I have reached my breaking point with v0 and am pausing my usage until the execution loops and overspending behavior are addressed.

Earlier on, v0 was great for quick iterations, small prompts yielded clean, targeted changes. Recently, even with budget-conscious models, the autonomous runner gets stuck in continuous execution loops that burn credits unchecked until the money runs dry.

A recurring issue is during testing: the runner repeatedly attempts actions it lacks authorization to perform (such as trying to open an admin route without credentials), fails, and immediately loops back to try again instead of yielding control back to the user. Automated testing should not trigger an infinite paid loop when an authentication barrier is hit, let users test and authenticate manually.

The breaking point was updating my portfolio: a single generation executed continuous calls that resulted in $29.53 charged in one go (Chat ID: u24AJhaUIMN).

v0 desperately needs:

Hard per-turn spend caps: A prompt must never burn dozens of dollars in autonomous retries without explicit user confirmation.

Intelligent loop break / auth failure exits: When a browser/preview action fails due to missing credentials or access gates, the runner must stop and hand control over rather than retrying.

Manual test-only toggle: An option to run pure code changes without letting the agent loop through autonomous UI testing.

Until guardrails are implemented to stop runaway autonomous charges, continuing to build here is simply too EXPENSIVE.