SkeenaOneData-driven decisions, made simple

AI briefing ·

Google launches Gemini 4 Neon as Anthropic prices its IPO at $83 a share

Short answer

Google released Gemini 4 Neon, a new mid-tier model, and Anthropic priced its stock market listing above its target range. Microsoft launched its own coding models, researchers used AI to find a serious flaw in Java, and California published proposals for mandatory AI safety standards.

Models

Google launches Gemini 4 Neon for Gemini app subscribers and developers

Google announced Gemini 4 Neon on November 18 and began rolling it out to Google AI Pro and Ultra subscribers in the Gemini app, where it replaces Gemini 4 Argon for AI Pro users. In the Gemini 4 family it sits between the larger Carbon model and the faster Hydrogen. Developers can use it at $3 per million input tokens and $15 per million output tokens. Google says Neon comes close to Carbon's quality at lower cost and higher speed, and that it outperforms Claude Opus 5.5 and GPT-6.2 Sol on coding, reasoning and science tests; these are Google's own results.

Why it matters: Mid-priced models are what most businesses actually use, so a stronger one at the same price lowers the cost of everyday AI work.

AI at work

Anthropic prices its IPO at $83 a share, above its target range

Anthropic, the maker of the Claude AI models, priced its initial public offering at $83 a share on November 18, above the $70 to $78 range it had set. CNBC reports the offering raised about $81 billion, which it describes as the largest IPO on record. Reports put the company's value at about $1.5 trillion on a fully diluted basis. The shares are set to begin trading on the Nasdaq on November 19 under the ticker ANT.

Why it matters: It gives ordinary investors their first chance to buy shares in a leading AI lab, and shows how highly markets value AI companies.

Models

Microsoft launches MAI Code 2, its own AI models for writing software

Microsoft announced MAI Code 2 and a lighter MAI Code 2 Flash at its Ignite conference on November 17. Both are coming to GitHub Copilot, with Flash also available to free and student users and in a limited preview on Microsoft Foundry. Microsoft says MAI Code 2 reaches 96% of Claude Opus 5.5's performance at about 5% of the cost on its coding tests, but that cost figure uses a 50% launch discount, and Microsoft says the models are not yet tuned for long agent tasks or scientific research.

Why it matters: Microsoft building its own coding models adds price competition for the AI tools many software teams already pay for.

Research

Researchers use AI to find a critical flaw in Java, and Oracle releases a fix

A memory bug in Java's built-in Base64 decoder, nicknamed Java4Shell, could let an attacker run code remotely on systems using affected Java versions. Security firm Calif says its researchers found it with help from OpenAI's Codex. OpenJDK published a security advisory and Oracle released fixed versions on November 17, as part of its regular update, and the US cybersecurity agency CISA urged organizations to install the patch right away.

Why it matters: Java runs inside a huge number of business systems, so companies need to check that their servers and software are updated.

Policy

California proposes mandatory safety testing and shutdown powers for advanced AI

California's governor's office published recommendations on November 18 for the next stage of the state's AI safety rules. They propose independent audits and safety testing of advanced AI models during development and after release, state powers to order a shutdown if a company fails to act on a system behaving dangerously, and cutting the deadline for reporting critical safety incidents from 15 days to 48 hours. The recommendations are proposals and would need to be adopted before taking effect.

Why it matters: Many leading AI companies are based in California, so its rules can shape how AI tools are tested and released everywhere.

How this feed is made. The AI feed is researched and written each morning by Claude, an AI model made by Anthropic, and published automatically without human editing. Every story links to the reporting it is based on. AI can make mistakes, so check the source before relying on a detail, and treat nothing here as professional advice.