VPS MVP Launch: First Real Users

Today we launched VPS to the first external developers. After a year of building, real users are actually using it.

What We Launched

Capabilities:

• 6DoF localization in pre-mapped areas
• Sub-meter accuracy outdoors
• Integration with Quest Guardian system
• Developer SDK for building localized experiences

Coverage:

• 50 locations globally (pilot sites)
• Mix of indoor and outdoor
• Partners include: museums, campuses, tourist sites

Limitations:

• Limited geographic coverage
• Requires network connectivity
• Daytime only (night accuracy poor)
• Beta stability expectations

Launch Day Metrics

Below target on all counts, but within acceptable range for MVP.

Immediate Feedback

Positive:

• "When it works, it's magic" - developer
• Accuracy exceeded expectations when successful
• SDK integration straightforward

Negative:

• "Fails too often in bright sunlight" - developer
• Latency spikes during high load
• Unclear failure messaging (why did it fail?)

First Week Bugs

1. Memory leak in feature extraction causing Quest crash after 30 mins
2. Race condition in pose verification causing occasional wrong pose
3. Network timeout handling not graceful (app hung instead of failing)
4. Geographic mismatch - some locations mapped in wrong coordinate system

Fixed 1 and 3 immediately (hotfix). 2 and 4 in next release.

Learnings

Testing wasn't enough: We tested extensively, but real developers found issues in 24 hours that we missed in months.

Coverage matters more than accuracy: Developers would rather have 70% success everywhere than 95% in limited areas.

Failure UX is critical: When VPS fails, the user experience should be graceful. We didn't invest enough here.

Latency variability hurts: Average latency is okay, but p99 latency is terrible. Need to smooth out tail.

What's Next

Short-term (1 month):

• Bug fixes for top issues
• Improve failure messaging
• Reduce latency variability

Medium-term (3 months):

• Expand coverage to 200 locations
• Improve sunlight robustness
• Add indoor support

Long-term (12 months):

• Self-serve mapping tools
• Consumer-ready quality
• Integration with core Quest experiences

Celebration and Reality

We shipped. That's worth celebrating.

We shipped something imperfect. That's reality.

The gap between MVP and product is still significant. But you can't cross that gap without launching. Now we iterate with real feedback.