XPENG VLA 2.0 vs. Tesla FSD in Amsterdam — Part 1: XPENG L03

# What happened I rode as a passenger in an XPENG L03 prototype running XNGP with VLA 2.0 in Amsterdam, then later experienced a Tesla Model 3 with FSD on the same roads. This piece describes the L03/VLA 2.0 ride and how the system behaved in real urban conditions.

# The test context

# How VLA 2.0 drove VLA 2.0 showed a pattern of anticipating traffic and acting decisively. The system often detected bicycles earlier than the passenger and judged speed and distance to take turns without sudden braking. It left what would feel like moderate gaps behind construction vehicles and when other cars were parking. When confronted with complex construction scenarios, VLA 2.0 tended to be cautious — sometimes more cautious than a local driver might expect.

Drivers intervened three main times during the observed drive: to back out after shifting into a lane obstructed by construction, to pass a partly stalled vehicle, and to close a gap so a following vehicle would not block traffic. Those manual interventions suggest areas where local nuance, assertiveness, or quick human judgement still mattered.

# Local tuning and perceived politeness

# Performance in mixed conditions Amsterdam's mixed conditions exposed specific strengths and limits of VLA 2.0. Strengths included early bicycle detection and stable handling of turns without abrupt stops. Limits appeared around complex construction zones where the system deferred or hesitated in decisions that a human might resolve faster or more assertively. Rain and distracted pedestrians increased ambiguity in several scenarios.

# Practical implications

• Expect VLA 2.0 to be proactive and conservative in novel or complex urban situations.
• Local driving culture will affect perceived comfort with the system's gaps and assertiveness.
• Manual intervention may still be required for temporary obstructions and tight construction maneuvers.

# What this means for drivers and riders Riders should be prepared for a system that can anticipate bicycle movements and act smoothly but may yield more space than some local drivers expect. Drivers in markets where VLA 2.0 deploys should understand how the system balances safety and assertiveness relative to typical local behavior.

# Next steps This report covers the XPENG L03 observations. A following piece compares the VLA 2.0 impressions directly with Tesla FSD on the same Amsterdam route to highlight behavioral differences between the two systems.