Shipping the IndexFox embed widget: iframes, postMessage, and small bundles
You drop one script tag. Behind it: an iframe, a postMessage protocol, and a deliberate choice to ship as little code into your site as possible.
Read post →The IndexFox engineering blog
Notes from building a search-and-RAG product. Hybrid retrieval, the rerankers we ship, the crawl pipeline behind the widget, and how we read the 2026 SEO landscape.
Search intent: typing vs committed — the funnel nobody measures
Treating every keystroke as a "search" is how product teams end up with 1.4M monthly searches and no idea what users actually wanted.
Read post →
Query fan-out: Google's 2026 mechanic was already inside our widget
Google publicly named the mechanic in 2026. We have been quietly doing it inside the widget since 2024 — for the same reason Google does it: one user query is rarely the right one.
Read post →
AEO and GEO are still SEO: lessons from 2026
Half the LinkedIn posts of 2025 told you AEO was a brand-new discipline. The other half were right: it is the same playbook, with one extra muscle.
Read post →
Our reranker stack: Jina, Voyage, and the quantization speedup
Reranking turns "decent results" into "the right answer." Here is how we make it cheap enough to run on every query.
Read post →
Hybrid search explained: vectors and BM25 beat both alone
Pure vector search is great until someone searches for a SKU. Pure BM25 is great until someone asks a question. The honest answer is: run both, fuse the scores.
Read post →
How we crawl: Crawlee, Readability, and segment-level reindexing
Crawling a site is the easy half. The hard half is deciding what changed since last week and not re-embedding 80,000 paragraphs that didn't.
Read post →
Why we wrote our own RAG on ManticoreSearch instead of buying one
Pinecone, Weaviate, Algolia, Vertex Search — all great. None of them let us hit sub-200ms hybrid retrieval on a $20/mo plan. So we wrote our own.
Read post →