (timed); take Academy assessments if ≥ 80% |
Mirrors the standalone Academy *AI Leadership* course (180 min, no pathway). For owners and sponsors. About 10–14 hours. Pairs a little hands-on fluency with governance and adoption vocabulary.
| Day | Reading | Hands-on | Checkpoint |
| --- | --- | --- | --- |
| **Week 1** | | | |
| 1 | [Credential landscape](/openai/credentials/); the three credential mechanisms | Confirm which pilots or workspaces your org can access | — |
| 2 | Foundations track D1–D3 (fast, for fluency) | Do one real task in ChatGPT end to end so demos do not fool you | Foundations in-page questions D1–D3 |
| 3 | [Leadership track](/openai/leadership/) D1–D2 (opportunity selection, strategy) | Draft an opportunity shortlist with value and feasibility notes | Leadership in-page questions |
| 4 | Leadership D3 (governance); [governance & security](/appendix/openai/governance-and-security/) | Fill the sign-off matrix for one proposed use case | Leadership in-page questions |
| 5 | Review week 1 | Refine the shortlist | Leadership mock 1, untimed |
| **Week 2** | | | |
| 6 | Leadership D4–D5 (adoption, workforce enablement) | Draft an enablement plan and an acceptable-use policy outline | Leadership in-page questions |
| 7 | Leadership D6 (measurement) | Define two adoption metrics and one quality metric | Leadership in-page questions |
| 8 | Re-read weakest domain | Redo missed items aloud, explaining each distractor | Retry-incorrect-only on mock 1 |
| 9 | Light review, no new material | — | Leadership mock 2 (timed) |
| 10 | — | — | Take the Academy AI Leadership assessment if ≥ 80% |
## Weekly review ritual
End each week with the same 45-minute ritual, whatever the plan:
1. **Cold re-take** of the week's in-page questions you got wrong, without looking at the explanations first.
2. **Explain each distractor aloud** — if you cannot say why the wrong options are wrong, you have not learned the item.
3. **Re-do the weakest hands-on task** from scratch, faster this time.
4. **Update your gap list** — the two topics you are least sure of become the first reading next week.
5. **Check the gate** — if a mock is due, sit it; record the per-domain breakdown, not just the total.
## Daily micro-routine (15 minutes)
1. Read five [OpenAI glossary](/openai/glossary/) entries you cannot define from memory.
2. Re-answer two questions you got wrong yesterday, aloud, explaining every distractor.
3. Skim one decision table from a cookbook appendix and state the signal → choice rule it encodes.
## Readiness gates and how to read them
| Band | Overall | What to do |
| --- | --- | --- |
| Keep learning | under 70% | Gaps are structural; re-read the two weakest domains and redo their in-page questions before another mock |
| Building confidence | 70–79% | You know it but not reliably; drill weak domains, retry incorrect items only |
| Assessment ready | 80–89% | At or above the Academy badge threshold; take the Academy course then its assessment |
| Strong readiness | 90%+ | Comfortable margin on a short randomised form; move to the next track |
:::tip[Per-domain trumps the total]
A 90% overall with one domain at 3 of 8 still risks failing a short Academy form that happens to sample that domain twice. Weight your revision by *domain weight × your error rate*, never by weight alone, and never advance with a domain under 70%.
:::
## Honest limits of these plans
- Hour budgets, mock lengths and domain weights are **our design choices**, not published OpenAI parameters. The published Academy parameters are exactly two: 10–20 items from a 50-item bank, and 80% to pass.
- Academy course times are the published estimates (for example AI Foundations 60–75 min, AI Leadership 180 min); they will drift, so confirm on the [resources page](/appendix/openai/resources/).
- Nothing here guarantees a badge, a pathway certificate, or eligibility for any future OpenAI certification — that is OpenAI's own wording, and these plans repeat it rather than blur it.