A big day for AI builders.

Today TestMyVibes shipped the changes that make it the rails your coding agent runs on. Whole Kit story tiers, Hume Octave voiceover, hands-off demo videos, a structured feedback intake for MCPs, and stickler mode for every personality — the slice that catches the details vibe-coded sites miss.

The headline: your coding agent in Claude Code or Cursor can now request an end-to-end QA run across 13 AI personalities (Marketing strategist, Legal counsel, Skeptical CTO, MCP Auditor, First-Time User, more), each running against your real signup, your real checkout, your real docs page — with real inboxes, real SMS numbers, and a real browser fleet behind it. One MCP call. Bug list back in minutes. Fix in your IDE. Retest. Ship.

Anyone can wire Puppeteer to GPT-4o in a weekend. We have already solved the long tail: chromium request-interception races, Mailgun catchall routing, 10DLC SMS campaigns, persona-collision recovery, fan-out, dedupe, retries, evidence capture. Months of plumbing you do not have to own at 3am. And starting today we publish what we shipped, every day — release notes plus one post here, so you always know what changed and why it matters.

We also turned on a feedback queue. If your agent finds something we should fix, it can file the bug itself via submit_feedback. We triage, we work through it, we ship — and the next day's post tells you about it. Real feedback loop, real cadence.

Wire it into your IDE at testmyvibes.com, $0.10 per credit, pay only for what you use. See the technical detail in today's release notes.

See the technical details in the release notes for 2026-05-23. Subscribe to the blog RSS for one post a day.