What will be the data source to establish whether Tesla was able to perform 10,000 or more rides?

@MuGoGonzalez I don’t know yet, do you have ideas? Whatever seems most reliable at year end I imagine. There may not be high quality data in which case I’d make a guess from testimonials or chatting with ChatGPT’s DeepResearch.

@aashiq I don't know whether there will be more than a general statement in a shareholder meeting. One could use that and taking whatever number they say in a "more than X rides since June" statement as the endresult.
Or you simplify it by saying that if at the end of the year Tesla offers a robotaxi service (i.e. no driver in the driving seat) somewhere that is open to the general public, the question resolves as yes. As so far, Robotaxi Austin is invite-only and the planned service in California will be more like a "classic" ride-hailing service with somebody in the driver seat, this would still be a milestone which isn't reached yet and can be verified easily by just looking at the Robotaxi service website: Get Started With Robotaxi | Tesla Support

@MuGoGonzalez certainly, public availability will play in but l don’t want to prematurely operationalize the resolution criteria. It will resolve based on my best estimation of the true number, and after some deliberation in the comments here.

@aashiq Well, you are the creator. However, under these circumstances I won't participate as this adds too much uncertainty whether our perceptions of what is "sucess" are the same.

@MuGoGonzalez that’s fine! However, I don’t think we’re disagreeing about what constitutes success so much as epistemology. Namely, how do we know how many rides they did?

I am inclined to make an informed guesstimate based on a model that I will concoct at year end given the available info after vetting it with the holders for suitability.

You’d rather reduce expressivity in the market and lean on public sources, which reduces modeling uncertainty and subjectivity. That’s a fair tradeoff but it introduces different kinds of uncertainty that would be unacceptable to some of us. Namely, if we do it your way, there may not be an official statement so now we are introducing the annoying possibility that Tesla is widely believed to have succeeded but without credible info we can use to definitively resolve the market.

@aashiq I agree, it's more of an underlying philosophy issue - I'm agnostic about the truth, I just want to know which arbitrary and flawed data source I have to match to be judged as "right" or "wrong" 😉

Especially in a case like this where Tesla will OBVIOUSLY claim success no matter what, but given their history, it's rather unlikely that there will be any objective and reliable source to corrobate the claim.

@MuGoGonzalez
Yes, I agree, and yet I don’t want throw the baby out with the bathwater. Manifold is a rationalist playground. In the real market, we have to satisfice with arbitrary and flawed data sources. Here, we can do our best to bet based on the actual truth as operationalized by other people who are trying their best to get at it. Where else can you do that?

We have a classic bias variance tradeoff. If you do it my way, you have a bunch of variance based on the very good points you are bringing up. But you have less bias that moves you away from the intuitive probability “Will Tesla succeed?”. Your way has less variance but introduces a bunch of bias to the downside, which makes the interpretation of the market price less satisfying to one’s intuition.

@aashiq Two thoughts on that:

1. I think your explanation made it clearer to me where we diverge here. For me, the "truth" is quite self-evident: No, Tesla does not have the technological edge and camera-based AV technology has limitations that will make it impossible to go the last step from the 99% that FSD offers today to the 100% that you need for the kind of AV that Tesla envisions, therefore making a safe and reliable driverless ride-hailing service impossible for them right now. However, I concede that this only applies if you see the safety aspect of AV as paramount. If you're okay with "probably not much different from the safety level of a human driver, even though the accident-prone behaviours are different", Tesla is probably already there.
Long story short: For me, the question can only resolve as "No" given all the circumstancial evidence as we have of today, because everything else is maybe a marketing scam but will not satisfy what I understand as a fully AV system: A reliable vehicle that will never kill you in an accident that a human could have easily avoided.
However, I now understand you as far more agnostic on the state of the technology - maybe Tesla is delivering, maybe it isn't. In this case, it makes absolutely sense to steer clear from any random metric that can be played to discern the real capability of the system.
But for me, the only way that Tesla can suceed is by faking it and coming up with a weird metric that somehow sounds plausible and might tell shareholders that yes, Tesla's share price is still warranted because it is on the right path and will capture the market for AVs any minute now. But I have no idea whether people - whether in general, as Tesla shareholders or as Manifold members - will believe it or not. However, if there is a clear metric in place, it doesn't matter whether I believe Tesla managed to clear that hurdle fair and square or by gambling with the lives of real people, I just have to think about whether it is plausible that Tesla can show they satisfied the preset expectations.
Therefore, a preset basket of conditions that Tesla has to fulfil for the question to be resolved as yes seems to me the fairer approach to the question as I can compete with people who think I'm a complete moron for not seeing the genius of Tesla's approach without neither of us feeling duped at the end no matter what the outcome.

2. As someone coming from a pure Brier score forecasting background, I'm struggeling with the whole market aspect anyway. I still haven't figured out how I can translate my approach of "update your forecasts regularly" into market prices. With the percentages I can somehow cope, although I'm bothered that the same amount of Mana can imply completely different leverages for swings into a certain direction depending on liquidity of the market.
But I haven't really figured out yet what the implications are for updates of my assumed probabilities - what bet is the right thing to do if I think the probability for an outcome has changed AND the market doesn't reflect these probabilities? I really haven't gotten the hang of that yet...

@MuGoGonzalez
Ok, thank you for clarifying your position. I don’t want this to be a market solely about whether Tesla develops a proper AV system. I am equally interested in whether regulators and the public will tolerate Tesla’s technology as it is in 2025.

On your point about betting strategy, liquidity (and similarly fees) are always an important consideration in traditional finance as well. There are volumes written about optimal portfolio construction given a set of fair values, liquidity, time horizon, fees, correlations between assets, bankroll, etc. On the margin, I think manifold bettors would do better to bet much larger fractions of their bankroll on the highest conviction markets with deep liquidity.

Ultimately, my market doesn’t matter very much strategically because the liquidity is terrible. In general as to whether to update when the market doesn’t reflect a change in your fair value, it really depends why your fair value changed and on what time horizon you expect the market to converge. Sophisticated trading systems shouldn’t just buy or sell everything that diverges from fair value. Simplifying, they should develop alpha signals that predict price changes at a given time horizon and participate in markets with a large alpha for the time horizon where there is liquidity to close the position. So if the market may never converge to your change in fair value, consider just doing nothing. On the flip, if a highly liquid market tends to track your changes in fair value and the change is due to some news event that the market tends to react to quickly, bet big.

@aashiq Thank you very much for sharing your insights on betting strategies - that's good food for thought! I already came slowly around to the fact that I had to build a more sophisticated "translation" tool for my (simply probabilistic) forecasts that advises me on how to translate eventual changes into actions on betting markets. Now that I feel I have a grasp of the markets I'm active in that's actually a good next step to challenge myself to dig deeper and not being content with "more or less understanding what drives market development".

Video of an intervention: no mogic door-handle button, just a click on the screen https://x.com/edgecase411/status/1937655601624154411

"the removal of the safety driver is the biggest milestone in development of a true robotaxi, not an incremental step that can be ignored." https://www.forbes.com/sites/bradtempleton/2025/06/22/tesla-misses-robotaxi-launch-date-goes-with-safety-drivers/

This market must resolve NO in the case of on-board safety monitors.

What a joke that an "autonomous taxi" still requires human labour in the car.

@JackAllison agreed

@aashiq

https://www.youtube.com/watch?v=yXv5xz0bHYQ

@StCredZero bullish

Observers seem to have right hand in same place. Probably a software 'hack' to change function of a button probably to being an emergency stop. So easy to revert if/when observers removed.

Presumably no rides count yet until observer no longer in cars?

Only 14 invites for first day. Impression I have is that streamers had impression that it wasn't a time or number of rides limited offer, they could do this from now on. Perhaps suggests a seamless expansion as restrictions are lifted but we will still have to wait and see if there are any breaks in service. $4.20 flat fee was being charged but maybe that will change when longer distances are possible.

Currently limited to 6am until midnight. 35 cars so far and if they do a ride per hour (maybe 1.5-3 if customers are waiting?) for maybe 10 of 18 hours a day. Not sure, but seems quite likely that number of cars goes up before or when observers are removed, but even if not, could get to 10,000 within a month of removing observers.

Big thing now would seem to be if emergency stop or accidents occur in anything other than bizarre rare instance(s) that couldn't really be avoided - that would likely delay plans to remove observers.

@ChristopherRandles FYI, they are sending new invites every day and anyone can sign up on a waiting list https://www.tesla.com/robotaxi

@MarkosGiannopoulos I do not see a place to sign up for a wait list. The link you provided does not day it's to join a wait list.

@WrongoPhD "Be among the first to take an autonomous ride in Model Y. Complete the form below to get updates about when our Robotaxi service is launching near you."

@ChristopherRandles yes need to have observers out of the car

@aashiq Is it your intention that waymo would resolve no currently, because I think it currently might?

This, from perplexity:

Can you say more? I don’t see why the screenshot shows that Waymo would resolve No. do you have stats on Waymo rides?

• Rides that require a human to be present or to take control will not count as driverless.
  
  This means waymo would currently fail this, right?

@NathanpmYoung what I've heard is that Waymo allows humans to tell the cars what to do at a high level (E.g. Reroute, or go through a construction zone) but they can't actually take control.

@SeekingEternity Agreed. And it would be quite a monkey wrench in these self-driving car markets if Waymo were declared to not count because it's SAE level 4 and not level 5. Other traders can chime in if I'm wrong but I think the universal assumption for this and related markets is "driverless ride-hailing à la Waymo".

Okay but that is not what it says:

"Driverless rides must be completely autonomous with no human available as a backup."

IIUC waymo rides have people in an HQ who can take control.

@dreev I consider Waymo as counting. When I wrote no human available, I meant "no human continuously available" so teleoperation or a backup driver in the car don't count.