> Our new engine is based on intercepting SQLite's disk writes and doing clever stuff with them. It was so easy because the file format is not just well designed but amazingly well-documented.
Quite the hobby project for a lot of people too! Other folks doing this:
I'm sure there's a lot of really cool local-first databases out there, but SQLite has the benefit of being incredibly widely battle-tested, with literally billions of installations worldwide. It has received thorough security research and fuzzing (it's part of Chrome's attack surface after all). And there's tons of resources online to help people understand how to use it. Although I'm sure there are alternatives that serve certain use cases better it's hard to imagine anything coming close for ours.
That said, the storage engine we've built is not that heavily dependent on SQLite specifically. Any database that uses a write-ahead log like SQLite does should be possible to adapt to it in the future. So maybe we'll eventually open it up to a variety of choices, or even let you bring your own as a Wasm module.
Oh, I've been informed that DuckDB uses SQLite under the hood, so maybe compiling DuckDB to Wasm and running it on top of this will be possible, we'll see.
> DuckDB is indeed a free columnar database system, but it is not entirely built on top of SQLite. It exposes the same front-end and uses components of SQLite (the shell and testing infrastructure), but the execution engine/storage code is new.
Correct. DuckDB is really interesting technology, but it's not a direct successor to SQLite for transactional workloads. It's also very new: there's a LOT of new code in DuckDB on top of the (heavily fuzzed) SQLite parts.
(I use it personally, but it's not the same thing as what we're building with D1)
Also, for OLAP workloads there's Workers Analytics Engine. Analytics Engine is arbitrarily scalable as opposed to a D1-like solution - I can almost guarantee I've inserted more data into its internal variant (which is built on extremely similar underlying infrastructure) than just about any customer would think of doing, and it handles it like a champ.
Basically, we already have a product for doing DuckDB-like things, and the D1 architecture isn't great for high volume OLAP workloads.
Disclaimer: I'm an engineer at Cloudflare, but not on a developer platform team. I'm not speaking for our developer platform strategy or anything, I'm just commenting on how it looks from where I sit.
A few comments have noted that it's common for non-profits own for-profits (Mozilla Foundation, college endowments, etc).
What seems different here is that rather than funneling profits up to the non-profit, the intent seems to be to make payouts to fellow board members and employees:
> Only a minority of board members are allowed to hold financial stakes in the partnership at one time. Furthermore, only board members without such stakes can vote on decisions where the interests of limited partners and OpenAI Nonprofit’s mission may conflict—including any decisions about making payouts to investors and employees. [1]
IANAL, but has anyone seen this sort of arrangement before?
Every for-profit pays its employees, right? I don't think the fact that they make these payments is unusual, they're just making it clear that partners with a stake in the for-profit recuse themselves when the nonprofit is approving these payouts. It's similar to a company CEO not voting when the board approves their own pay package, even though they're a board member.
I agree that payments on their own (e.g. salary, bonuses) don't seem counter to the non-profit's interest. But what if we're talking about 100B of equity -- equity that otherwise would have gone to the parent non-profit?
That both Fred from Electrek and Taylor from Snow Bull ignore this distinction shows me their intent is less than neutral.
Looking at the original data sources [1], FSD seems to be improving over time at the metric that matters most:
* City miles per CDE have gone from ~50 to ~120.
* % of drives over 5 miles with no CDE have gone from 72% to 93%.
I've been pleasantly surprised that human oversight while the software improves seems to be a viable approach. FSD doesn't appear particularly close to Waymo/Cruise at the moment, but it's not as if they are crashing left and right (you'd certainly hear about it if they were).
Personally I have it, I don't find it enjoyable to use -- but I also don't feel unsafe when using it. Highway autopilot on the other hand I find immensely reliable and valuable.
It's sort of meta to this whole discussion, but it has been interesting to see "journalists" like Fred Lambert and the evolution of their coverage.
Fred was an early Tesla fanboy and investor; he used to gush over Tesla and Elon Musk in his writing at Electrek. It seemed like most of the time when Tesla was mentioned in an article, he also mentioned TSLA, the stock ticker...but he didn't seem to do that as regularly for say, Ford or GM.
It was rumored that Fred had a direct line to Elon and there were also occasionally public Twitter interactions between the two, but Fred gradually become more critical of Elon, especially over FSD stuff. After one critical article, Elon blocked Fred on Twitter: https://twitter.com/fredericlambert/status/14176561698297487...
In case it goes away, or if you're international and can't see, this seems like a good usage of Community Notes for added context and is currently published:
Regulatory Credits on account for a small percentage of Tesla Revenue.
2020: 6.2%
2021: 3.2%
2022: 2.7%
As of FY Q3 2022 Credits were only $287m against profits of $3.3b
https://twitter.com/FonsDK/status/1591489889924976640
TSLA Q3 2022 IR Deck: shorturl.at/mow47
Not sure, it depends on what the numbers look like going back to the founding of the company. Tesla probably wouldn't have survived in the past without the credits, but as others in the comments noted the point of the credits is to enable companies selling EV's to survive and grow during their difficulties in the switch to EV's. So I definitely wouldn't consider this a mark against Tesla, the credits are doing what they're supposed to, to an extent.
> Listen, weird nerds and $8 blue checks, I don't care about the 3rd Quarter of this year. I'm talking about the history of the company, its foundations and how it arrived where it is now. Which is a story most people clearly didn't know.
Wow, does he seem salty when called out. You can't untangle these profits from the car business. The credits won't exist without the cars.
The credits are not a magical widget that they are selling independent of the cars.
The actual implementation is a little more nuanced than always roll or never roll, with seven conditions required to avoid coming to a complete stop:
1. The functionality must be enabled within the FSD Beta Profile settings; and
2. The vehicle must be approaching an all-way stop intersection; and
3. The vehicle must be traveling below 5.6mph; and
4. No relevant moving cars are detected near the intersection; and
5. No relevant pedestrians or bicyclists are detected near the intersection; and
6. There is sufficient visibility for the vehicle while approaching the intersection; and
7. All roads entering the intersection have a speed limit of 30 mph or less.
If all the above conditions are met, only then will the vehicle travel through the all-way-stop intersection at a speed from 0.1 mph up to 5.6 mph without first coming to a complete stop. If any of the above conditions are not met, the functionality will not activate and the vehicle will come to a complete stop
Speaking as a city runner, you are 100000% wrong, and I have had numerous close calls demonstrating this.
If you roll through at 5.6mph while I am jogging into the intersection from your side, I will often be hidden by your A pillar the entire time right up until you hit me. And a 5.6mph point-blank hit to a pedestrian can be serious.
(edit: I have avoided these hits so far by watching people's eyes. If I can't see them see me, I stop regardless of right of way. This usually gives them a nice scare as they're rolling into the intersection and glance out their side window to see a person standing right there 'out of nowhere'. Not sure how I'll tell with FSD.)
As someone who did actually get run over when running through a crossroad as a kid, I _never_ run through crossroads since then, regardless of eye contact or not.
Runners (and the much loved sidewalk cyclists) are much farther from the road than a walking pedestrian. The drivers are NOT seeing you coming.
> Would you rather be hit by a semi-truck going 1mph or a bike going 28mph?
I'm actually not sure. The initial impact would hurt more with the bike, suggesting that perhaps momentum or something else proportional to velocity (not v^2) is a good heuristic there.
But if I tripped after impact, the bike would do no further damage to me and the semi would be ~1s away from killing me.
5.6 miles per hour is a fast jogging speed if you're on foot, twice as fast as typical walking speed. It is not a stop, doubly so if you're in a big metal box. Please don't do this.
Never having been stopped for doing this and confirming with police that they do not stop people for this are separate things, and I'm willing to bet that in your case it's the former, not the latter.
So presumably the Tesla is aware of all local laws and, more importantly, how strictly they are enforced from town to town, in every market the Tesla is sold or imported, in order to use this functionality safely and without resulting in ticketed traffic violations?
> more importantly, how strictly they are enforced from town to town
Does this dataset actually exist anywhere? I would bet that at best it might be able to be inaccurately inferred from other data.
Regardless, if we as a society decide to build self-driving cars, should we really be optimizing for the financial well-being of the driver (via tickets) over the physical wellbeing of the humans in the society (who are hit by FSD cars that roll stop signs in areas with less traffic enforcement).
Presumably because it matches the behavior of human drivers.
If self-driving cars were widely rolled out already, any large manufacturer could probably completely break traffic nationwide by rolling out an update that makes the cars actually follow every law and speed limit. (Just like work-to-rule can be as effective as a strike.)
Which, to me, just means the laws are wrong. The police often don't obey basic traffic laws like speed limits and signaling (in my experience of following police on the freeway).
I recently read an interview with a highway patrol officer whose opinion was that it's not really feasible for them to drive at the speed limit. Since a marked police car is already effectively a pace car -- no one wants to pass it for fear of being pulled over -- they would just distort the natural flow of traffic and probably create tailbacks everywhere they went, which would end up being more dangerous.
I'm not sure I agree completely with this reasoning, but it was an interesting perspective.
That's the same as saying that the speed limit (on freeways) is wrong because if everyone obeyed the speed limit, it would cause significant traffic congestion. In fact, you can see this when there IS a police officer driving the speed limit on the freeway and no one passes them. Traffic backs up for miles.
I recall seeing a forums post by a traffic engineer once and they said part of the process of deciding speed limits is, once the actual optimal speed for the road segment has been determined, subtract ~10mph to pre-emptively avoid complaints by the elderly that the speed limit is too high.
That was regarding suburban road planning though, not sure if the same applies to freeways and other major arteries.
It's not the same. The police officer is a moving speed limit restriction so cars driving at the normal speed will catch up to it and slow down and they will continue banking up for as long as the car is driving slower than normal.
You wouldn't get that banking up with everybody following the limit.
It becomes harder to change though because if the culture is to go x over the speed limit, people will probably still go x over the new higher speed limit.
This is ironic because next time there is a bike article, you get all the professional drivers in this thread commenting "but they never stop at stop signs!".
Can't have it both ways. The statistical evidence is clear: the average US driver is tremendously unsafe, untrained, unobservant and unskilled despite their country being built around driving everywhere. Their remonstrations on all the rule breaking they can perform safely stems purely from ignorance of their own inabilities.
in my country (the netherlands) i have seen people stop at stop signs at times when there is very, very little traffic (04:00 at night).
The point is also much more about creating a habit in which this kind of behaviour is just done, regardless of the state of the traffic on the road.
The law says you must stop for a stop sign, stop signs are placed in places in which sudden traffic participants could enter your field of vision at a time in which it is too late to react properly.
Also, people get fined for ignoring stop signs, even if no one is present.
Driving education in the netherlands is quite strict and so are punishments for drivers.
For instance, the driver of a car is always at fault for an accident with a "weak" traffic participant (foot/bike traffic), even if technically they werent at fault. (there is process to fight this in court if you assume ill intent/fraud is at play, although it is rarely used).
the reasoning being that the driver of a car has had a drivers education and can thus act responsible while driving a hunk of metal down the road at lethal speeds.
> in my country (the netherlands) i have seen people stop at stop signs at times when there is very, very little traffic (04:00 at night).
This is also normal in the US, despite what a few commenters on HN may have you believe. Hell, if anything, regular people out and about at 4am are better about following the basic laws, because you stand out a lot more when you flout them, and in the wee hours of the night the proportion of drunk drivers is far higher and so cops are looking for it.
i wonder if it’s a city vs suburbs thing. when i lived in the ‘burbs, doing highway driving, i got super accustomed to spotting the next red light and then coasting to a lower speed, hoping it’d turn green before i got there so that my average speed through the light was actually higher, city driving, you accelerate fast and break fast. once you’re accustomed to breaking with force and breaking often, a hard stop at a four way stop becomes a whole lot more natural than the alternative.
i don’t mean to overgeneralize. i expect there is at least some small correlation between density and how much you break at a four-way stop, and i’m curious what the other correlates are and how dominant full-stops are.
We humans often roll a stop sign (hopefully at a lower speed than 5.6mph) because we are impatient and because we have quite a lot of faith in our general intelligence and situational awareness to be able to make a reasonable call in these situations.
Still we often get this wrong and somebody gets hurt.
A computer is not cursed with our human impatience so why program it in?
But more importantly the computer has none of the general intelligence and situational awareness that a dim human has, and they won’t for a long time.
Have you ever been a pedestrian? The fact that most drivers don't give a shit about stopping for crosswalks isn't a reason for Tesla to say "fuck it let's go".
I wouldnt say outrageous but its pretty audacious to actually program in law breaking behavior. I would imagine this would instantly expose the company to liability?
Maybe my part of LA/SoCal is weird, but almost all of the people at the nearest stop signs stop at the signs, even when no other cars are there and no pedestrians are near the intersection.
In fact, the most annoying thing is that they will stop for too long despite the lack of cross-traffic or pedestrians traversing the intersection (in any direction).
> As the market cooled between August and September, Opendoor and Offerpad purchased fewer houses, while Zillow purchased more.
> The iBuyers also adjusted to changing market conditions by paying less for houses. The median purchase price in Phoenix peaked in August. Opendoor and Offerpad's median purchase price also peaked in August before tracking the market and declining in September. But Zillow kept paying more and more.
Tesla has also recently added wi-fi to their existing chargers, and will facilitate payment processing:
> With a growing number of Tesla cars on the road, a Wall Connector can pay for itself over time. Property managers will soon be able to set the price of charging sessions while Tesla handles payments automatically and securely – with no monthly fees. [1]
This is a model similar to what chargepoint offers, though I suspect it will have lower fees as their main goal is to sell more cars.
It seems media companies have a different interest in political races: if they call it a tight race, the candidates will pay them to run ads. And indirectly, if they report that it's tight, people will keep tuning in, and they can sell these eyeballs for a better price to their advertisers.
For example the 2008 Dem primaries with Obama vs Hillary. Obama was sure to win it months before the Dem conference, but I can recall CNN still calling it a race...
> if they call it a tight race, the candidates will pay them to run ads.
All the spending happens well before news channels start reporting ballot counts. Multi-million dollar political campaigns do not rely on the news media to tell them how they should spend their ad budget over a months-long election cycle.
By the time news orgs start reporting actual ballot counts, it's too late to spend any more money. A good chunk of votes have already been cast via mail. Poll stations have either closed or are at best a few hours away from closing.
