Public MCP servers fail ~40% of the time; Merge benchmarks 99.999% success vs ~60% for public tools
pk_iv · x · 2026-10-23
Merge benchmarked its tools at 99.999% success versus roughly 60% for public MCP servers, which fail 40% of the time because most are unmaintained side projects. Key points from the Navigators podcast episode: AI security hasn't had its big incident yet because tools break before data leaks; using AI without a DLP layer is "irresponsible"; most enterprise AI spend is in coding agents; agents only see "what's on the menu" when picking tools; you can't vibe code critical infrastructure; and "humans will become the blocker" to the singularity.
Related event: Public MCP Servers Fail About 40% of Calls, Benchmark Finds(2 posts)→
More from coding & agent
- Blinded agents compare formal verification vs code review: empirical results are in, and they're striking — rickasaurus · 2026-10-24
- GitHub ships HydraFusion Max, orchestrating Claude Opus 5.5 and GPT-6 models into one response — unixterminal · 2026-10-24
- kimo: an open-source no-GUI video editor that coding agents drive over MCP — LieTop7965 · 2026-10-24
- tinyfool builds a personal shell app: fold remote output, edit remote files locally with auto-diff — tinyfool · 2026-10-24
- Dev shares how they built a Claude client for the iPhone Duo dual-screen setup — rudrank · 2026-10-24
- Hanging Context: keyless MCP server for citable news lands in Claude's directory — bshivarthy · 2026-10-24