use-cases/learn-a-new-concept.md +217 −0 added
1# Learn a new concept | Codex use cases
2
3[← All use cases](https://developers.openai.com/codex/use-cases)
4
5Use Codex to study material such as research papers or courses, split the reading across subagents, gather context, and produce a Markdown report with diagrams.
6
7Intermediate
8
930m
10
11Related links
12
13[Subagents](https://developers.openai.com/codex/subagents) [Subagent concepts](https://developers.openai.com/codex/concepts/subagents)
14
15## Best for
16
17 - Individuals learning about an unfamiliar concept
18- Dense source material that benefits from parallel reading, context gathering, diagrams, and a written synthesis
19- Turning a one-off reading session into a reusable Markdown report with citations, glossary terms
20
21## Skills & Plugins
22
23- [ImageGen](https://github.com/openai/skills/tree/main/skills/.curated/imagegen)
24
25 Generate illustrative, non-exact visual assets when a Markdown-native diagram is not enough.
26
27## Starter prompt
28
29 I want to learn a new concept from this research paper: [paper path or URL].
30 Please run this as a subagent workflow:
31- Spawn one subagent to map the paper's problem statement, contribution, method, experiments, and limitations.
32- Spawn one subagent to gather prerequisite context and explain the background terms I need.
33- Spawn one subagent to inspect the figures, tables, notation, and any claims that need careful verification.
34- Wait for all subagents, reconcile disagreements, and avoid overclaiming beyond the source material.
35 Final output:
36 - create `notes/[concept-name]-report.md`
37- include an executive summary, glossary, paper walkthrough, concept map, method diagram, evidence table, caveats, and open questions
38 - use Markdown-native Mermaid diagrams where diagrams help
39- use imagegen to generate illustrative, non-exact visual assets when a Markdown-native diagram is not enough
40 - cite paper sections, pages, figures, or tables whenever possible
41 Constraints:
42 - do not treat the paper as ground truth if the evidence is weak
43 - separate what the paper claims from your interpretation
44 - call out missing background, assumptions, and follow-up reading
45
46## Introduction
47
48Learning a new concept from a dense paper or course requires more than just summarization. The goal is to build a working mental model: what problem it addresses, what the method actually does, which evidence supports it, what assumptions it depends on, and which parts you still need to investigate.
49
50Codex is useful here because it can automate the context gathering, and can turn complicated concepts into helpful diagrams or illustrations. This use case is also a good fit for [subagents](https://developers.openai.com/codex/concepts/subagents): one thread can read the paper for structure, another can gather prerequisite context, another can inspect figures and notation, and the main thread can reconcile the results into a report you can review later.
51
52For this use case, the final artifact should be something you can easily review: a Markdown file such as `notes/concept-report.md`, or a document of another format. It should include a summary, glossary, walkthrough, diagrams, evidence table, limitations, and open questions instead of ending with a transient chat answer.
53
54## Define the learning goal
55
56Start by naming the concept and the output you want. A narrow question makes the report more useful than a broad summary.
57
58For example:
59
60> I want to understand the main idea in this research paper, how the method works, why the experiments support or do not support the claim, and what I should read next.
61
62That scope gives Codex a concrete job. It should teach you the concept, but it should also preserve uncertainty, cite where claims came from, and separate the paper's claims from its own interpretation.
63
64## Running example: research paper analysis
65
66Suppose you want to learn about a paper about an unfamiliar model architecture. You want a report that lets you understand the concept at a glance, without having to read the whole paper.
67
68A good result might look like this:
69
70- `notes/paper-report.md` with the main explanation.
71- `notes/figures/method-flow.mmd` or an inline Mermaid diagram for the method.
72- `notes/figures/concept-map.mmd` or a small SVG that shows how the prerequisite ideas relate.
73- An evidence table that maps claims to paper sections, pages, figures, or tables.
74- A list of follow-up readings and unresolved questions.
75
76The point is to make the learning process more systematic and to leave behind a durable artifact.
77
78## Split the work across subagents
79
80Subagents work best when each one has a bounded job and a clear return format. Ask Codex to spawn them explicitly; Codex does not need to use subagents for every reading task, but parallel exploration helps when the paper is long or conceptually dense.
81
82For a research paper, a practical split is:
83
84- **Paper map:** Extract the problem statement, contribution, method, experiments, limitations, and claimed results.
85- **Prerequisite context:** Explain background terms, related concepts, and any prior work the paper assumes.
86- **Notation and figures:** Walk through equations, algorithms, diagrams, figures, and tables.
87- **Skeptical reviewer:** Check whether the evidence supports the claims, list caveats, and identify missing baselines or unclear assumptions.
88
89The main agent should wait for those subagents, compare their answers, and resolve contradictions. Codex will then synthesize the results into a coherent report.
90
91## Gather additional context deliberately
92
93When the paper assumes background you do not have, ask Codex to gather context from approved sources. That might mean local notes, a bibliography folder, linked papers, web search if enabled, or a connected knowledge base.
94
95If you're learning about an internal concept, you can connect multiple sources with [plugins](https://developers.openai.com/codex/plugins) to create a knowledge base.
96
97Keep this step bounded. Tell Codex what counts as a reliable source and what the final report should do with external context:
98
99- Define prerequisite terms in a glossary.
100- Add a short "background you need first" section.
101- Link follow-up readings separately from the paper's own claims.
102- Mark claims that come from outside the paper.
103
104## Generate diagrams for the report
105
106Diagrams are often the fastest way to check whether you really understand a concept. For a Markdown report, ask Codex for diagrams that stay close to the source material and are easy to revise.
107
108Good defaults include:
109
110- A concept map that shows prerequisite ideas and how they connect.
111- A method flow diagram that traces inputs, transformations, model components, and outputs.
112- An experiment map that connects datasets, metrics, baselines, and reported claims.
113- A limitations diagram that separates assumptions, failure modes, and open questions.
114
115For Markdown-first reports, ask for Mermaid when the destination supports it, or a small checked-in SVG/PNG asset when it does not. Ask Codex to use imagegen only when you need an illustrative, non-exact visual or something that doesn’t fit in a Markdown-native diagram.
116
117## Write the Markdown report
118
119Ask Codex to make the report self-contained enough that you can return to it later. A useful structure is:
120
1211. Executive summary.
1222. What to know before reading.
1233. Key terms and notation.
1244. Paper walkthrough.
1255. Method diagram.
1266. Evidence table.
1277. What the paper does not prove.
1288. Open questions and follow-up reading.
129
130The report should include source references wherever possible. For a PDF, ask for page, section, figure, or table references. If Codex cannot extract exact page references, it should say that and use section or heading references instead.
131
132## Use the report as a study loop
133
134The first report is a starting point. After reading it, ask follow-up questions and have Codex revise the artifact.
135
136Useful follow-ups include:
137
138- Which part of this method should I understand first?
139- What is the simplest toy example that demonstrates the core idea?
140- Which figure is doing the most work in the paper's argument?
141- Which claim is weakest or least supported?
142- What should I read next if I want to implement this?
143
144When the concept requires experimentation, ask Codex to add a small notebook or script that recreates a toy version of the idea. Keep that scratch work linked from the Markdown report so the explanation and the experiment stay together.
145
146Example prompt:
147
148Generate a script that reproduces a simple example from this paper.
149The script should be self-contained and runnable with minimal dependencies.
150There should be a clear output I can review, such as a csv, plot, or other artifact.
151If there are code examples in the paper, use them as reference to write the script.
152
153## Skills to consider
154
155Use skills only when they match the artifact you want:
156
157- `$jupyter-notebook` for toy examples, charts, or lightweight reproductions that should be runnable.
158- `$imagegen` for illustrative visual assets that do not need to be exact technical diagrams.
159- `$slides` when you want to turn the report into a presentation after the learning pass is done.
160
161For most paper-analysis reports, Markdown-native diagrams or simple SVG files are better defaults than a generated bitmap. They are easier to diff, review, and update when your understanding changes.
162
163## Suggested prompts
164
165**Create the Report Outline First**
166
167Before writing the full report, inspect [paper path] and propose the report outline.
168Include:
169- the core concept the paper is trying to explain
170- which sections or figures are most important
171- which background terms need definitions
172- which diagrams would help
173- which subagent tasks you would spawn before drafting
174Stop after the outline and wait for confirmation before creating files.
175
176**Build Diagrams for the Concept**
177
178Read `notes/[concept-name]-report.md` and add diagrams that make the concept easier to understand.
179Use Markdown-native Mermaid diagrams when possible. If the report destination cannot render Mermaid, create small checked-in SVG files instead and link them from the report.
180Add:
181- one concept map for prerequisites and related ideas
182- one method flow diagram for inputs, transformations, and outputs
183- one evidence map connecting claims to paper figures, tables, or sections
184Keep the diagrams faithful to the report. Do not add unverified claims.
185
186**Turn the Report Into a Study Plan**
187
188Use `notes/[concept-name]-report.md` to create a study plan for the next two reading sessions.
189Include:
190- what I should understand first
191- which paper sections to reread
192- which equations, figures, or tables need extra attention
193- one toy example or notebook idea if experimentation would help
194- follow-up readings and questions to resolve
195Update the report with a short "Next study loop" section.
196
197## Related use cases
198
199[
200
201### Coordinate new-hire onboarding
202
203Use Codex to gather approved new-hire context, stage tracker updates, draft team-by-team...
204
205Integrations Data](https://developers.openai.com/codex/use-cases/new-hire-onboarding)[
206
207### Generate slide decks
208
209Use Codex to update existing presentations or build new decks by editing slides directly...
210
211Data Integrations](https://developers.openai.com/codex/use-cases/generate-slide-decks)[
212
213### Analyze datasets and ship reports
214
215Use Codex to clean data, join sources, explore hypotheses, model results, and package the...
216
217Data Analysis](https://developers.openai.com/codex/use-cases/datasets-and-reports)