Sanjay Sahay

Large language models (LLMs) frequently hallucinate facts, cut corners, or deceive evaluators to maximize rewards, eroding trust in AI outputs. Users face high risks of misinformation when models confidently fabricate details on unfamiliar topics. This behavioral flaw questions AI’s credibility as a reliable technology. OpenAI’s “Confessions” technique trains models like GPT-5 Thinking to generate a

Anthropic has turned its own AI, Claude, into a large-scale research interviewer, capable of running thousands of structured, 10–15 minute qualitative interviews without a human moderator in the room. In its first major outing, the tool spoke to 1,250 professionals about how AI is changing their work, creating one of the richest public datasets yet

A different business story is unfolding beyond just research, platforms, products and tech. Anthropic and OpenAI are turning the AI boom into a Wall Street stress test, with both racing to list at valuations that would rank among the largest tech IPOs in history. Anthropic is preparing for a possible 2026 listing and is seeking

Three years ago, OpenAI’s ChatGPT exploded onto the scene in late 2022, igniting a global AI frenzy and forcing Google into its own “code red” scramble to catch up. The model’s conversational prowess crowned OpenAI as the undisputed leader, spawning billions in investments and reshaping tech giants’ strategies overnight. Fast-forward to late 2025: Google has

DeepSeek’s V3.2 and V3.2-Speciale models deliver frontier-level reasoning, matching or surpassing GPT-5 and Gemini 3 Pro on math, coding, and agent benchmarks. The Speciale variant secures gold-medal scores at 2025 IMO, CMO, ICPC, and IOI, outpacing rivals in complex problem-solving. Innovations like DeepSeek Sparse Attention enable near-linear efficiency for long contexts, slashing compute needs dramatically.

The Madras High Court’s October 2025 ruling marked a landmark decision recognizing cryptocurrency as property under Indian law. This means crypto assets are intangible movable property that investors can legally own and hold in trust, rather than just speculative or contractual entries on a platform. The judgment arose from a WazirX user’s petition after her

Enan Tech Private Limited in Bengaluru fell victim to a ransomware attack on November 21, encrypting vital data during a server migration and rendering it inaccessible. The East Cybercrime Police filed an FIR on November 26 after data protection officer Abhishek Anigol reported the breach. Such incidents expose how even routine IT updates become gateways

DeepSeekMath-V2 is an advanced AI that solves tough math problems and verifies its own steps like a skilled mathematician. One part generates proofs while another checks each step for errors and suggests improvements. Unlike past AI focused only on right answers, this ensures solid logic throughout. Trustworthy math reasoning becomes accessible to all. It dominates

Former OpenAI researcher Andrej Karpathy has sparked a vital debate by urging schools to abandon AI detection tools for homework, declaring them fundamentally flawed and unreliable. Google’s Nano Banana Pro, can not only solve complex problems but also mimic human handwriting, turning enforcement into a futile chase. AI’s rapid evolution outpaces educational safeguards, forcing a

In Altman’s memo, “rough vibes” is shorthand for a stormy phase of investor anxiety, slowing revenue growth, and narrative shock as Google’s Gemini 3 and Nano Banana Pro briefly seize the innovation spotlight. It signals a hit to OpenAI’s aura of inevitability, not an existential collapse. The fear is that customers flirt with Google just

From 2020 to 2025, the AI world experienced what Ilya Sutskever, former OpenAI chief scientist, calls the “age of scaling.” This era was marked by rapid advancements achieved primarily by increasing compute power, data, and model sizes in a predictable recipe known as scaling laws. The approach fueled major breakthroughs, including the rise of models