Category: Google

  • Live Blog: Hacking Gemini Embeddings

    Live Blog: Hacking Gemini Embeddings

    —

    by

    Dan Petrovic
    in ,

    Prompted by Darwin Santos on the 22th of May and a few days later by Dan Hickley, I had no choice but to jump on this experiment, it’s just too fun to skip. Especially now that I’m aware of the Gemini embedding model. The objective is to do reproduce the claims of this research paper…

  • Google’s New URL Context Tool

    Google’s New URL Context Tool

    —

    by

    Dan Petrovic
    in ,

    Google’s just released a new system which allows Gemini to fetch text directly from a supplied page. OpenAI had this ability for a while now, but for Google, this is completely new. Previously their models were limited to the Search Grounding tool alone. Gemini now employs a combination of tools and processes with the ability…

  • How Google grounds its LLM, Gemini.

    How Google grounds its LLM, Gemini.

    —

    by

    Dan Petrovic
    in , ,

    In previous analyses (Gemini System Prompt Breakdown, Google’s Grounding Decision Process, and Hacking Gemini), we uncovered key aspects of how Google’s Gemini large language model verifies its responses through external grounding. A recent accidental exposure has provided deeper insights into Google’s internal processes, confirming and significantly expanding our earlier findings. Accidental Exposure of Gemini’s Grounding…

  • Google Lens Modes

    Google Lens Modes

    —

    by

    Dan Petrovic
    in ,

    lns_mode is a parameter that classifies Google Lens queries into text, un (unimodal), or mu (multimodal). Google Lens has quietly become one of the most advanced visual search tools in the world. Behind the scenes, it works by constructing detailed, context-rich search queries that include a growing set of parameters. One of the newest additions…

  • Chrome’s New Embedding Model: Smaller, Faster, Same Quality

    Chrome’s New Embedding Model: Smaller, Faster, Same Quality

    —

    by

    Dan Petrovic
    in ,

    TL;DR Discovery and Extraction During routine analysis of Chrome’s binary components, I discovered a new version of the embedding model in the browser’s optimization guide directory. This model is used for history clustering and semantic search. Model directory: Technical Analysis Methodology To analyze the models, I developed a multi-faceted testing approach: Key Findings 1. Architecture…

  • I think Google got it wrong with “Generate → Ground” approach.

    I think Google got it wrong with “Generate → Ground” approach.

    —

    by

    Dan Petrovic
    in ,

    Grounding Should Come Before Generation Google’s RARR (Retrofit Attribution using Research and Revision) is a clever but fragile Band‑Aid for LLM hallucinations. Today I want to zoom out and contrast that generate → ground philosophy with a retrieval‑first alternative that’s already proving more robust in production. Quick Recap: What RARR Tries to Do Great for retro‑fitting citations onto an existing model;…

  • Introducing Grounding Classifier

    Introducing Grounding Classifier

    —

    by

    Dan Petrovic
    in , ,

    Using the same tech behind AI Rank, we prompted Google’s latest Gemini 2.5 Pro model with search grounding enabled in the API request. A total of 10,000 prompts were collected and analysed to determine the grounding status of the prompt. The resulting data was then used to train a replica of Google’s internal classifier which…

  • How Google Decides When to Use Gemini Grounding for User Queries

    How Google Decides When to Use Gemini Grounding for User Queries

    —

    by

    Dan Petrovic
    in , ,

    Google’s Gemini models are designed to provide users with accurate, timely, and trustworthy responses. A key innovation in this process is grounding, the ability to enhance model responses by anchoring them to up-to-date information from Google Search. However, not every query benefits from grounding, and Google has implemented a smart mechanism to decide when to…

  • Cross-Model Circuit Analysis: Gemini vs. Gemma Comparison Framework

    Cross-Model Circuit Analysis: Gemini vs. Gemma Comparison Framework

    —

    by

    Dan Petrovic
    in , ,

    1. Introduction Understanding the similarities and differences in how different large language models represent and prioritize brand information can provide crucial insights for developing robust, transferable brand positioning strategies. This framework outlines a systematic approach for comparative circuit analysis between Google’s Gemini and Gemma model families, with the goal of identifying universal brand-relevant circuits and…

  • Neural Circuit Analysis Framework for Brand Mention Optimization

    Neural Circuit Analysis Framework for Brand Mention Optimization

    —

    by

    Dan Petrovic
    in , ,

    Leveraging Open-Weight Models for Mechanistic Brand Positioning 1. Introduction While our previous methodology treated language models as black boxes, open-weight models like Gemma 3 Instruct provide unprecedented opportunities for direct observation and manipulation of internal model mechanics. This framework extends our previous methodology by incorporating direct neural circuit analysis, allowing for precise identification and targeting…