Sections

Search

OpenAI introduces GPT-6 Sol and Luna modelsEngineer Describes Workplace Where 'Everything Is Made by Claude Code'OpenAI claims GPT-6 has better prompt caching and diagnosticsllm CLI plugin llm-anthropic 0.29 adds support for Claude Opus 5.5TechCrunch AI reports on two years of AI data center fights in Pennsylvania
All stories

Models·

Simon Willison pushes back on 'MCP was always a bad idea' Hacker News post

Simon Willison argues MCP still matters for access control, auth, UI, and audit logging, not terminal agents calling APIs directly.

What happened

Simon Willison published a comment responding to a Hacker News piece titled 'MCP was always a bad idea?', arguing it 'entirely misses the value that MCP brings today.' He concedes terminal agents with unfettered internet access have little need for MCP: 'just let it call APIs directly.' [1]

What Willison says MCP is for

For setups less autonomous than a full terminal agent, Willison lists four needs MCP addresses: control over which external services the agent can access; authentication that keeps API keys out of the agent's reach; a UI for users to connect and authenticate further services; and strong audit logging. He frames these as the things 'we might want to build' beyond coding agents. [1]

Sources

  1. Simon Willison · Reporting · Sep 20, 2026
    MCP was always a bad idea?
    MCP was always a bad idea?
    MCP makes all of that so much easier to provide.