PinnedHarsh Chandekar·Oct 20, 2024Text to SQL Evaluation Techniques: A Comprehensive GuideText to SQL is an innovative technology that transforms natural language queries into structured SQL commands, enabling users to interact…
PinnedInTowards AIbyHarsh Chandekar·Sep 3, 2025How Hackers Hide Malicious Prompts in Images to Exploit Google Gemini AIImagine uploading a cute cat photo to Google’s Gemini AI for a quick analysis, only to have it secretly whisper instructions to steal your…
PinnedHarsh Chandekar·Jan 27Recursive Language Model (RLM) vs Large Language Model (LLM)Rather than attempting to fit large context into a single prompt, an RLM treats text as an external env managed within a programmable code…
PinnedInTowards AIbyHarsh Chandekar·Nov 30, 2025Why Google Thinks Our Entire Approach to Training LLMs Needs to ChangeA New Kind of Learning for AI: Multi-Timescale Models Explained Simply.
Harsh Chandekar·Mar 29The Full-Stack Observability Guide for Production GenAI Application1. Introduction: The Black Box Problem in GenAI
Harsh Chandekar·Feb 8Reasoning Models Will Blatantly Lie About Their ReasoningWe’ve been told that Chain-of-Thought (CoT) is the “glass box” of artificial intelligence . The logic is simple: if a model shows its work…
InTowards AIbyHarsh Chandekar·Jan 11Apple’s CLaRa: Bridging Retrieval and Generation with Continuous Latent ReasoningIntroduction: Why RAG Still Isn’t “Solved”
Harsh Chandekar·Dec 23, 2025The LLM is Now a Project Manager: Why Asynchronous Thinking is the Future of AI ReasoningWhen we talk about groundbreaking AI, we often focus on the sheer power of a single Large Language Model (LLM) — the “solo genius” that…
Harsh Chandekar·Nov 15, 2025Train for Truth: How Binary Retrieval-Augmented Reward (RAR) is Solving the LLM Hallucination…Large Language Models (LLMs) are revolutionary tools, but they often struggle with a fundamental reliability issue: hallucinations. This…
