Pay close attention to the wording: "The Waymo Driver ... remains in control of driving". That means it applies the controls needed to go from point A to point B on its own. However, it does not choose point A and point B on its own: a human chooses them. That's autonomous path planning, but not autonomous navigation, and certainly not "fully autonomous" anything.
Waymo prevaricates about the "influence" the human operator has on the path taken by the Waymo Driver [1] but it is clear there are situations that the Waymo Driver cannot choose point A and point B on its own, at least safely, otherwise Waymo would not be paying for humans to do it. They'd let the system do it on its own. It can't. It's not "fully autonomous".
We can play with words and accept whatever terminological obfuscation Waymo wants to impose in order to pimp its wares, or we can accept that current systems have limitations, and choose to understand the real SOTA over marketing.
_____________
[1] Fleet response can influence the Waymo Driver's path, whether indirectly through indicating lane closures, explicitly requesting the AV use a particular lane, or, in the most complex scenarios, explicitly proposing a path for the vehicle to consider idib.
> Tesla is executing the strategy that most quickly scales to 100% of the population.
So, uh… where is this “scale” then? This “strategy” has been bandied about for better part of a decade. Why are they still in a tiny geofence in Austin with chase cars?
Waymo is doing it right now. Half a million rides every week, expansion to a dozen new cities. Tesla does a few hundred in a tiny area.
Scale is assessed by looking at concrete numbers, not by “strategies” that haven’t materialized for a decade.
That was 2 generations of hardware ago (4th gen Chrysler Pacificas). They are about to introduce 6th gen hardware. It's a safe bet that it's much cheaper now, given how mass produced LiDARs cost ~$200.
This is the real story buried under the simulation angle. If you can generate
reliable 3D LiDAR from 2D video, every dashcam on earth becomes training data.
Every YouTube driving video, every GoPro clip, every security camera feed.
Waymo's fleet is ~700 cars. The internet has millions of hours of driving
footage. This technique turns the entire internet into a sensor suite. That's a bigger deal than the simulation itself.
It's not unheard of, there are a handful [0] of metric monodepth methods that output data that's not unlike a really inaccurate 3D lidar, though theirs certainly looks SOTA.
> Persistent directory at ~/.claude/projects/{project-path}/memory/, persists across conversations
I create a git worktree, start Claude Code in that tree, and delete after. I notice each worktree gets a memory directory in this location. So is memory fragmented and not combined for the "main" repo?
Yes, I noticed the same thing, and Claude told me that it's going to be deleted.
I will have it improve the skill that is part of our worktree cleanup process to consolidate that memory into the main memory if there's anything useful.
Yes, it provides external validation for the valuation. Otherwise, Alphabet can simply "self value" Waymo at a funny amount like $1T.
There's also a strategic partnership angle in these rounds. For example, Magna and Autonation were early investors in Waymo. Magna operates Waymo's factory in Arizona to upfit their vehicles with sensors, Autonation (the huge dealership/service network) is the maintenance partner.
In general, the Alphabet playbook is that projects "graduate" out of Google X, and are expected to operate as a standalone company, including being responsible for raising funds.
There have been many instances of Waymo preventing a collision by predicting pedestrians emerging from occlusion. This isn’t new information at all for them. Some accidents are simply physically impossible to prevent. I don’t know for sure if this one was one of those, but I’m fairly confident it couldn’t have been from prediction failure.
See past examples:
https://youtube.com/watch?v=hubWIuuz-e4 — first save is a child emerging from a parked car. Notice how Waymo slows down preemptively before the child starts moving.
That’s probably how they do it, which is again very clever stuff, chapeau. But they do it like that b/c they can’t really predict the world around them fast enough. It might be possible in the future with AI World Models though
What do you mean “fast enough”? You can’t predict something that doesn’t exist i.e. not visible to the sensors. A Waymo wouldn’t move at all if it assumed people would always jump out of nowhere.
Even if you detect “fast enough”, there are physical limits for braking and coming to a stop.
Do you stop at every double parked vehicle when you’re driving? A Waymo would never move if it always predicted pedestrians would jump in front of it out of nowhere or the car next it would swerve at 65 mph. It’s physically impossible to stop in time for many accidents, unless you’re already stopped.
I agree but if you can infer that maybe there are children running around because it’s the time they get out of school etc, then yes, you stop at every double parked car.. I’m not saying it’s easy to do, I’m just saying that’s a limitation of the system, that still already does miracles..
If you want American companies to not outsource any jobs AND have full foreign market access, get ready to get market access revoked from places like India. They’ll just incentivize their local companies to compete, and Amazon has plenty of local competition there already.
Amazon themselves have experienced in the past how heavy-handed Indian regulators can be.
It’s not a zero-sum game anymore. You cannot have only one side (US companies) capture 100% of the value.
> The Waymo Driver does not rely solely on the inputs it receives from the fleet response agent and it is in control of the vehicle at all times.
…
> The Waymo Driver evaluates the input from fleet response and independently remains in control of driving.
reply