Week 1 first, then week 2. Same four prompts every time.
Week 1 · 9/1 to 9/6
logs/wk01.md
What I worked on
Mostly just getting the boring setup out of the way. GitHub, clone, commit, push, Pages. I already had roycemy.github.io live so the "first deployed site" part felt like something I'd already done by vibe coding a Next site. Thursday we actually walked through what happens when you hit enter on a URL (DNS, request, status codes) and poked around in DevTools. Then we made the oim3690 course repo from my laptop and shoved it up to GitHub, which is the opposite direction of Tuesday.
What gave me trouble
I kept treating this like one project. It's not. roycemy.github.io is my personal site. oim3690 is the class dump for exercises and these notes. By Thursday night the course repo was basically empty. No real index.html, no logs/wk01.md. I just assumed if one GitHub thing was working, the other one was too.
Something new I picked up
A website is just a folder of files a server hands you. That's it. GitHub is the server for this class. If there's no extra path in the URL it looks for index.html. Also opened DevTools on babson.edu and the Network tab was unhinged. Like a hundred plus files, and if you have an ad blocker on the number changes. I always thought a "page" was one page.
Working with AI this week
I'm used to prompting Cursor, getting a site, and moving on. Class was talking about agents vs chatbots and the part that actually stuck was: how far does it go before it comes back to you. This week I still used it like ChatGPT. Ask for a page. Get the page. Stop. Didn't make it hit an error and keep going.
Week 2 · 9/8 to 9/13
logs/wk02.md
What I worked on
Tuesday we did Markdown, turned on Pages for oim3690, and the main thing was: have AI spit out a plain index.html, then write down at least five things in it you don't get. That's the opposite of how I usually vibe code (prompt until it looks good, don't read it). Thursday was git stuff, .gitignore / secret.txt, going through the HTML head line by line, and having the agent add a nav, interests, and footer to hello.html. I also dumped the AI index.html into Cursor and asked it to explain the weird lines like I was dumb on purpose.
What gave me trouble
I gitignored the whole logs/ folder at first, which is actually the one folder they grade. Would've been so embarrassing. Also .gitignore does nothing if you already committed the file, which I did not know. I still mess up save vs commit vs push. I'll think I shipped something and it's just sitting on my laptop.
Something new I picked up
I actually sat with the AI index.html and hello.html instead of just refreshing the preview.
<!DOCTYPE html> doesn't show up on the page. It's just telling the browser this is modern HTML.<head> is backstage. <body> is what people see.charset="UTF-8" is why text doesn't turn into random garbage. Would've never guessed that.- The viewport meta thing is for phones. Without it, your phone pretends the site is a tiny desktop and zooms it out. That's why a lot of vibe coded pages look cooked on mobile.
href = where the link goes. src = where the image is. The words you actually click are between the <a> tags.<img> and <meta> don't get a closing tag. <ul> is the list, <li> is one bullet.<nav>, <section>, <footer> are just labeled boxes. I swapped the fake interests for mine so it didn't still say whatever the agent made up.
Working with AI this week
The Tuesday assignment was basically "stop vibe coding for a second." Find the line you don't understand and ask about that line, not "make it nicer." I did that instead of restyling it five times. Thursday we had to add tags I hadn't seen, then change one thing myself so I wasn't just accepting the dump. The first index.html it gave me worked. That's the trap. Working is not the same as I could explain it on a walkthrough.
