This is the core problem. The agent writes its own memory while working, so it has blind spots about what matters. I've had sessions where it carefully noted one thing but missed a bigger mistake in the same conversation — it can't see its own gaps.
A second pass over the transcript afterward catches what the agent missed. Doesn't need the agent to notice anything. Just reads the conversation cold.
The two approaches have completely different failure modes, which is why you need both. What nobody's built yet is the loop where the second pass feeds back into the memory for the next session.
Examining the cumulative hours waiting over time, it is a bit staggering just how much time Waymos are spending without a passenger or even assigned to pick one up. Peaking in March 2025 with over 304,000 hours, the California Waymo vehicle fleet is spending the equivalent of 12,700 days every month operational but without an assigned passenger trip.
If we assume 1,000 Waymos were deployed for public rides during this period (on the conservative side given recent fleet announcements), that ends up being around 12.7 days2 of waiting per vehicle per month. Further, the bias here is to be forgiving, as Waymos are not operational 24 hours a day.
It's mentioned in the article, the real problem was they kept trying to contact remote support to "verify" the light was out. Leading to a backlog of requests which they couldn't get through fast enough.
My only * to this would be Google Chromecast devices directly if you already have them.
They have an option (buried way under settings) to make the home-screen apps only.
> Turn on Apps only mode
> From the Google TV home screen, select Settings Settings and then Accounts & Sign In.
> Select your profile and then Apps only mode and then Turn on.
It also makes the device significantly more performant.
> I ride their robotaxis daily and see them performing better than Waymo, but it's obviously meaningless until we see accident stats after they remove safety monitors.
I've seen this claimed a lot but never have gotten a definitive answer.
Is this like "overall better but hard to pinpoint" or "this maneuver is smoother than Waymo" or something in between?
Would love to hear experiences with them since they're so limited currently.
Yeah Tesla has more smoothing, but IMO that's less interesting than the ability to navigate tricky scenarios and model other actors. Here's my collection of interesting videos, Tesla only because those are the ones I get forwarded to me. I'd love to see a similar collection for Waymo.
Smoother maneuvers, or things like seamlessly backing up a bit when it predicts that a large vehicle turning from an intersecting street won't have enough room to turn unless the car moves out of its way. It's really cool.
https://www.reddit.com/r/SelfDrivingCars/comments/1pem9ep/hm...
reply