Independent analysis of economics, technology, business, and public policy. Original essays informed by academic research, historical evidence, and data-driven reasoning.
Information Systems
A language model asked the same factual question three times gave three confident, wrong answers. The real culprit isn't broken software — it's benchmarks that reward bluffing over honesty, a flaw retrieval-augmented generation can narrow but never actually cure.
Vector-search benchmarks measure the one part every engine already optimized and ignore the thing that actually breaks production retrieval: filtering against live business data. That reframes the entire pgvector vs Qdrant vs Pinecone debate.
Precision is easy to report; recall requires counting every relevant document that exists, which nobody has done since the web outgrew ten blue links. The AI search industry hasn't solved that problem — it has stopped mentioning it.
Vector search misses exact strings; keyword search misses meaning. The reason hybrid retrieval works isn't that it's smarter — it's that it refuses to trust either retriever's confidence, and hands the decision to a score-blind referee.
The trick behind billion-scale AI retrieval isn't reading everything faster. It's learning what to ignore—and accepting a system that misses one match in twenty by design.
Semantic search didn't just make search smarter. It converted language into geometry, gaining a genuinely new ability while losing the razor-sharp precision of the old index — which is why the best systems now run both.
The trade press defines a hyperscaler by server counts and square footage. But the word's root — scale — was always about how gracefully a system grows, not how big it is. Trace the slippage and you find an industry that named its firms for the one thing they made look effortless.
Everyone frames local versus cloud AI as privacy versus convenience. The real variable deciding which one bankrupts you is how many tokens you actually push through hardware you've already paid for.
A chatbot with a million-token window can still lose the one fact you need, because attention isn't storage—it's a spotlight with a U-shaped blind spot in the middle.
Retrieval-augmented generation was sold as the fix for AI hallucination. Five years of deployment show it just moves the failure from the model to an unaudited retrieval pipeline — and a wrong answer with a footnote is more dangerous than an obvious bluff.
PostgreSQL's two supposed weaknesses—academic extensibility and the total absence of a corporate parent—are exactly why it outlasted every rival built to beat it.