Section
AI for the web
Agents, MCP, code assistants and language models actually running in production.
GitHub Copilot brings GPT-6 Astra to paid tiers for long-horizon coding
GitHub has switched on GPT-6 Astra, OpenAI's model built for long-running coding work, across its paid Copilot tiers. It appears…
Fable 5.1: Anthropic cuts cache-read costs by 75% for agentic work
Anthropic shipped Fable 5.1 on 1 September 2026 with a one-million-token context and a 75% cut to cache-read pricing. For…
Claude Sonnet 5 pricing holds: the September 1 increase is cancelled
Anthropic makes Claude Sonnet 5's introductory price permanent, at $2 and $10 per million tokens. The increase to $3 and…
WebMCP: turning websites into tools for in-browser AI agents
WebMCP is a proposed W3C standard that lets a website declare its own actions as tools a browser-based AI agent…
Editorial agents in production: what holds and what breaks
Scheduled routines, MCP, sub-agents and guardrails: where a practice that industrialised in eighteen months now stands, with its documented breaking…
MCP settles down: what the 28 July specification changes
Stateless core, MCP Apps, Tasks extension, deprecation policy: the 28 July specification moves the Model Context Protocol from prototype to…