A serious summer — even if you don't have one locked down yet.
You didn't land the brand-name internship. Or you did, and it's less interesting than you hoped. Or you'd rather spend the summer actually building something than shadowing someone who ships next quarter. Orchid was built for you.
You're probably a fit if…
You want to ship something real
Not a toy project. Not a class assignment. Something a real host actually needs, with a deadline and consequences and a name at the top of the repo that isn't yours alone.
You're willing to write
Daily journaling, weekly digests, decision notes, PR descriptions. If that sounds like a chore, this isn't your program. If it sounds like a skill worth building, keep reading.
You're curious about AI-first workflows
Claude Code sits at the center of how you'll work. Not as a cheat, not as a crutch — as a patient senior collaborator who's always available. You'll learn to use it well.
You can self-direct with scaffolding
You won't have a manager watching your screen. You will have a clear Problem Brief, a weekly rhythm, and a host you can ask. The rest is on you.
Steady rhythm, visible progress.
-
Monday — plan
Open last week's digest. Pick the next concrete piece of the project. Write a 3-line plan in your journal. Raise any ambiguity with your host before building.
-
Tuesday–Thursday — build
Session by session: restate the goal, sketch the approach, build the smallest next piece, verify it works, journal what happened. Commit atomically. Open PRs when something's ready for review.
-
Friday — reflect and digest
The weekly digest drafts itself from your commits and journal. You edit for voice and accuracy, then send it. Your host reads it over the weekend and flags anything for Monday.
-
One 30-min 1:1 per week
Agenda drafted from your journal. Used for coaching, real questions, and the things that don't belong in the sponsor update. Not a status meeting — you already sent the status.
The things that transfer.
A portfolio piece
A real project, shipped, with your name in the commit log. If the host agrees, the repo becomes public at program end.
A reference letter
Drafted from your actual journal and your host's observations — specific, not generic. Worth more than the "intern of the summer" plaque.
Habits that outlast the program
Planning, verification, decision capture, async communication, stuck-protocol discipline. These don't show up in the CS curriculum. They'll show up on your next team.
Claude Code fluency
Not "I've used Copilot." Fluency — how to scope a task for a model, how to verify its output, when to push back, how to make it your pair programmer without making it your ghost-writer.
A cohort
Other Orchid interns across projects, meeting weekly in a demo-and-feedback format. The network is often the most valuable thing you take home.
Domain knowledge
Whatever your host is working on — infrastructure, bioinformatics, fintech, climate, civic tech — you'll leave knowing it more deeply than any class ever teaches.
The commitments.
Daily journaling
Non-negotiable. Start-of-day and end-of-day entries in a private journal. It's how the digests write themselves, how the reference letter has substance, and how you'll notice your own growth.
Weekly sponsor digest
Drafted automatically; you review and send. Protecting your host's time is part of the job.
The stuck protocol
Before pinging a human, confirm: you asked Claude clearly, you tried something, you have a specific question. Paying this cost makes the help you get dramatically better.
Teach-back before merging
Before you commit code you didn't reason through yourself, you walk through what it does in your own words. If you can't, you're not ready to ship it.
Decision notes for non-trivial choices
Five bullets: context, options, choice, tradeoff, rollback. These aren't busywork — they're how you remember why you made the calls you made six weeks later.
Honesty about what you don't know
No fabrication. No smoke. When you're unsure, you say so — to Claude, to your host, to yourself. This is a learning program. Pretending defeats the point.
What "AI-first" means — and doesn't.
It means: Claude Code is open during most of your working sessions. You plan with it, debug with it, review code with it, learn with it. Your repo has a CLAUDE.md file that tells it how to work with you specifically — your level, your project, your guardrails.
It means you get to ask as many "this is probably obvious, but…" questions as you want. It means the loneliness of being a junior in a senior's codebase goes away. It means you can read an unfamiliar file, ask Claude to walk you through it, and ship a fix by lunch.
It doesn't mean: Claude writes your code and you merge it. The program enforces a teach-back rule, and your host will ask you to walk through what you shipped. Code you can't explain isn't done.
It also doesn't mean Claude is always right. Part of fluency is learning when to push back, verify, or discard what the model suggests. The best interns leave the program less impressed by AI, not more — because they've learned where it actually helps and where it confidently leads you astray.
How to apply.
We collect a calibration profile — experience, tools, learning goals, communication style — so we can match you to a project that's a real fit, not just an available one.
-
Email us to request access
Your application lives in a private repo. Email apply@orchid-initiative.com from your school address with a one-paragraph intro. We'll add you as a collaborator within a couple of days.
-
Fill out the Intern Profile
Once you have access, open a new issue in the applications repo using the Intern Profile form. About 15 minutes. Be honest — the match is better when we know where you actually are.
-
We review and suggest matches
We'll read your profile against open project briefs and suggest 1–3 potential matches. You can also browse and indicate interest in specific projects yourself.
-
Intro call with a host
A 30-minute conversation with a potential host. If both sides want to proceed, we seed your pair repo and kick off Week 1.