Introduction
In a fast-moving AI industry, missing a week's worth of news can leave you behind. But nobody has time to read everything. This article organizes the major AI news of 2025 and 2026 by category so the big picture is easier to track quickly.
1. The Model Launch War
The period from 2025 to 2026 was the most intense stretch yet for model releases. Major tech companies were announcing new models almost every week:
- GPT series: 5.1 -> 5.2 -> 5.3 -> 5.4 launched in rapid succession, with direct computer control arriving in 5.4
- Claude series: Claude Sonnet 4.6 launched, 1M context officially supported, Computer Use introduced as a research preview
- Gemini series: Gemini 3.1 Pro released with reasoning performance roughly doubled, later selected as the foundation model for Apple's new Siri
- Open source: Llama, Qwen, Mistral, and others quickly narrowed the gap with commercial models
2. Competition Over Agent Platforms
Big tech companies are now fighting over standards for the AI agent ecosystem:
- Anthropic: moved MCP to the Linux Foundation and pushed for standardization; its Agent Skills system was adopted by Codex and Gemini CLI as well
- Google: announced the A2A protocol, released the ADK framework, and pushed UCP for agentic commerce
- OpenAI: adopted MCP for interoperability with Anthropic and launched Agent Builder plus the Agents SDK; also advanced ACP for agentic commerce
- Microsoft: expanded the Autogen framework and built out a .NET agent ecosystem
3. Memory and Context Innovation
AI memory capabilities are improving quickly:
- Claude Code's Memory system and Auto Dream feature
- Official support for 1M-token context windows, up from 200K
- Persistent cross-session memory and project-specific memory management
4. The Fragmented AI Coding Tool Market
The AI coding tool market is crowded:
- Claude Code: rapid feature expansion with Skills, Memory, Hooks, Auto Mode, and Agent Teams
- Cursor: Cursor 2.2 introduced Debug Mode and Multi-Agent Judging
- Kiro: AWS's SDD-based coding tool
- Codex CLI: positioned around 24-hour autonomous coding
- Gemini CLI: tightly integrated into Google's ecosystem
- Open Code: a free and open source alternative
5. Security Threat Trends
Security threats have risen with AI adoption:
- Supply chain vulnerabilities found in AI-generated code
- Prompt injection attacks becoming more sophisticated
- Growth in AI-powered social engineering such as deepfake phishing
- New attack vectors emerging through MCP servers
6. Industry and Social Issues
- QuitGPT movement: 1.5 million users left after OpenAI's Pentagon contract controversy
- Apple + Google partnership: Gemini selected as the foundation for the new Siri
- KakaoTalk ChatGPT Pro promotion: a $200/month service sold for KRW 29,000 before ending early
- AI certification market: Anthropic, Google, IBM, and OpenAI all competing with their own AI certificates and completion programs
7. How to Track the News Efficiently
Tips for staying on top of the constant flow of AI news:
- Use bots: connect AI news bots to Slack or Discord for automatic collection
- Subscribe to weekly newsletters: examples include The Batch by Andrew Ng and TLDR AI
- Scan headlines first: read titles broadly, then go deep only in relevant areas
- Group by trend: understand individual news items by grouping them into models, tools, business, regulation, and similar categories
Closing
The pace of AI news keeps accelerating, but the big picture is clear: the rise of agentic AI, protocol standardization, mainstream AI coding, and growing emphasis on security and ethics. It is more useful to understand these broader patterns and apply them to your own domain than to react to every headline. A well-curated weekly summary is often more valuable than trying to read news continuously every day.