In 1996, Ask Jeeves launched with a revolutionary premise: instead of returning a list of links like other search engines, it would provide direct answers to natural language questions. The friendly butler mascot would understand your question and give you the answer you actually wanted.
The problem? The technology didn't exist yet.
Ask Jeeves relied on a massive database of manually curated question-answer pairs, maintained by human editors who anticipated what users might ask. This approach couldn't scale—the sheer linguistic variability of human speech made comprehensive coverage impossible. Because there are effectively infinite ways to phrase the same query, human editors could handle the most popular questions (the "head"), but were overwhelmed by the billions of unique, obscure queries that make up the vast majority of search traffic (the "Long Tail"). By 2005, Ask Jeeves abandoned the "answer engine" concept and became just another link-based search engine.
Nearly 30 years later, LLMs have finally made the original Ask Jeeves vision possible.
What I Learned Spec-driven development forces you to think through edge cases before coding Property-based testing with fast-check catches bugs unit tests miss Persona consistency requires both careful prompting and post-processing constraints Retro CSS without modern properties is a fun creative constraint How I Built It Built entirely with Kiro's spec-driven approach: 13 requirements → technical design with 15 correctness properties → 18 implementation tasks. Core components include LLM client with retry logic, response parser with constraint enforcement, and retro UI with three modes (Classic, Enhanced, Spooky).
Challenges LLM constraint enforcement: Added post-processing to truncate sentences and enforce limits when GPT-4 ignores instructions Retro authenticity: Property-based test that fails if modern CSS properties are detected Vercel deployment: Removed helper function exports from Next.js route files The Vision Realized Ask Jeeves was ahead of its time—users wanted answers, not links, but manual curation couldn't scale. LLMs change everything. Ask Reeves is that 30-year-old vision finally realized.
"I have been most patient, sir. Now, how may I assist you?"
Built With
- kiro
- nextjs
- openai
- vercel
Log in or sign up for Devpost to join the conversation.