@RutgerDeMaeyer Resolve please.
Q3 accounts shows on a graph around 1.8 million cumulative unsupervised robotaxi miles. No way there can be less than 10,000 rides now.

3.11 million driverless miles to 18 October 2026 per Tesla safety report.

Largest geofence area is Austin at ~272 square miles, so if you imagine this as very roughly a 20 mile by 13.6 mile rectangle, you might be able to get a maximum ride distance of 25 miles but most and certainly average ride distance will be shorter. If there are as many as 3 times more miles between rides as during rides then that would seem as if they are being inefficient. 100 miles seems an extreme large estimate for miles driven per ride.

I believe we have confirmation of >1M fully unsupervised robotaxi miles in 2026. Since 100 miles is a high upper bound on the average number of miles per ride, that puts us comfortably above the threshold of 10k rides needed for a YES for 2026.

@dreev Is there anything you're waiting on for resolution? Just more detail on the announcement, or some specific clarification, or...?

@SeekingEternity dreev is not the market creator.

A slightly more detailed calculation:

Announced during Tesla's Q2 2026 earnings update on July 22–23, 2026, Tesla's Robotaxi fleet had officially crossed 380,000 fully unsupervised miles.

A high proportion of these 380k miles over 6 months to July might be testing rather operating as robotaxis, but I think that is unlikely to be the case recently e.g. with 620k+ miles in the 1.5 months from 22 July to 3 Sept.

Robotaxi tracker suggests about 5 miles per trip but maybe that is low with people testing out the system in short trips. 20 miles seems a high upper bound for trips but the robotaxis also travel to customers and maybe drive around waiting for customers. This seems unlikely to be more than 3 times the ride miles. So 60 miles per ride. 620k miles / 60 miles is over 10,000.

Other ways of arriving at a number also suggest it is over 10,000 rides. e.g. robotaxis likely do about 10 rides a day. There seem to have been well over 100 vehicles if not at least 200 in operation for around 10 days before and after the 3 Sept event. 170 spotted in last 30 days.

I don't really see any way the number of rides is less than 10,000. The only question is whether to wait until Q3 earning call in 3 weeks for just a bit more data and assurance that it is over 10,000 rides. @RutgerDeMaeyer any thoughts on resolving or waiting a little longer?

Looks like the only remaining uncertainty for resolving this to 2026 is the number of rides with no safety monitor in either seat. This is a tricky number to estimate (since Tesla is infuriatingly opaque) but it's probably above 10k by now. Unless something goes very wrong and they pull the plug on the whole program, it's safe to say it will happen by the end of the year if it hasn't already.

@dreev @RutgerDeMaeyer
per https://robotaxitracker.com/?provider=tesla&area=austin

19+5+5 =29 unsupervised vehicles since 11 May
If 10 rides per day per vehicle 290 rides a day
46 days at 290 rides would be > 13,000 rides

But maybe some vehicles are only in operation for an hour or 2 at peak demand times so maybe not 10 rides a day. So could still be close rather than over 10k yet. Another couple of weeks and it might start to get hard to estimate below 10k.

In case this hasn't been established
https://assets-ir.tesla.com/tesla-contents/IR/TSLA-Q4-2025-Update.pdf

page 10
"Automotive and Other Software The Robotaxi iOS app no longer has a waitlist in the areas we serve."

I would suggest that anyone in US can buy an apple iphone if they really want to in order to get access to robotaxis. Some people might not want to get an iphone but that is a different thing, most people in US could if they wanted to. So I think this should suffice for general availability for people in US that might want to use the service and no waitlist.

Rides without safety monitor and no chase car in Austin. 35000 rides in December and about 20% in Austin.

@brianwang Are you getting that 20% from robotaxitracker.com? (I see 17.5% of the fleet in Austin at the beginning of December, up to 19% at the end, based on the graph there. Potentially the Austin cars see higher utilization than the Bay Area ones -- not sure if that could get us a bit higher.)

But only a small fraction (6.1% since Jan 22 apparently, none before that) are fully unsupervised rides, which this market requires.

Do we have a good way to get the current cumulative total of fully unsupervised rides?

There seems some confusion so I will clear things up:

No humans in the chain = no safety driver.

Also seems to be still waitlist based/not really commercially at this moment. More like pilot testing.

Correct me if I am wrong.

@RutgerDeMaeyer To be totally clear, you're calling the passenger seat safety monitor a safety driver?

@dreev Yes

Phew! A definitive verdict. Thanks!

For those wanting to agonize more, see "Will we conclude Tesla launched level 4 robotaxis in summer 2025?". I think the remaining confusion is:

1. How fair is it to count the passenger seat safety monitors as safety drivers? My litmus test is whether the safety monitor has their finger on a kill switch, which there's some decent evidence is the case.

2. My understanding is there's technically no waitlist but Tesla isn't serious about meeting demand yet, in terms of number of cars and the app being iOS-only, no Android. But in my market we decided to say the current offering in Austin does count as sufficiently public.

3. Is Tesla cheating by using tele-operators? I think we'll be able to rule that out when they scale up but at current scale it's a question mark.

It's been pretty brutal figuring out how to adjudicate all this stuff. I'm probably overcomplicating it. The passenger-seat thing was just such a confusing curveball and people had such strongly divergent intuitions about it. But the longer Tesla goes without removing them (most recently Musk promised them gone by the end of the year and missed that deadline), the more clear it becomes that they were there out of more than mere "abundance of caution", I think.

Given the waitlist remained throughout 2025, I think 2025 must resolve NO

ping @RutgerDeMaeyer

@DavidFWatson Has the waitlist remained? I thought it had been reported people were getting invites fairly soon after requesting.
https://www.notateslaapp.com/news/3341/robotaxi-opens-up-to-the-public-tesla-removes-waitlist-for-ios-users

I agree that 2025 should eventually resolve NO but for the reason of 'humans in the chain' so none of the rides by end of 2025, or currently, count towards the 10000 rides needed.

@ChristopherRandles I wasn’t sure you’d agree on whether the passenger seat person counted as in the chain.

But my thinking was that “getting invites soon after asking” isn’t the same as “no waitlist”

@DavidFWatson My opinion doesn't matter. "Fully driverless and FSD" seems somewhat ambiguous re safety monitors in passenger seat. Seems like the software does all the driving and even when safety monitors intervene it is done in a manner similar to a change of destination and the driving decisions remain with the cars software rather than a driver taking over. Does that make it fully driverless? Alternately, is the need for a safety monitor mean it is not fully driverless as a human is still needed?

I thought "Human in the chain" seemed clearer that the presence of a safety monitor meant it did not qualify, but I suppose it depends on what you think is "the chain". If the chain is restricted to the driving then maybe there is no human in the chain but if the chain is the whole system then there is a human in the whole system.

I would suggest the phase being aimed at is more meant to be 'human in the loop' and the definition I first found suggested
The human "in the loop" might: 

• Provide data for training machine learning models.

• Validate or correct the outputs of an AI system.

• Handle complex "edge cases" or unforeseen scenarios that the AI is not programmed to manage.

  To me that seems to point towards the safety monitor being a 'human in the loop' and no rides should count yet. However what I think doesn't matter a great deal, it is what the creator thinks or intended.

@ChristopherRandles

😆 sorry, I didn’t check carefully and thought you were @RutgerDeMaeyer

@Bayesian Can’t seem to resolve one year I think.

@RutgerDeMaeyer sorry, you’re right. Linked mc options cannot be early resolved

@RutgerDeMaeyer Do you have a verdict on whether the robotaxi rides in Austin, Texas with passenger-seat safety monitors count as fully driverless? In case you're unsure I can recommend my own market that's still agonizing about the fairest way to answer that question.

@dreev I don't know what creator will think but is 'fully driverless' a condition or is the condition here 'commercially operational'. It is operating but not profitably as it has nowhere near the scale to cover costs of safety monitors, vehicle expenses and system development even if fsd driving development is excluded. So doesn't seem commercially operational yet to me, I would say that so far, it is only testing out the systems level of operation.

@ChristopherRandles commercial operation doesn't require profitability (though without it you'll stop operating sooner or later) but the scale definitely seems pretty small.