Summary
Highlights
The speaker argues that Google is destroying its core product, search, by integrating AI. Many alternative search engines exist, such as DuckDuckGo, Quant, Startpage, Ecosia, and Bing. The speaker uses a self-hosted search engine that provides clean, ad-free results without AI overviews, contrasting it with the current state of Google. This shift echoes a broader trend where major tech companies are 'shitifying' their products, pushing users to open-source alternatives like Linux for gaming due to their less bloated and more user-friendly experience.
Google is incorporating AI search features, particularly relying on it for a significant portion of its income. This AI search often generates 'AI overviews' that take up significant space in search results. These overviews frequently extract information from websites, including older fan sites or Wikipedia, without driving traffic to the original sources. This practice stifles the revenue and traffic of content creators who depend on those visits, effectively killing the internet economy the AI relies on for its data. The speaker exemplifies this by showing how Google's AI might pull a year-old Reddit post for 'best Linux distro' rather than actual, up-to-date information, making Google's AI search akin to simply looking up Reddit answers.
Google claims over a billion monthly users for its AI mode, which the speaker finds surprising given the perceived quality. Google is integrating its Gemini Flash model directly into the search bar, promising fast analysis for complex queries like 'available pottery classes near me on Tuesday nights.' However, this leads to an 'AI overview' that again pulls information from third-party sites, potentially providing inaccurate results and explicitly stating 'for informational purposes only.' The introduction of 'search agents' aims to allow users to create and manage multiple AI agents, further emphasizing Google's AI-centric future.
The speaker demonstrates a local AI search agent using Seir XNG, which can report JSON files of search results to a local AI model (Quen 3.6 6 35B A3B). This local setup shows how AI can execute searches, pull results, and combine information, similar to Google's process but on a local machine. The speaker argues that Google's approach is 'f***ing ass f***ing the entire internet' by stifling traffic with hallucinated AI results. Google's goal is to replace traditional internet search with an AI chatbot, which the speaker believes is problematic because users still need to verify AI-generated information, defeating the purpose. This move is seen as Google destroying its own key product.
Google's AI push is detrimental, increasing bot traffic and decreasing human engagement with the internet. Similar to how YouTube videos are summarized by AI, stifling content creators, Google search AI sometimes hallucinates information. Even features like 'agentic coding' are presented as innovative but are ultimately another avenue for companies to inflate user numbers. The speaker suggests that Google reports high AI usage by counting any interaction with their AI tools on Android devices or Google-related products. To combat this, the speaker advocates for de-Googling by switching to alternative browsers like Brave, Helium, Firefox, or Waterfox, and using self-hosted search engines like Seir XNG (easily deployable with Docker). For operating systems, alternatives like CashioOS for gamers or Fedora Linux are recommended to avoid AI bloat and privacy concerns. The speaker concludes by questioning why users continue to use Google search given its increasingly poor results, constant sponsorships, and potential security risks.