In Austin during SXSW, Tesla’s robotaxis did the simple thing well and the hard thing poorly. The rides were generally calm, clean, and competent, with human safety operators in the front passenger seat. Booking those rides through Tesla’s iPhone-only app, however, was a different story—intermittent availability, odd errors, and a juvenile user experience that felt out of step with a service asking for real-world trust.
Smooth Rides With Human Chaperones in Austin During SXSW
Over five point-to-point trips—three at night, two during the day, all on city streets—the modified Model Y vehicles behaved like cautious, occasionally timid drivers. They yielded predictably, accelerated smoothly, and only nudged past posted limits to match traffic flow. I never saw a safety operator take the wheel or hit the big red pause button; two chatted freely, two stayed silent, and one made it clear they were there purely for safety.
The in-car visualization showed what the system “saw” via its camera-only perception stack. It consistently tracked nearby vehicles and pedestrians, though it sometimes labeled cyclists as walkers. Before each departure, the car waited for my seatbelt click and a tap on “Start Ride,” a now-familiar ritual in autonomous fleets that mirrors Waymo’s cautious gating of passenger actions.
Tesla’s no-radar, no-LiDAR design keeps hardware costs low but sacrifices sensor redundancy. That trade-off didn’t surface in my rides, but it looms large in the wider safety debate, where rivals like Waymo and Zoox tout multi-modal sensing as a backstop when vision alone isn’t enough.
Availability And Pricing During SXSW in Austin
Getting a car at peak moments was the real friction. On SXSW’s busiest Saturday, the app repeatedly flashed “High service demand,” cycled through 20+ minute waits, canceled one ride seconds after it was confirmed, and threw a puzzling “No route available” error on another. It even demanded I re-enter a credit card already on file with my Tesla account.
By Monday and Tuesday, reliability improved: waits hovered between 10 and 15 minutes. That tracked with a small local fleet—an independent tracker counted roughly 38 Tesla robotaxis midweek, with only a handful running “unsupervised.” At that scale, hiccups were inevitable as crowds moved en masse between venues, day parties, and late shows.
Pricing was competitive but not a blowout. A 2.15-mile Monday night ride from downtown to East Austin ran $6.01, compared with $6.98 on Lyft and $16.92 on Uber at the same moment. Other trips in the 1.2 to 2.2-mile range cost between $5.79 and $6.48. Those are solid numbers for short hops, though frequent riders might still opt for Capital Metro’s $1.25 bus fares when routes align. The service area remained centered on downtown and nearby neighborhoods, ruling out airport runs.
Pickups were generally tight—within a car length or two—yet four of my drop-offs landed a block away despite open curb space closer to the pinned destination. The software may be conservatively avoiding complex curb maneuvers, but for a passenger hauling gear between sessions, a one-block shortfall is more than a rounding error.
An App That Undercuts the Experience and Trust
The most baffling element wasn’t reliability—it was tone. The app rates trips with lightning bolts and prompts tipping with meme-bait amounts like $0.69 and $4.20, only to flash “Just Kidding” when tapped. It’s a gag that might get a laugh internally, but it erodes confidence when you’re judging the maturity of a system navigating busy city streets.
That adolescent flourish, coupled with basic pain points—iOS-only access, occasional authentication loops, and vague failure messages—makes the service feel like a beta, even when the driving is better than fine. In mobility, UX isn’t garnish; it’s the on-ramp to trust.
Safety Data Transparency Remains The Test
Tesla’s Austin robotaxi operations are subject to federal Automated Driving Systems incident reporting. The most recent National Highway Traffic Safety Administration filings list 15 crashes in the city since last summer, with Electrek estimating roughly one collision every 57,000 miles—about 4x the minor-collision rate for human drivers. None of the January incidents involved injuries.
By contrast, peer-reviewed research cited by Waymo found its vehicles reduced injury-causing crashes by about 80% versus human baselines in tested markets. Even so, no operator is spotless; multiple reports have flagged Waymo cars mistakenly passing stopped school buses with extended stop signs. Autonomy’s safety story is getting better—but it’s not linear, and edge cases still bite.
Transparency is where Tesla lags most. In NHTSA’s public database, the company routinely redacts the narrative field that could explain circumstances around each crash. That opacity makes it harder for riders, regulators, and Tesla owners considering a Full Self-Driving subscription to assess actual risk. And first-person accounts, like a former Uber autonomy lead’s essay in The Atlantic about his Tesla veering toward a concrete barrier in FSD mode, keep skepticism alive.
Bottom Line From SXSW After Real-World Testing
On the road, Tesla’s robotaxis in Austin mostly did what city cars should do: drive politely, stop predictably, and get me where I asked without drama. Off the road, the app wobbled between spotty and sophomoric. If Tesla steadies its dispatch reliability, ditches the gimmicks, expands coverage, and shares clearer safety data, the service could compete on price and convenience. Until then, the car feels ready for prime time more often than the product around it.
