Can anyone tell if this architecture makes sense to ship in their actual cars to replace whatever is doing the computation there now? That is the only way I can see this making any sense to develop.
They have already developed and deployed custom silicon for their cars.
There is probably some overlap in the IP between these chips, but Dojo is optimized for operating in a large data center cluster for training purposes, whereas their car chips are optimized for running pre-trained models.
I'm not sure if this is the point you're trying to make or if you're implying private companies aren't up to it, but... the other launcher of the same class, which will probably be launching within a year and is much more capable, is Starship.
That video uses a pretty strange calculation to estimate the battery weight. Here is a different way: 500 mile range * 2 kWh/mile means they need a 1000 kWh pack. The most recent packs they are shipping (Texas Model Y) are 543 kg [1] for 67 kWh [2], or 8.1 kg/kWh. If they have the same energy-to-weight ratio for the truck, its should weigh something like 8104 kg. As opposed to 17000 kg in the video.
You are right to be confused about that claim, but it because the units of the claim have been mixed up. The actual claim is 2 kWh/mile. That is a measure of energy per distance, kind of like gallons per mile.
If a truck is using 2 kWh/mile while traveling down highway at 70 miles per hour, its power draw will be 140 kW -- that is the one to compare to your 2 kW idling or 24 kW hard pull.
My EV uses 0.2 to 0.3 kWh/mile, so 2 kWh/mile for a truck seems plausible to me.
I ran into the case of needing to back up a write-heavy database without blocking anything, and came up with a solution: Writing a VFS ( https://www.sqlite.org/vfs.html ) that makes the application move the Sqlite journals file to a directory for processing instead of deleting them. Another process reads them to see what pages were touched and can very quickly get the changes to update the backup.
My house is set up like you describe. The whole solar/battery system never interconnects with the grid, and is instead set up like a generator. There is a transfer switch to choose between "generator" power and grid power. If there is some exceptional weather and the battery gets too low, we switch to grid. If the battery is full and the day is sunny we often dump power into the car, air conditioning, or whatever.
For pressure vessels like rockets, the mass of the structure and the mass of the fuel scale together as far as square-cube reasoning goes. The surface area of the structure scales with the square while the volume of the fuel scales with the cube, but the thickness of the cylinder walls must also increase, so you end up with cube vs cube.
Here is a graph of Texas's generation by source during that time and the surrounding time. It seems borderline dishonest to have an article about this topic not include this. The sharp dip in the middle on the image (the beginning of Feb 15th) is when my power went out. https://postimg.cc/FfG4XdDy
Just to clarify: this is about insurance for the payload, not the booster. So, the insurance costs only depend on how likely the insurance company thinks it is for the satellite to be blown up and how much the satellite costs, not how much the launch costs.
That’s interesting because I would assume a reused structure (rather then reused design) would be at greater risk for failure. More fatigue cycles etc.
I believe in the case of govt payloads they are self insured to around 80% or so but that’s obviously a different case. I also wonder if that affects the business plan for how to decide which payloads go on which rocket
I’d like to be at or near the first flight of a new plane using a time-tested design.
Failure rates (more specifically hazard rates) in most components tend to increase with time (I.e., follow a time-dependent failure model like a Weibull distribution and not a constant rate model). The “bathtub” curve is usually more associated with components with a constant failure rate (where the ideal place is after infant mortality but before wear out).
There are two types of uncertainties here and I think we’re talking about different ones. One is a design uncertainty that can be mitigated using a well-used and understood design. The other is a component failure uncertainty which increases as service life increases. So the ideal place to be is in a well-vetted design with new-ish components.
I not arguing about design uncertainty, rather production uncertainty. A failure in the production is more likely to fail on the first flight then the 3rd.
I don’t think reliability theory, particularly in aerospace, supports this claim unless the quality control is lacking. To be fair, SpaceX has had quality issues in the past but they seem to learn quickly from it.
To reiterate, the hazard rate tends to increase with service life. So as components age, there is an increased risk of failure.
