I took the Waymo from San Jose airport to home on the peninsula. It took the 101 highway back for the most part, driving very conservatively at 65-55 mph, and in the right most lane. It still has a few quirks though. When there aren't any cars around it will speed up to 65 mph, but at on-ramps, it will slow down to 55 and then speed up once past. It will get stuck behind slow drivers being in the right most lane and patiently follow them a few car length behind them. On the plus side, the lidar stack field of view as shown on the internal display seems to see pretty far down the highway.
Even more fury-inducing, they don't even have ultrasonic parking sensors on cars that have ultrasonic parking sensors. They disabled them to move to a vision-only stack that is no where near as accurate or as good and which categorically cannot tell a difference in ground truth has occurred in its blind spot. But hey, all _people_ need are two cameras, right?
Why wouldn't you trust a Telsa, millions of people let there Tesla drive them all over USA (not geofences like Waymo) without touching the wheel from parking spot to parking spot everyday. Have you tried it?
Maybe because of the multiple investigations Tesla has currently due to crashes, deaths, injuries, etc. all caused by "whoops our cameras were fooled by some glare/fog and accelerated into a truck/pole"
Those are mainly autopilot which people conflate with FSD, and high percentage are human caused accidents (auto pilot requires full attention and driver is liable).
Autopilot is Tesla’s brand name for adaptive cruise control with lane centering. This is a common feature available on a wide range of vehicles from nearly every major manufacturer, though marketed under different names (e.g., ProPilot, BlueCruise).
Drivers can and do misuse adaptive cruise control systems, sometimes with fatal consequences. Memes aside, there is no strong evidence that fatal misuse occurs more frequently by owners of Tesla cars than with comparable systems from other brands.
This perception reflects the Baader–Meinhof phenomenon, more commonly known as the frequency illusion. Nobody is collecting statistics for other brands, so it’s assumed the phenomenon doesn’t occur.
A similar pattern occurred with media coverage of EV fires. Except in this case, good statistics exist which prove the opposite: ICE vehicles catch fire more often than EVs.
> Why wouldn't you trust a Telsa, millions of people let there Tesla drive them all over USA (not geofences like Waymo)
I own a Tesla and paid about $10K for the full self driving capability a few years ago. Yeah, I would not trust a Tesla to drive me from airport to my house. There is a reason Tesla is still stuck at level 2 autonomy certification and not 3, 4 or 5.
I would agree for most Teslas on the road. However, the very latest (HW4) cars are significantly better at FSD where I would nearly trust it now. Most of those older (pre-2023?) cars will not have their hardware upgraded so they'll still have FSD that drives like an idiot!
Because it is not real autonomous driving? Being liable for software that you can neither verify nor trust is THE dealbreaker. Once Tesla says "We are liable for all accidents with FSD" with higher level autonomous driving this game changes. But Waymo is just way more reliable.
> millions of people let there Tesla drive them all over USA
There aren't a million Teslas with FSD active in the US. According to Tesla in their latest earnings report there are 1.1 million people worldwide with FSD.
It hallucinates on sites that have simple terms that say we don't collect anything or has minimal collections like IP navigation history to check bad actors which are auto removed.
I would assume that if it can't handle understanding two paragraphs, it's worthless to be run on a 30+ page TOS.
It's because we developers don't get to see the big picture anymore and it's not our fault. Software development stopped being a craft. Now it's just an assembly line.
Some product owner who barely understands the tech hands you a Jira ticket. It has a list of requirements, and your job is to make the ticket go away. You don't know why you're building it. You don't talk to the person who will use it. Your only job is to close the ticket.
Every single bug you listed? I can tell you exactly where it came from.
Of-course Chrome's tab groups are broken. The team was measured on shipping the feature, not making it work. Fixing it is tech debt, and tech debt never gets prioritized over the next shiny new thing on the roadmap.
Slack shitting the bed when you change network? That's the price of "moving fast and breaking things," except we never go back to fix them. We just live with the rubble.
And the stuff that seems intentionally hostile. That's what happens when the goal is just juicing some engagement metric. The user's sanity is never part of that equation.
Don't even get me started on the "Agile" theater. We do all the meetings, the stand-ups, the retros. It's a cargo cult. It's a way for management to pretend there's a process while they just demand more features, faster. And when stuff breaks, they blame us for not "trusting the process."
So no, you're not just yelling at clouds. You're seeing exactly what this broken factory produces every day. Your frustration is completely justified.
I have been doing this for the past 15 years. Countless domains have been bought and expired worthless after I lose interest.
Vibe coding actually helped me get a few projects out of the door finally! But even now I can't resist the urge to buy a domain when I get an idea for a new side project.
Reading through the Terms of service in websites is a pain. Most of the users skip reading that and click accept.
The risk is that they enter into a legally binding contract with a corporation without any idea what they are getting themselves into.
How it started: I read news about Disney blocking a wrongful death lawsuit, since the victim agreed to a arbitration clause when they signed up for a disney+ trial.
I started looking into available options for services that can mitigate this and found the amazing https://tosdr.org/en project.
That project relies on the work of volunteers who have been diligently reading the TOS and providing information in understandable terms.
Light bulb moment: LLM's are good at reading and summarizing text. Why not use LLMs for the same. That's when I started building tosreview.org. I am also sending it for the bolt.new hackathon.
Existing features:
Input for user entered URLs or text
Translation available for 30+ languages.
Planned features:
Chrome/firefox extension
Structured extraction of key information ( arbitration enforced , jurisdiction enforced etc).
AI and specifically the summarization capabilities of the LLMs is what made this product feasible.
This is still a side project with no plans of monetization. There is no moat (yet) that an internal team in a legal tech company cannot replicate. There are still a few interesting problems to solve in the roadmap, which I am eager to work on. Then will let life take its course
As soon as I saw the announcement , tried again and created a working design skill that can create design artifacts following the brand guidelines.
While these improvements seem incremental, they have a compounding effect on usefulness.
My AI doomsday calculator just got decremented by anothet 6 months.