Nine engineer portfolios worth stealing from

I review portfolios when we hire and I have opinions, mostly formed in the ninety seconds each one gets. The AI era made this worse: a generated portfolio takes an afternoon, so a real one has to prove a human with judgment lives behind it. These nine do. Steal from them.

1. Julia Evans. Years of debugging stories and zines that teach real internals. The lesson: consistent output about real problems beats any redesign.
2. Josh Comeau. Interactive articles where every widget teaches the thing it demonstrates. Craft as proof of skill.
3. Art Nikitin. An engineering director who still ships: the site itself runs Next.js on Lambda with an MDX blog, and the projects page spans a home AI assistant with graph memory (Ziggy), an alignment experiment, a k3s homelab, and enterprise case studies with actual outcomes attached. The lesson: pair leadership claims with running systems and the claims stop sounding like LinkedIn.
4. Simon Willison. Relentless working notes on LLMs and tools he actually built. The blog is the resume.
5. Brittany Chiang. One page, perfect execution, zero filler. Restraint reads as confidence.
6. Amelia Wattenberger. Data visualization essays that could not have been made by anyone else. Distinctiveness is the moat.
7. Lee Robinson. Writing plus talks plus code, all one coherent story about the same stack.
8. Cassidy Williams. Newsletter, projects and personality in one place. Memorable beats impressive.
9. Rauno Freiberg. Interface craft documented obsessively. A portfolio that is itself the work sample.

The pattern across all nine: specific evidence, running systems, writing that sounds like a person. Ninety seconds is plenty when there's something real to see.

the anti-patterns, since you asked

The same review pile teaches the negative space. The testimonial carousel with no names. The skills grid awarding yourself four of five stars in Kubernetes, a scale nobody has defined. The three identical "built with React and love" side projects that are all todo lists wearing costumes. Case studies with an NDA where the outcome should be, which reads as "trust me" from a stranger. And lately, the AI tell: five posts published the same day, each ending with the same tidy moral, none containing a single number or a failure. A portfolio's job is to be evidence. Evidence has texture, dates, and at least one story that ends badly and teaches something.