Yes, I Used AI to Build This. Here's Exactly How.
Start with the honest part
I used AI to build this portfolio. Specifically, I used Claude Code — an agentic coding assistant that can read files, run commands, and edit a repo — steered by me over an afternoon. I’m writing this down because “I used AI” can mean anything from one lazy prompt to careful collaboration, and for a software engineering class the difference is the whole point. So here are the receipts.
What the AI actually did
I did the TechFolios setup myself: forked the template, named the repo sozodennis01.github.io, and configured GitHub Pages and Actions. Then I handed the AI my resume and links to my real GitHub repositories and put it to work. It read the assignment and all three style guides and took notes, rewrote bio.json from my resume, and drafted seven project pages straight from my actual repo READMEs — my CubeSat flight software, the PDU firmware, the Selenium login bot, the passport-photo tool, and the rest. It generated a square headshot and seven square project cards so the site meets the image requirements, removed every trace of the Molly Maluhia template, validated the JSON, and checked that no image links were broken. Fast, and genuinely useful.
What I actually did
Here’s the part that matters: I supplied every fact and made every call. The accomplishments on this site are mine — the AI organized and phrased them, it didn’t invent them. I reviewed everything, edited bio.json by hand (reworded my security clearance, set my phone to private), and when the first draft of my career essay came back I told it straight up that it was “a bit corny.” To fix that, I fed it my own writing — my Slack intro and the short-answer essays I wrote for a federal job application — so it could match how I actually talk instead of generic AI polish. I took breaks, came back with fresh eyes, and re-read. The AI moved quickly; my job was to steer it and catch what was wrong.
The actual timeline
People assume “AI did it” means instant. It wasn’t, and the breakdown is the interesting part. Here’s the real clock, all on the night of June 2, 2026 (Hawaii time):
- 8:28 PM — Opened the assignment with Claude Code. It read all three style guides and the TechFolios docs and wrote up notes and a plan.
- 8:34 PM — We settled which GitHub account to use for the long haul.
- 8:34–8:56 PM — I did the TechFolios initialization myself: used the template, named the repo, and set up GitHub Pages, Actions, and the config.
- 8:56 PM — Handed over my resume and links to my real repos. It built
bio.json, the first project pages, the images, and a first essay while I reviewed. - 9:39 PM — Had it add the rest of my projects (seven in total) and delete the template content. By then I’d already hand-edited my bio.
- 9:45 PM — Gave it my Slack intro and a federal-application essay so it could rewrite my career essay in my actual voice.
- 10:08 PM — Had it draft this essay and my take on AI.
- ~10:20 PM — Done. A little under two hours, start to finish.
The honest takeaway from that clock: the AI’s actual generating was a small slice of it. Most of those two hours was me — setting up the repo, reviewing output, hand-editing my bio, feeding it samples of my writing, and stepping away to come back with fresh eyes. The tool was fast. The steering was the work.
Why I’m telling you this
Even this essay was drafted from my notes and edited by me, and I think being upfront about that is the responsible move. AI is a tool. Used lazily, it spits out a generic site full of claims that aren’t true — exactly the thing a professor or recruiter can smell from a mile away. Used well, it’s a forcing function: it let me refresh a portfolio of real work in an afternoon while I stayed accountable for every line on the page. If you want to see the gap, compare a simple “make me a portfolio” prompt to handing an AI harness (tool) your real resume, your real repos, and your real voice — and then reviewing every word it writes. That second version is the one I did, and it’s the one worth learning.
