Code Assignment

# 🧩 Code Assignment **Turn a rough hiring idea into a polished, self-checking coding exercise — in minutes.** Writing a good take-home or interview coding problem is slow work. You have to invent something interesting, write a clear problem statement, build starter files, and then prove that the whole thing actually works before you put it in front of a candidate. Code Assignment does the heavy lifting for you. Describe what you want, review a short list of candidate ideas, pick your favorite, and let the app build and test the complete package for you. --- ## What you get - **Fresh ideas on demand** — a batch of candidate problems tailored to your language, topic, and difficulty level. - **Complete problem packages** — a written problem statement (README), starter files, and an automatic checker that can tell whether a submitted solution is correct. - **Proof that it works** — every problem can be validated automatically, so you never hand out a broken or ambiguous exercise. - **A searchable library** — everything you create is kept in one place, filterable by language, topic, difficulty, tags, and review status. - **One-click downloads** — grab a whole batch or a single problem as a zip file. --- ## Getting started ### Step 1 · Draft ideas Open the **1 · Draft ideas** tab. 1. Give your batch a short name (for example `2025-q1-python-graphs`). This just keeps your work organized. 2. Fill in the brief: - **Language (s)** — e.g. `python`, or `any`. - **Domain / topic** — e.g. `graphs`, `string processing`, `concurrency`. - **Difficulty target** — `easy`, `medium`, `hard`, or something descriptive like *"45-minute interview screen"*. - **Constraints** — anything that must be true, such as "no external packages" or "must run offline". - **Reference context** *(optional)* — paste a job posting or a list of required skills, and the ideas will lean toward that role. - **How many ideas** — leave blank for a handful (around 5–8). 3. Click **Draft ideas →**. In a moment you'll see a set of idea cards, each with a title, a one-line summary, the skills it tests, why it's interesting, and where candidates might get confused. > 💡 Tip: You can come back to any earlier batch from the list on the left, tweak the > brief, and draft again. ### Step 2 · Pick an idea and expand it Found one you like? On that idea's card, adjust the suggested problem id if you wish and click **Use this idea →**. The app writes the full problem package — statement, starter scaffold, and automatic checker — and then takes you straight to the Library. ### Step 3 · Review and validate In the **2 · Library** tab you can browse everything you've created. - Use the filters at the top (language, domain, difficulty, status, tags, or a text search) to narrow the list. - Click any card to open it. You'll see the problem details, the full README as a candidate would read it, the latest test results, and any written feedback. - Click **▶ Validate this problem** to have the app solve and test the problem end-to-end. When it finishes you'll see a pass/fail summary and per-test details. - Click **Generate report** for a written critique of the problem — clarity, difficulty, and anything worth fixing. - Click **Download problem (zip) ↓** when you're ready to share it. --- ## Understanding the status badges | Badge | What it means | |------------------|----------------------------------------------------------------------| | **unexpanded** | An idea has been chosen, but the full package hasn't been built yet. | | **draft** | The package exists but hasn't been tested. | | **needs-review** | It was tested and something didn't pass — worth a look. | | **validated** | Tested and passing. Ready to use. | --- ## Handy details - **Models**: The two dropdowns at the top let you choose which AI models to use — a "smart" one for the heavier writing work and a "fast" one for lighter tasks. Your choices are remembered. - **Activity log**: At the bottom of the page you can watch what's happening in real time, including links to monitor a running task live. - **Longer runs**: Drafting and validation can take a little while. That's normal — the app is genuinely writing and running code. Just leave the tab open. - **Nothing is lost**: Batches live under `drafts/` and finished problems under `problems/`, so you can always come back later or share the folders directly. --- ## A typical session > *"I need a medium-difficulty Python exercise about graphs for a backend screen."* 1. Name the batch `backend-screen-graphs`, fill in the brief, click **Draft ideas**. 2. Skim six ideas, pick the one about shortest paths in a road network. 3. Click **Use this idea** — the full package is written for you. 4. Click **Validate this problem** — all tests pass. ✅ 5. Click **Generate report**, read the notes, and download the zip. Total effort: a few minutes of reading and three clicks. --- **Ready?** Head to the **Draft ideas** tab and describe the problem you wish you had.