This site: a personal website with a built-in AI twin
A website that answers interviewers' questions on my behalf — the website itself is the proof of AI engineering ability.
Problem
Everyone writes "proficient with AI" on their résumé, but an interviewer can't verify it. How do you turn "knows how to use AI" from an empty claim into evidence you can experience live, watch unfold, and inspect for design trade-offs?
Approach
Done end-to-end in collaboration with Claude Code: first a multi-agent deep-research workflow (5 search angles, 22 sources, 12 cross-validated conclusions) to map how this is done across the Chinese- and English-language worlds, then the architecture design, then the implementation. Core designs: content/ as a single source of truth driving both page rendering and the AI system prompt; real AI and demo mode sharing one streaming protocol with automatic graceful degradation; the anti-hallucination boundary hard-coded into the system prompt — for anything outside the material, it plainly says it doesn't know.
Results
1 day (done on 2026-06-11)
From research to live
Just 1 (the chat box)
Client-side JS islands
0 (no database / vector store)
External infrastructure
AI's role in this project
AI (Claude Code) handled the market research, solution design, and all of the coding and verification; I owned the requirements, the key decisions (role targeting / language / how AI is wired in / deployment), and the content sign-off. The real prompt at every step is published in the build log.
Why I built this
When you're job hunting, "knows how to use AI" is already a buzzword on every résumé — saying it carries no weight. This website turns the claim into evidence:
- An interviewer can talk to the AI twin live, and experience how it answers from real material and how it refuses to make things up;
- They can see the design — the system prompt, the fallback chain, and every technical trade-off are published on the "How the twin works" page (source available on request);
- They can see the process — the build log holds the real prompts, the AI's missteps and corrections, and how long each stage took.
Three design decisions worth a mention
1. v1 deliberately skips RAG. The résumé corpus is only a few KB; putting all of it into the system prompt is more accurate than vector retrieval (no recall misses), cheaper (a fixed prefix hits the provider's prefix cache), and needs zero infrastructure. When RAG would actually be worth it is written up in the roadmap on the architecture page.
2. A single source of truth. The material shown on the page and the material the AI reads come from the same set of files, and the prompt shown on the architecture page is the exact prompt running in production — the output of one function, leaving no room to fake it.
3. A graceful degradation chain. Real AI → demo mode → static pages. With no key, an expired key, or an upstream outage, the site stays fully usable — and every AI reply honestly labels whether it came from a real model or a preset Q&A.
Verification
A four-scenario matrix: no key (demo mode), real key (streaming + anti-hallucination spot-checks), bad key (seamless degradation, no 5xx), and rapid-fire requests (rate-limited 429). See the build log for details.