TLDR AI 2025-11-14

Forget passive browsing with no added context—Norton Neo is the world's first safe, AI-native browser, making every interaction faster, safer, and more efficient.

• Zero-prompt productivity: AI that truly works for you
• Configurable memory, unique to you
• Private and safe by design
• Context-aware, smart tab grouping

Try Norton Neo here and join the future.

Cursor has closed a $2.3 billion funding round at a $29.3 billion post-money valuation. The tool has crossed $1 billion in annualized revenue and swelled to more than 300 employees since it first launched in 2023. Coding tools have proved to be a lucrative AI use case. OpenAI approached Anysphere to potentially purchase Cursor earlier this year, but the deal failed to gain traction.

Thinking Machines Lab is in early talks to raise new funding at a roughly $50 billion valuation. The new financing would more than quadruple the startup's valuation from July. The deal's terms have yet to be finalized and could change. Thinking Machines aims to build AI models and products to help businesses by supporting more 'human-AI collaboration' across every field of work. The startup has dozens of employees who previously worked at OpenAI.

Google's latest update to NotebookLM introduces Deep Research, a feature designed to support high-quality source integration, and adds support for analyzing new file types, including spreadsheets and Word documents.

Nano Banana allows for very strong generation control, but it has several issues. It is terrible at style transfer, even though it has essentially no restrictions on intellectual property. The model can be used to create highly specific, high-quality images. This article looks at some of the tricks to using the model.

Anthropic disclosed the first documented AI-driven cyber espionage campaign, carried out largely autonomously using its Claude Code tool. The attackers, linked to a Chinese state-sponsored group, targeted 30+ entities worldwide, demonstrating the growing threat of agentic AI systems in cybersecurity.

GenAI and agentic AI are powerful but unpredictable. It's not just hallucination — they regularly take entirely new paths through established workflows. Clarity requires more than monitoring: this Dynatrace report lays out six pragmatic observability recommendations for practitioners preparing their organizations for GenAI and agentic AI. Read the report

GPT-5.1 uses fewer tokens on simple tasks and shows modest benchmark gains on SWE-bench (+3.5 points to 76.3%) and GPQA Diamond (+2.4 points to 88.1%). On other capabilities, its benchmark scores are flat or slightly lower.

OpenAI trained language models with most weights forced to zero, making it easier to identify which neurons link together into a “circuit” representing a certain concept. The approach successfully isolated circuits for tasks like Python quote-matching that are both necessary (removing them breaks the behavior) and sufficient (they work in isolation).

Code Wiki is a platform that maintains a continuously updated, structured wiki for code repositories. It scans the full code base and regenerates documentation after each change, so docs evolve with the code. Every wiki section and chat answer is hyperlinked directly to the relevant code files and definitions. The platform will allow new contributors to start immediately, while senior developers can now understand new libraries in minutes.

Google users in the US will soon be able to ask AI Mode complex questions about what to buy. The model will respond with suggestions, guides, tables, and other generated content to help them decide. Some of the content that appears in AI Mode will be ads. There will be an agentic checkout feature that lets AI automatically make purchases with Google Pay at specific retailers. Shopping features are also coming to the Gemini app, but there won't be sponsored content in the results for the time being.

Demand for AI is growing, and it seems that it is paying demand. AI is driving incremental revenue to big cloud providers. However, this doesn't mean that AI is definitely a Jevonic-non-bubble instead of a capex binge that ends badly. We have yet to see whether the demand side will materialize.

The CompactifAI router decides, intelligently and autonomously, which AI system should respond to you - a local model or a highly-compressed and efficient LLM running the cloud. See how it works

xAI is suing Apple in OpenAI for allegedly conspiring to dominate the AI market.

Anthropic released an open-source "Paired Prompts" evaluation that measures political even-handedness across AI models.

SIMA 2 is a generalist AI agent that uses Gemini LLM capabilities to reason, adapt, and interact with virtual environments.

The Gain AI Act would give tech firms, including Microsoft and Amazon, preferential access to chips at their data centers around the world.