How I Use Kagi

Original Article ↗ Hacker News Discussion ↗

Yandex Partnership & Geopolitics

  • Major thread focus is Kagi’s continued use of Yandex as a data source and paying them for API access.
  • Some users cancelled subscriptions or won’t subscribe because they do not want any of their money flowing into the Russian economy during the invasion of Ukraine; Yandex is described as deeply entangled with the Russian state and propaganda.
  • Others argue this is disproportionate or impractical given that many Western companies and governments also fund or enable wars; accusations of “whataboutism” fly in both directions.
  • A recurring split:
    • One camp prioritizes moral boycotts, even if inconsistent or small in impact (“I draw a line somewhere”).
    • Another camp stresses moral complexity, inevitable complicity, and says a good, small, privacy‑respecting search engine may do more net good than harm.
  • Kagi’s founder defends the Yandex integration as ~2% of costs and justified purely on search quality, not politics; argues that once geopolitics drives inclusion/exclusion of sources, it stops being a neutral search engine.
  • Some users find this principled and “refreshing”; others see it as technocratic, evasive, or outright disqualifying, and request at least a per‑user “no Yandex” toggle.

Search Quality, Features & Comparisons

  • Many subscribers report Kagi consistently matching or beating Google on difficult, specific, or technical queries, especially when Google’s “fuzzy intent” and SEO spam get in the way.
  • Kagi’s killer features for fans:
    • Per‑user block/boost of domains and “lenses” (custom source filters like “Academic”), which can also constrain its AI assistant.
    • Ability to globally nuke sites like Pinterest, tabloids, or obvious AI‑spam; some say going back to Google + uBlock feels awful after getting used to this.
  • Some users see little or no difference versus Google or DDG once ad‑blocking is enabled, and don’t find a subscription justified.
  • Performance is occasionally cited as slower than Google; Kagi staff engage and offer to investigate.

AI, LLMs & Changing Search Habits

  • Several users now lean on LLMs (ChatGPT, Claude, Perplexity) for many queries and feel that “search is being squeezed out.”
  • Others highlight Kagi’s integrated AI features (? quick answer, !ai to send to Kagi Assistant with multiple models) as a strong hybrid approach.
  • Some cancel Kagi because their primary “search” is now an LLM; others subscribe to both Kagi and AI tools and see Kagi’s higher‑quality web retrieval as a good substrate for AI.

Blocklists, Content & UX Gaps

  • The article’s shared blocklist is controversial: some see it as over‑broad (blocking large news sites, social networks, Amazon, etc.), mixing “SEO spam” with sites the author personally dislikes.
  • Users advise treating such lists as starting points, not gospel, and emphasize downranking over hard blocks.
  • Kagi’s weak spots repeatedly mentioned: maps and local business/restaurant search, where Google Maps is still preferred.
  • Other frictions: required sign‑in for any search, lack of easy prepaid/anonymous options beyond a constrained Privacy Pass, and Yandex being a hard blocker for some otherwise enthusiastic users.