Tesla Completes Coast-to-Coast FSD Cannonball Run

A 2024 Model S completed a coast-to-coast Full Self-Driving run from Redondo Beach, Los Angeles, to midtown Manhattan, New York, covering 3,081 miles, according to reporting by The Drive and CryptoBriefing. The on-road driving time was 58 hours and 22 minutes, with an additional 10 hours and 11 minutes spent charging, as reported by The Drive and Autoblog. The vehicle was running FSD v14.2.2.3 on HW4, per The Drive, Teslarati, and CryptoBriefing. The attempt was led by Alex Roy and a small team of autonomy experts; Roy and The Drive documented prior Cannonball attempts that required 32 human interventions in December 2024 and a subsequent run with roughly five minutes and 20 seconds of intervention, per CryptoBriefing and The Drive.

Per reporting in The Drive, Teslarati, and CryptoBriefing, the car ran FSD v14.2.2.3 on HW4. Articles describe the route encountering varied conditions including snow, ice, slush, and active winter storms on portions of the trip, and note the team cleaned exterior cameras during charging stops to preserve sensor performance, per Teslarati and The Drive. Multiple outlets emphasize that despite the zero-intervention run, Tesla's FSD is generally categorized as SAE Level 2 autonomy in public coverage, meaning human supervision remains the regulatory and technical baseline, according to Autoblog and Yahoo Autos.

Editorial analysis: Companies and observers tracking autonomy milestones treat coast-to-coast, zero-intervention runs as high-visibility demonstrations rather than peer-reviewed safety validations. Reporting frames this run as the latest entry in a series of attempts led by Alex Roy that show iterative improvement in Tesla software versions, with coverage noting a reduction from dozens of interventions to near-zero as FSD versions advanced from v12.5.6 to v14.2.2.3, per CryptoBriefing and The Drive. Public reporting also contrasts Tesla's data-driven, unconstrained-road approach with competitors such as Waymo, which operates in geofenced, pre-mapped urban areas, as described by CryptoBriefing.

Editorial analysis: News articles uniformly caution that a single documented run does not by itself establish safety across fleet-scale operations or diverse edge cases. Reporting highlights that strict test rules and the presence of autonomy experts on the team shaped the conditions of the attempt, and that charging logistics and winter detours materially affected total elapsed time, per The Drive and Autoblog. Multiple outlets reiterate that regulatory classification and real-world supervision expectations have not changed solely because of this demonstration, per Autoblog and Yahoo Autos.

For engineers and teams building perception and planning stacks, the reported details that the crew cleaned cameras during stops and that software updates correlated with reduced interventions underscore two recurring operational themes: sensor cleanliness and rapid iteration of perception/planning models. Reporting-based accounts of the run provide concrete telemetry-like datapoints (distance, elapsed driving time, charging time, software build) that practitioners can use as benchmarks for long-range route-handling performance, per The Drive and CryptoBriefing.

Editorial analysis: Observers should look for independent validation, larger-sample fleet metrics, and formal disengagement logs released by third parties or regulators. Reporting to date is based on the team narrative and journalistic coverage; neither Tesla nor an independent regulator has published exhaustive telemetry in the sources reviewed. Also watch how future software releases and weather-seasoned runs change intervention counts in reporting from the same operators and independent testers.

What was reported by The Drive, CryptoBriefing, Teslarati, Autoblog, and Yahoo Autos documents a high-profile, zero-intervention coast-to-coast run using FSD v14.2.2.3 on HW4. Editorial analysis indicates the result is an important demonstration milestone but not, on its own, a fleet-level proof of fully autonomous capability.