Week 16 of 16

Capstone: Build Day 2

Complete the build, handle edge cases, and wire everything together.

Day 79 90+ minutes Build

Day 79 of 80

Where You Should Be

By the end of yesterday (Day 78), you had the core working — the main thing your tool does. Today you complete it:

Handle the edge cases you deferred
Make error messages useful, not just "something went wrong"
Connect any pieces that are still separate
Get it to a state where someone else could use it

Don't add new features. Complete what you committed to in the plan.

The Completion Checklist

Before you write any new code today, read through your project plan from Day 77. For each thing you planned:

Does the core functionality work end-to-end?
Does it fail gracefully when given bad input?
Are error messages human-readable (not raw Python tracebacks)?
Does it do what your plan said it would do — and only that?

Common Things Left for Day 2

Error handling you probably skipped

Yesterday you focused on the happy path — what happens when everything works. Today add the unhappy paths:

What if the input file doesn't exist?
What if the API key is missing?
What if Claude returns an empty response?
What if the database is corrupted?

You don't need to handle every possible failure — just the ones a real user would plausibly hit. Wrap the most likely failure points in try/except and print a useful message.

The "what does this do?" problem

Run your tool fresh — as if you've never seen it. Does it tell you what to do? Does --help exist? Are command names obvious? A tool that requires reading source code to use is half-finished.

If you're using Typer, you already have --help for free. Make sure every command has a docstring and every argument has a help= string.

Connecting the pieces

You may have some things working in isolation that aren't connected yet. Day 2 is for those connections. Look at your plan — what was supposed to call what? Are all the arrows actually wired?

A common pattern: you built the data layer and the CLI separately. They both work. But the CLI calls input() instead of the data layer. Today you replace those input() calls with the real functions.

What "Done" Looks Like

You're done when:

You can demo the tool without looking at the code
Running the main command with no arguments shows help or a clear prompt — not an error
The failure cases you anticipated are handled with clear messages
The tool does exactly what the plan said — and nothing extra

If you finished early

Don't add features. Instead:

Read your code out loud — does every function name and variable name explain itself?
Delete any commented-out code, dead code, or debugging prints you left behind
Add logging to the operations that might fail silently
Write the README for Day 80 — what would you tell someone trying to use this for the first time?

If you're behind

Cut scope. Not quality — scope. Pick the two or three things that are most important to your use case and make those work well. A tool that does one thing perfectly is better than a tool that does five things badly.

Update your plan document to reflect what you actually built. The plan isn't a contract — it's a guide. Adjusting it is professional, not a failure.

End of Day Checklist

Core functionality works end-to-end
Error cases are handled with readable messages
Help text is present and useful
No dead code or debugging artifacts
You could demo this to someone unfamiliar with the code

Tomorrow — The Final Day

Day 80 is testing, documentation, and reflection. You'll write tests, write a README, and look back at how far you've come since Week 1.