Module 1 of 7 · 44 min

Navigate the OpenAI Ecosystem

Choose deliberately among ChatGPT workspaces, the OpenAI API platform, and Codex development workflows without transferring capabilities or controls between them.

OpenAI

By the end

You will be able to

  • Distinguish conversational, long-running work, API-integrated, and coding-agent surfaces.
  • Select an OpenAI surface from user, automation, control, data, tool, and operational requirements.
  • Map identity, credentials, data, permissions, review, and evidence for the selected workflow.
  • Verify volatile plans, models, features, limits, regions, and prices from current official sources.
01

One provider has several operating surfaces

ChatGPT supports direct conversational work and organized project context. Longer-running work surfaces can coordinate files, connected tools, and review checkpoints. The API integrates model behavior into software, while Codex supports repository-centered development workflows.

These surfaces can share provider capabilities without sharing authentication, billing, data handling, memory, tools, permissions, or operational ownership. Always name the exact surface when documenting a requirement.

02

Match the surface to the outcome

Use supervised chat for exploration, explanation, drafting, and analysis. Use a project or long-running workspace when related context, files, instructions, and follow-up work must remain organized. Use the API when an application must enforce input, output, tool, identity, or reliability contracts.

Use Codex when the authorized outcome requires repository discovery, code changes, commands, tests, review, and source-control evidence. A workflow may move across surfaces, but every transition needs an explicit handoff and control review.

03

Map controls per surface

Record the actor, account or service identity, credential boundary, trusted and untrusted data, storage and retention arrangement, available tools, permission scope, budget, audit evidence, human approvals, and failure owner.

Do not infer an API data policy from a ChatGPT setting or assume a ChatGPT capability is exposed through the API. Keep API keys out of prompts, examples, source control, browser code, logs, and learner evidence.

OpenAI surface decision record
text
outcome: [USER-VISIBLE RESULT]
surface: [CHATGPT | WORKSPACE | API | CODEX]
actor: [PERSON OR SERVICE]
identityAndCredential: [REFERENCE ONLY, NEVER A SECRET VALUE]
trustedData: [SOURCES + CLASSIFICATION]
toolsAndPermissions: [ALLOWLIST + LIMITS]
validation: [OUTPUT + POSTCONDITION CHECKS]
humanReview: [TRIGGER + OWNER]
currentEvidence: [OFFICIAL URL + VERIFIED DATE]
04

Separate stable principles from volatile facts

Authentication, least privilege, output validation, data minimization, evaluation, observability, and recovery are stable operating principles. Model identifiers, feature maturity, plan availability, limits, pricing, supported regions, and product names can change quickly.

Keep volatile facts in linked, dated evidence rather than long-lived routing rules. Revalidate the exact plan, account, deployment, and region before procurement, release, or user-facing commitments.

05

Build a reviewable surface map

Connect each user journey to the selected interface, input sources, output consumer, tools, approvals, evidence, exclusions, and operating owner. State what the workflow intentionally cannot do.

Review the map with product, security, data, accessibility, and operations stakeholders. Attach official evidence for every plan-specific or newly released dependency and assign a revalidation date.

Practice activity

Design an OpenAI surface map

  1. Choose one realistic workflow and document its user, outcome, data sensitivity, frequency, integration needs, review needs, and consequence of failure.
  2. Compare ChatGPT, an organized workspace, the OpenAI API, and Codex; mark unsupported and unverified assumptions.
  3. Select one surface and map identity, credentials, data, tools, permissions, validation, logs, approvals, budgets, and failure ownership.
  4. Attach at least three official OpenAI sources and date every volatile capability claim.

What to produce

  • A four-surface comparison with a justified selection, rejected alternatives, and explicit non-goals.
  • A secret-free control map linking volatile claims to dated official evidence and accountable owners.

Reflect before continuing

Which requirement changed the selected surface, and which product fact must be revalidated before release?

Evidence

Sources and verification

Knowledge check

Make it stick.

Pass at 80%

Choose the strongest answer for each question. Your attempts become part of your device-local transcript.

01What should determine the OpenAI surface?
02Which statement about product controls is correct?
03Where should an API key be stored?
04Which fact needs frequent revalidation?
05What makes the surface map implementation-ready?