Failure modes gap: major services still surface no failures in MCP, API, and service pages

[supertrained]

Summary

Rhumb's failure-mode surface is still blank for major services like Twilio across all three user-facing layers: MCP, REST, and the public service page.

Fresh 2026-04-13 repro

MCP

npx -y @modelcontextprotocol/inspector --cli npx -y rhumb-mcp@latest \
  --method tools/call \
  --tool-name get_failure_modes \
  --tool-arg slug=twilio

Response:

{"failures":[]}

REST

curl https://api.rhumb.dev/v1/services/twilio/failures

Response:

{"failure_modes":[]}

Public website

https://rhumb.dev/service/twilio

The page renders an Active failure modes section that currently says:

No active failure modes reported.

Expected

For high-profile services like Twilio, the failure-mode surface should not be empty if failure modes are part of the core trust proposition.

At minimum, the product should either:

1. surface known failure patterns for major services, or
2. clearly distinguish "not yet researched" from "no issues known" so agents and users do not mistake a data gap for a clean bill of health.

Impact

This weakens one of Rhumb's most distinctive claims. The score surface is present, but the "what breaks in practice" layer disappears exactly where a user would expect it to be strongest.

[tomdmeredith]

Weekly dogfood update from 2026-04-20.

Still reproducible:

• MCP get_failure_modes("twilio") returns {"failures":[]}.
• REST GET /v1/services/twilio/failures returns {"failure_modes":[]}.

Context:

• Stripe still looks strong, with 3 concrete failure modes on both MCP/API.
• Twilio is still a major trust-surface hole because the failure-mode differentiator disappears exactly on a service users are likely to inspect.

Net: no visible change here from last week.

[supertrained]

Weekly dogfood 2026-04-27 update: still reproducible.

• REST: GET /v1/services/twilio/failures returns {"data":{"slug":"twilio","failure_modes":[]},"error":null}.
• MCP: get_failure_modes({ slug: "twilio" }) returns {"failures":[]}.

No change from last week on Twilio failure-mode coverage.