Ship the Ugly Personal Site

By Rareș Gosman

I built an elaborate AI portfolio for a student I mentor. He ignored it, made his own plain site in Framer, and proved the point better than I could.

CareerMentorshipAI Agents

One thing I regret from university is never putting up a personal website.

This is especially funny because I went to Waterloo, where students are famous for elaborate portfolios. That was partly the problem. The bar around me felt so high that I thought my site needed to compete with the designs built by upper-years with design internships at Facebook.

So I waited until I could make something impressive.

Naturally, I made nothing.

That was the wrong comparison: my nonexistent first version against someone else's tenth. An ugly site would have been better. Once something is public, you have a foundation to improve. When you keep building privately, you often lose interest, or you keep raising the self-imposed quality bar until nothing sees daylight.

A personal site does not magically create an audience. You still need to meet people, publish work, and share the link. But it gives people who find you somewhere to land.

A resume and LinkedIn profile compress you into standardized fields. A site can show the projects behind the bullets, the decisions you made, what went wrong, what you learned, and what you are still learning.

A post by Nate B. Jones about "owning your interface layer" drove this home for me. Instead of presenting another document for someone to filter, give them something they can investigate.

Building one for a student I mentor

Last December, I spoke with a younger engineering student I had been mentoring. Afterward, he sent me a thoughtful email and his resume.

There was plenty of real material in it: high-voltage electronics for an electric race car, custom PCBs, a life-size Dune thumper, and a robot dog built during a hackathon. But, like most student resumes, years of work had been flattened into bullets that a recruiter might skim for a few seconds.

I reviewed and tightened the resume, but I also wanted to give that work more room to breathe. So I built him a site and bought the domain for three years. It had a recruiter-facing chat for questions about his projects, and a job-fit evaluator that would tell a recruiter "not a fit" when the match was not there. The goal was honesty, not flattery. AI cannot invent projects someone did not do; it can only make the real ones easier to find.

In a mildly embarrassing parallel to my university experience, my email explaining all of this sat in drafts for months. I eventually sent it. Better late than never applies to both mentorship emails and personal websites.

I generalized the site into an open-source AI Portfolio Template so anyone could use the same setup.

The template in action

The template's demo persona, live: ask the recruiter chat about a project, then paste a job description and watch it name the gaps instead of flattering.

What he actually did with it

He did not use it.

A month after I sent the email, he replied. He appreciated the site and the resume feedback, but he wanted more control over the presentation. So he built his own portfolio in Framer, wrote up his strongest electrical project, and published it. He asked me to release the domain so he could point it at his site.

That was better than anything I had built for him. He did not need the recruiter chat or the fit evaluator. He needed a nudge and a reason to start, and then he chose his own tools and shipped something he could maintain himself.

The elaborate version I built was a gift, and I am glad I sent it. But the version he built was proof: all the site needed to do was exist. His plain Framer page with one project write-up was the ugly personal site I wish I had shipped at Waterloo.

Start with four things

Your first personal site can contain four things:

  1. Who you are.
  2. What you are learning or working on.
  3. One or two projects with evidence.
  4. A way to contact you.

Put it online. Share it. Add to it whenever you finish something. Record measurements and outcomes while projects are fresh, because six months later you will forget the details that make the work credible.

Your first version may be plain. Good. The portfolio that competes with theirs can come later. First, create the address and let the public version teach you what the next version needs.

Something imperfect can compound. Something permanently private cannot.