Bots now outnumber human traffic on the internet, Cloudflare CEO Matthew Prince reveals; one AI agent visiting 5,000 sites for a single task signals the old web business model is breaking.

Cloudflare's AI Gateway processes 241 billion tokens daily, adding auditing and guardrails; edge inference on Workers is positioned as key to agent compute.

OpenAI and Broadcom unveil Jalapeño, a custom LLM inference chip designed to improve performance, efficiency, and scale across AI systems.

The chip signals OpenAI's move into custom silicon to reduce reliance on external GPU suppliers for inference workloads.

SpaceX completes IPO at $1.75T valuation, then acquires AI coding startup Cursor for $60B, positioning xAI with Cursor's talent, data, and code generation product.

The deal signals that coding agents are the most valuable AI application right now, with Cursor's huge user base and data likely strategic assets.

ChatGPT's market share drops below 50% for the first time, as Gemini and Claude gain ground, according to TechCrunch via Last Week in AI.

The shift indicates intensifying competition in chatbot usage, eroding OpenAI's once-dominant market position.

Dan Biderman and Jessy Lin (Engram) bake company knowledge directly into model weights, achieving frontier-level performance with up to 100x fewer tokens than prompt-based approaches.

Working with Microsoft, Notion, and Harvey, Engram envisions every user having a privately trained, continually learning model—the 'ChatGPT moment for memory'.

Fields medalist Timothy Gowers expected AI would prove a theorem, signaling 'all over for mathematicians very soon'—but it only found a counterexample, buying a year or two.

Surge AI CEO Edwin Chen, whose company trains frontier models, argues scaling laws hold and nothing humans can do AI won't do better; benchmarks like Riemann-bench measure progress.

Emil Lassen from AIUC introduces the AIUC-1 framework: standards, certification, audit, and insurance layered on AI agents to accelerate enterprise adoption.

Red teaming based on standards is pitched as key to building trust before the 'AI hailstorm' of agentic systems hits enterprises.