The Game Has ChangedThe Internship Has Changed

AI is sweeping through the world of software engineering. For students seeking an internship or their first role, that may appear intimidating. Experienced engineers have powerful tools, companies are moving faster, and expectations are rising.

But there is another side to the story. AI gives a well-prepared intern capabilities that once required months or years of experience. With the right preparation, an intern can build prototypes, explore unfamiliar codebases, automate processes, integrate platforms, test ideas, deploy software and communicate a credible product story.

Earlier: the internship as a learning experience

Internships were primarily designed around learning. Companies expected interns to attend training, complete courses and gradually become familiar with their tools. Students, understandably, arrived expecting to be taught.

Now: the internship as a performance showcase

Today, companies expect interns to complete sprints, solve process problems, create proofs of concept, explore product platforms, work like junior Forward Deployed Engineers (engineers who sit with a customer and build working solutions on top of a product), or create and maintain websites—on Day 1. All right, perhaps Day 7.

The expectation is not that interns know everything. It is that they can learn rapidly, use AI responsibly and turn an unfamiliar problem into a working outcome. Every completed project sends a signal:

• ✓  I can understand a problem.
• ✓  I can learn what is necessary.
• ✓  I can use AI responsibly.
• ✓  I can produce something that works.
• ✓  Yes, I can.

The PlanA Four-Week Intensive Transformation

This is not a couple-of-hours-a-day course. It requires approximately six to eight hours of focused work each day. It also assumes you have a paid AI subscription—ChatGPT Plus, Claude Pro or Google AI Pro—and can spend a small amount on API credits and, in Week 4, a domain. The free tiers will slow you down precisely in the weeks that matter most. Each week adds a new layer of capability, evidence and professional communication.

Week 1

Build Your AI Foundation

Assume that you know something about AI and have experimented with tools such as ChatGPT, but do not yet understand the landscape in depth. Week 1 builds that foundation.

Understand the AI landscape

Learn the fundamentals of generative AI, large language models, prompting, tokens, context windows, model weights, reasoning models, multimodality, agents, proprietary models and open-weight models.

Explore the offerings of OpenAI, Google and Anthropic. You should be able to explain how products such as ChatGPT, Codex, Claude, Cowork, Gemini, Nano Banana and NotebookLM differ, what they are designed to do and when one may be preferable to another. The objective is not to memorise a product list; it is to learn how to understand new AI products as they emerge.

Build an AI information system

Use three principal sources: official company websites and documentation, credible YouTube creators, and AI researchers and practitioners on X. Add one or two reliable podcasts or daily AI news programmes so that you begin following launches, open-weight models, Chinese AI developments and changes in agent capabilities.

YouTube can be valuable but unnecessarily long. Add selected videos and other sources to topic-specific NotebookLM notebooks. Read transcripts, compare sources and ask questions across them. Use summaries to accelerate learning, but return to the original material whenever a claim is important.

Learn prompt engineering properly

Complete at least one structured prompt-engineering course of approximately one to two hours. Take notes and practise giving the model context, a clearly defined task, examples, constraints, output format and success criteria.

Week 2

Prepare for Agentic Development

You already know the fundamentals of software development. This week focuses only on what is new: preparing to develop software with autonomous AI agents. None of it needs to be figured out alone: ask your AI system—ChatGPT, Claude or the like—for step-by-step help at every stage. The system will guide you.

Set up your AI-native development environment

Choose one primary coding agent: Cursor, GitHub Copilot, Codex, Claude Code, Gemini CLI or another capable tool. Configure it with your editor, terminal, GitHub account, runtimes, package managers, secure API credentials, build and test commands, and a simple deployment environment.

Confirm that the agent can inspect a repository, edit multiple files, execute commands, read failures and run tests. Understand the progression from code completion and conversational assistance to IDE agents, asynchronous agents and loop-based development.

Complete one small coding task with your agent

Fork an existing open-source project that uses a technology you already understand. Do not send experimental changes to the original project. Suitable repositories include:

First ask the agent to explain the repository structure, execution flow, tests and likely files involved. Verify its explanation yourself. Then ask it to plan, implement and test one small, clearly defined change. Review the diff and independently run the application and tests.

Optional free APIs

An API integration gives the agent an external system to understand. Appropriate choices include JSONPlaceholder for posts and users, DummyJSON for product data, Open-Meteo for weather, PokéAPI for Pokémon data, Open Library for book search, and the GitHub REST API for public repository information.

Study Ralph loops

A Ralph loop is a technique from the agentic-coding community in which a coding agent is run again and again against the same objective—each run picking up where the last one stopped—until the work is genuinely done. Learn how a Ralph loop differs from one autonomous task. In a loop, the agent repeatedly examines the objective and current state, chooses the next bounded task, implements a change, verifies it, records progress and begins another iteration until defined completion conditions are met.

Study why loops require clear objectives, persistent progress information, automated tests, controlled permissions and reliable stopping conditions. Understand where they can fail and why human verification remains necessary. You will apply the approach in Week 3.

Week 3

Build with an Agentic Development System

This is an intensive build week. You will move from a single coding agent to an orchestrator, builder and checking agent working across longer development assignments. The first version will probably fail in several ways. Your job is to improve the workflow until it can produce a workable system.

The three-agent workflow

The orchestrator assigns a task; the builder implements it; the checker accepts or rejects it. Rejected work returns to the orchestrator for correction. Accepted work is recorded as progress before the next iteration.

Use a locked technology stack

Do not spend the week comparing frameworks or allowing agents to suggest replacements. Professional engineers frequently work within an organisation’s existing architecture.

Choose one workable system

Define a workable first version.

Let AI coach the process

Do not run this week alone. Ask Claude or Codex to guide you through the process itself: how to plan the week, how to divide the work and what a productive day should look like. At the end of each day, ask the AI for feedback on your engagement and for suggestions to improve; incorporate the ones that make you more productive. Ask it to critique your work—code, decisions and workflow—and to provide insights for better development, better standards and a better product. You are not merely building with AI; you are being coached by it.

Talk to at least five potential users

During Week 3, show the idea or early application to at least five people outside your normal circle who could plausibly use or understand it. Ask what problem they think it solves, what is confusing, which feature matters, what is missing and whether they would actually try it.

Listen rather than defend. Record patterns, decide which feedback to accept or reject, and explain why. Product judgement means valuing feedback without turning every suggestion into a feature.

Week 4

Turn the Application into a Product

Working software is not yet a finished product. Week 4 is about securing it, deploying it under a proper identity and communicating why it matters.

Secure and solidify the platform

Strengthen Clerk login and session security, separate administrator, manager and attendee permissions, and implement Supabase Row Level Security. Protect environment variables, validate inputs and uploads, enforce server-side authorisation, handle errors, and add safeguards for AI requests.

Do not assume that hiding an administrator button secures an operation. Enforce permissions in the API and database. Use the checking agent to test multiple roles and deliberately attempt unauthorised access. Security must be demonstrated, not claimed.

Give the product a live identity

Deploy the production version on Vercel and confirm that authentication, environment configuration and the production Supabase database work correctly. If you can spend a few dollars, purchase a suitable .in domain and connect it to Vercel. The product will then have a proper public identity and a live HTTPS URL.

Build a one-page product website

Create a separate one-page website containing:

• ✓  A clear positioning statement: what the product does, who it is for and why it matters.
• ✓  Customer benefits: the outcomes created, not merely the technologies used.
• ✓  Product features, each connected to a benefit.
• ✓  Screenshots, a short demonstration or a link to try the product.
• ✓  A call to action inviting visitors to try it, request access or contact you.

You may use design inspiration and components from 21st.dev, while ensuring that the final site communicates your own product identity.

Prepare and deliver the product presentation

Create a concise presentation covering the problem, customer, positioning, solution, demonstration, benefits, features, technology, agentic development method, security, user feedback and next steps. Present it to classmates, teachers, developers, friends or potential users.

Show Your WorkBuild in Public from Week 1

Do not wait for the product to be complete before discussing the journey. At the end of every week, share genuine progress on LinkedIn and Instagram:

• ✓  Week 1: what you learned about AI and prompt engineering.
• ✓  Week 2: the agentic environment and autonomous exercise.
• ✓  Week 3: the product, multi-agent workflow and user feedback.
• ✓  Week 4: the secured, deployed product, website and presentation.

Show screenshots, diagrams, short demonstrations and lessons learned. Include difficulties and corrections. Do not fill the posts with AI terminology that you cannot defend, and never expose credentials, private user data or security details.

After the MatchThe Four Weeks Are Only the Beginning

Continue developing the product. Ask friends, classmates and real users to adopt it. Observe what they do, collect feedback and improve the experience.

• ✓  Strengthen coding harnesses, regression tests, security checks and AI evaluations.
• ✓  Improve reliability, performance and usability.
• ✓  Expand features based on evidence rather than excitement.
• ✓  Add analytics, monitoring and eventually payments.
• ✓  Develop a sustainable product and bring in more users.

Every week, continue updating the product, LinkedIn and Instagram, the white paper, the presentation and the résumé. Update the résumé with outcomes you can demonstrate—users acquired, features shipped, failures corrected, security strengthened and improvements measured—not merely with additional AI tool names.

Game, SetThe 360-Degree Intern

At the end of this journey, you will have practised AI and agentic development, product thinking, user conversations, engineering judgement, security, testing, deployment, positioning, feature articulation, website communication, presentation and public professional communication.

You are no longer asking a company to give you an opportunity so that you can begin learning. You are showing that you can understand a problem, speak to users, build a solution, secure it, deploy it and explain why it matters.