Tracking Tesla's Progress on Autonomy and Robotics

I think it’'ll be useful to have a thread tracking Tesla’s progress in these two important growth areas for the company.

Yesterday, we got news from AI insurance company Lemonade, announcing:

Lemonade Autonomous Car insurance, a first-of-its-kind product designed specifically for self-driving cars, starting with Tesla FSD.

The new offering cuts per-mile rates for FSD-engaged driving by approximately 50%, reflecting what the data shows to be significantly reduced risk during autonomous operation. Lemonade expects further reductions as Tesla releases FSD software updates, which are anticipated to make the cars even safer over time.

They’re putting their money where their mouth is, based on data.

1 Like

Today, riders have posted of their fully autonomous, driverless and on-board safety monitorless, paid-for rides in Austin:

And Elon confirms:

That’s less than 7 months from trials with safety monitors. Took Waymo almost 2 years to reach that milestone.

And confirms my expectation that driverless rides would start with the Model Y, not CyberCab.

2 Likes

That’s less than 7 months from trials with safety monitors. Took Waymo almost 2 years to reach that milestone.

Tesla has a chase car following each of these ‘safety-monitorless’ robotaxis. I suspect that all Tesla has done is move the safety monitor from the robotaxi to the chase car and given them a remote kill switch.

That’s hardly some great feat, but I’m not surprised that Musk would be deceptive this way and that bulls would buy it leading to the stock to pop today.

4 Likes

And it’s only a few cars within the fleet:

Not all of Tesla’s fleet in Austin will be fully driverless. Per Tesla’s AI lead Ashok Elluswamy, the company will be “starting with a few unsupervised vehicles mixed in with the broader robotaxi fleet with safety monitors, and the ratio will increase over time.”

Tesla is charging for the rides, according to one rider who posted on X. There also appears to be a chase car following the driverless vehicles.

Tesla launches robotaxi rides in Austin with no human safety driver | TechCrunch

2 Likes

Not unlike Waymo in its early driverless days.

1 Like

Are you claiming this is fully autonomous?

I’m unable to find anything supporting this claim. So far what I have found is that Waymo did not have “follow cars”, although they did have remote monitoring (as they do now). The pieces I read do not indicate whether the remote monitoring was 100%, as the Tesla follow-cars are.

Do you have any information on that?

This 5 year old video shows a chase car catching up to a Waymo:

There are also internet reports of a red stop button in chase vehicles, particularly in Arizona. Waymo claims its chase cars were there to have human support personel quickly available.

1 Like

They’ve scaled from one car per ride to two!

It’s a hyper-exponential doubling!

Didn’t they achieve the follow-car milestone in December?

The driverless delivery from June 2025 isn’t being repeated.

The “no safety monitor in the car” (never mind the chase cars behind the curtain) in December 2025 and today is the same.

This is why Tesla lacks credibility.

Earnings next week, coincidence?

When Tesla can provably demonstrate real L4 autonomy and start accumulating the millions of miles to statistically demonstrate human-level safety, then I’m happy to adjust my view.

Let’s see the real data.

1 Like

Hot and fast reported progress these past couple of days:

Alex Roy previously wrote a story claiming his Roomba was better than his Tesla in 5 ways.

Now it’s:

“CRAZIEST events in snow – but FSD did it! Holy s**t,” Roy said in one update. “Snow performance and recovery is unreal,” he later added.

“The video will be crazy,” Roy assured us.

2 Likes

These videos of “not true autonomy” just further reinforce that L4 autonomy is not imminent for Tesla.

It’s evidence of non-autonomy.

1 Like

There’s nothing wrong with using a chase car. What’s deceiving is making a post on Twitter saying the safety monitors have been removed when really all Tesla did was move the safety monitor into a close-following chase car with a kill switch.

2 Likes

What’s actually deceiving is you trying to put words in Musk’s mouth that he didn’t say.

What he actually DID say, and I quote:

Just started Tesla Robotaxi drives in Austin with no safety monitor in the car.

Which is 100% factually correct, despite your spin attempt.

3 Likes

We’ll have to check back in 2027.

This chase car video supports the view that the most likely outcome is little if any true autonomy for Tesla in 2026, certainly a very tough road to get mileage to even just a convincing start to statistically demonstrate autonomy in any meaningful operational domain with human safety levels this year (maybe they can go back and forth down main street?).

Even a modest 2 million miles seems a reach in 2026 (which Waymo can do in about a week).

Tesla doesn’t provide much data on their capability and safety, but ironically, by releasing these orchestrated hype videos they are confirming that true L4 autonomy is not coming soon.

Tesla has a march of 9s safety problem to work on.

Here’s a summary of Tesla taxi status.

(a 2027 checkpoint still seems reasonable)

They are not really out of the gate yet for true autonomy.

I’ve been wondering exactly what hyper-exponential looks like - now we know!

Maybe they can pivot to marketing, is that their true calling?

March 7, 2026

See, the thing there is you can see 1 car multiple times. However, other people ( literally dozens, probably well over 100 actually. ) are actually actively tracking the individual vehicles. ( You can check details out at robotaxitracker.com )

They have added 25 cybercabs in Austin, but only a handful of them are still in operation, and none seen particularly recently. They are also only in testing, not in service.

They added 8 unsupervised, but only 1 is still in operation. A single 1, operating in a very tiny area, is a publicity stunt.

89 Robotaxis have entered operation, but only 45 have been in service within the last month, and evidence suggests even less than that are currently in operation, but it’ll take a bit more time to be certain. ( Some quick calculations based on approximate daily sightings suggests it takes ~28 days to have a 95% confidence that all active vehicles have been seen. )

Tesla has averaged a pretty terrible ~45% availability.

You’ll also notice how Tesla has gone relatively silent mode ever since they announced the first driverless rollout. Almost like they have no good news to report.

3 Likes

The robotaxi tracker is showing that Tesla did 23 rides in Austin over the last 7 days and 6 of them were “unsupervised.” So, barely 3 rides a day, and not even 1 ride per day unsupervised. It’s almost like everyone who knew how to do anything quit.

3 Likes

I have a friend here in the SF Bay Area who for a week decided to use Tesla’s robotaxi service whenever he could. What he found was that every ride he took, except 2, that week (about 30 in total iirc) was on a vehicle that had not been reported previously on the tracker.

And those 2 were ones that he had reported for the first time on the tracker earlier that week.

The robotaxi tracker depends on riders to report their rides. Most don’t, especially now that the influencers have mostly left Austin.

1 Like

This certainly sound totally, or mostly made up. Especially given this:

[ Tesla Robotaxi ride reports] are generated through an automated, camera-based data pipeline, capturing vehicle telemetry, safety events, and operational metrics in real-time. These reports are compiled via external camera recordings (Safety Events and Fleet Learning) and cloud-processed metrics, allowing for analysis of passenger experiences and vehicle performance, particularly following recent deployments in cities like Austin

This is the second, perhaps third time of “I have a friend” or “I know someone with inside information” I have seen. It’s not credible, particularly when you cap it with “all the influencers have left Austin”. What is that about? Austin thrives. I have a friend who was just there! (Ha ha, I make a structured joke, referred to as a “call back” in the comedy world.)

This is true but for full accuracy, I understand that vehicles can be submitted manually too by riders. So I suspect what smorgasbord1 has said about friend is actually true. But such user submissions are investigated first to substantiate validity.

The app has a published methodology:

How ironic, as I am the most credible poster here. No-one quotes from and links to more primary souce documents than I. Whether it’s NHTSA documents such as FMVSS or Proposed Rules, or NHTSA Forum keynotes and recordings, DMV documents, or SAE’s J3016, I always quote directly and provide links where possible.

Despite my almost 15-year track record (on the paid board initially), you doubt the veracity of my posts? Well, here’s a photo my friend took at the NHTSA AV forum:

And, here he is inside the demo CyberCab at the event:

You can see a Zoox and a Holon vehicle in the background. Go ahead - run a reverse image search if you think these are from some public source.

You are certainly free to disbelieve what you want to disbelieve. I can only guess how upsetting it must be to have your world view disrupted.

Quoted with substitutions, without a link, and without context. In fact, you are wrong. In your zeal to call me out you defy logic: You think that a small private website has access to Austin and the SF Bay Area’s police Flock cameras such that they can grab not only license plates but whether there’s someone in the driver’s seat, not to mention whether the person in the passenger seat is a Tesla safety monitor or a passenger?

Really?

Here’s a quote from the actual methodolgy link, which @gledmo posted:

Vehicles are added to the tracker through community submissions. When a user spots a robotaxi, they can:

  • Sign in and submit the license plate number (validated by provider-specific formats)
  • Upload at least one photo of the vehicle for verification (up to three)
  • Specify exterior and interior colors (Tesla submissions only; Waymo uses defaults)
  • Indicate the service area where it was spotted
  • Add optional notes about the sighting

You get that? “Community Submissions.” Not “external cameras.” Heck, the word “camera” doesn’t even appear on that page.

To back up my prior assertion on most riders not reporting their rides, here’s what the Robotaxi Tracker page shows about its own data:

You reading that 0.6% of rides being logged? I’d say 99.4% of rides counts as “most.”

Hey, for extra credit, I leave it up to you to find the ride reporting leaderboard, which includes per-user stats of “rode first” vehicles reported, just to see if what I reported is “credible,” lol.


It’s becoming more and more apparent to me that there’s not much point in my posting here. I could continue to ignore posters writing nonsense, not following logic, etc., but I think it’s become clear that there’s really no point in me continuing. Not only is the sheer garbage repeatedly posted useless, but that almost all of the audience here is incapable of a nuanced discussion is perhaps the most disappointing of all. And now being challenged in an aggressive manner kind of puts a bow on it. There were many other respectful ways to ask and probe for my data sources, but that didn’t happen and I feel it’s emblematic of this board.

3 Likes