Starship "making LEO" is not a significant challenge--the existing flights have explicitly targeted a (very slightly) suborbital trajectory. They could have done otherwise at any point, but for now it's more important to guarantee that the stage comes down immediately. None of their current objectives require more than ~1/2 of an orbit.
Starship v3 flying will be a significant leap, though. It's the first with the Raptor v3 engines and has many other improvements as well, such as updated grid fins and hot staging ring. It will be the first that achieves close to the intended capacity of ~100 tons.
Propellant transfer is indeed a significant challenge. They have already demonstrated internal transfers between tanks, but not between spacecraft.
> Starship "making LEO" is not a significant challenge
Of course it is. I say this as someone who sturdied astronautics.
You’re broadly correct, though. My point is the action shifts to Hawthorne and West Texas for the next year or so. Then pivots back to NASA for Artemis IV.
It’s not a significant challenge compared to what they’ve already done.
Each of those previous tests could have easily gone to LEO running the engines just a tiny bit longer.
OPs point is that they intentionally didn’t.
achieving LEO means you need a relight to have a controlled reentry. You don’t want that if you want to avoid countries being mad at you while you iron out those controls
> It’s not a significant challenge compared to what they’ve already done
I don't know an aersospace engineer, within SpaceX or without, who would agree. When you increase speeds you increase energies faster. That has an effect on everything from pump performance to re-entry physics.
> Each of those previous tests could have easily gone to LEO running the engines just a tiny bit longer
Which risks recovery. Given they were replacing their Raptors in the next refresh, pushing an already-obsolete engine for shits and giggles doesn't make sense when you can get good data on e.g. skin performance.
> achieving LEO means you need a relight to have a controlled reentry. You don’t want that if you want to avoid countries being mad at you while you iron out those control
There is zero indication diplomatic pressure has been a constraint on the U.S. space programmes in the last couple years.
They didn't have to increase speeds, they already achieved orbital velocity. To circularize all they need to do is relight. Relighting an engine is very difficult for an engine like Raptor, but they've already demonstrated relight.
> They didn't have to increase speeds, they already achieved orbital velocity
My undertstanding is Starship didn't hit 17,000 mph [1]. LEO orbits tend to be 17,500 mph and up.
Like, I'm not arguing that SpaceX couldn't have circularised on previous tests. But it would have added material risk without any reward. And taking a ship, particularly a re-usable one, particularly a novel one, into its first orbital flight is always exhausting and novel.
It is like a runway taxi test on a plane that is fully capable of flight. Sometimes the plane takes off unexpectedly but the plan is not to do it. Starship can do orbital insertion now despite no plan to do it yet.
Odd. As a side note, your comment was posted [dead]. I vouched it to restore it back to life.
This is the second time I’ve seen such insta-dead comments. (One was my own, and I thought I did something wrong. Now it looks like there’s some kind of bug in HN that’s killing on-topic comments when they’re posted.)
Your comment wasn’t deep or insightful, but not every comment should be. A simple rejection of a premise is certainly on-topic. So it’s hard to argue that your comment was “bad”. That narrows the possibilities down to a bug in the algorithm. Maybe the mods are experimenting with ML auto classifying whether new comments should be killed or not.
Nothing. Now that I’ve seen it once for me and once for you, both on comments that seemed lightweight-but-harmless, I’m convinced there’s some sort of bug. So don’t take it personally.
Also HN != YC. They’re separate organizations, iirc. When Sam Altman was running YC one of the first things he did was “refactor” HN so that it has editorial independence.
Either way, it would be hard to imagine someone from YC telling Dan “you should boost so-and-so” and him going along with it unless it directly benefitted the HN community.
I guess I was a little distracted by the tangent to starship over the orion/Artemis
I was disappointed to see that after all these years NASA trying the old trick again and hoping people get excited.
As for spaceX and starship, I haven't kept up with it but I trust it's still putting NASA to shame wrt setting the state of the art.
Their objectives keep shifting and starship is far behind schedule. Sure, it's a success if you keep objectives small. They could have tried for LEO ages ago but didn't. Each launch should maximize learning and having small objectives is anathema to that. And very wasteful.
If you think Starship is behind, look at the 'competition'.
Learnings per flight may not be maximal, but they are measured with enough risk so that bureaucrats will approve it (not restrict future launches) and other countries won't be impacted by a failure.
What would going into LEO have taught them? They have been there hundreds of times.
They don’t have small connectives, or was catching the Super Heavy booster and then reusing it too small for you? Not everything they are doing is public.
How do they hope to make prop transfer work without a working heat shield to enable reuse of the tankers? Unless SpaceX pulls a hat trick, Starship is borderline useless.
They have a working heat shield (see last flight). It may not be quickly reusable, but that doesn’t matter at this stage.
For the transfer test, just left over fuel in two Starships is enough. They aren’t full blown finished tankers yet.
For HLS, if they are unable to get Starship reuse working in time, they can use expendable tankers.
There were 19 successful unmanned Dragon 1 missions before Crew Dragon, and an unmanned Crew Dragon mission before the first crewed one (actually two missions, but one didn't reenter from orbit). The heat shield material and design was essentially the same and so there was a great deal of flight heritage.
In particular I don't think its physically possible to test Orion components in flight very many times. It relies on SLS which chews through 4 space-shuttle engines every time, which even with unlimited money I don't think you could acquire a large supply of very quickly.
Not only that, but it has to reach much higher altitudes in order to also reach the much higher re-entry velocities that it will have IRL. That makes testing Orion very expensive. Testing Crew Dragon was much much cheaper.
SLS is required to get Orion to the moon, but there are other options for LEO tests. Exploration Flight Test-1 was performed on a Delta IV Heavy, and Falcon Heavy is also capable of launching Orion to LEO (and now New Glenn, although that wouldn't have been an option at the time NASA needed to start work on another Orion test).
That doesn't seem right to me. Sodium (and mercury) vapor lamps are the color they are due to physics, and were chosen because they're very efficient (and long lasting). Low-pressure sodium is the best and worst of these; essentially monochromatic but fantastic efficiency. Their only advantage, color-wise, is that the light can be filtered out easily (they used to be widely used in San Jose because Lick Observatory could filter out the 589 nm light).
The monochromatic light emitted from sodium lamps is also close to the peak sensitivity of the human eye. Colours are not distinguishable, but contrast is much enhanced compared to “cooler” light sources.
*edit: but it’s the overwhelmingly larger lifespan (20-30k hrs) that led to the wide adoption as streetlights. And I guess, the same is true for the change to led today, because of less power consumption.
It's not especially close to the peak sensitivity of the human eye (in either bright or dim conditions), but that's entirely okay. The goal should be to not affect people's level of dark adaptation.
If you use shorter ("bluer") wavelengths, as happens with white LEDs which consist of a blue LED + phosphor, it causes people's eyes to become bright adapted and effective night vision is ruined, causing people to have much worse vision in the shadows.
Also, if you use bluer light, the lights themselves cause dramatically more glare in peripheral vision, because the shorter-wavelength-sensitive "S" cone cells and rod cells are mostly absent from the fovea (center of the retina), and prevalent in the outer areas of the retina. This is why LED headlamps on cars are so obnoxious for drivers going the opposite direction.
Also, the LEDs clobber people's circadian rhythms and are extremely disruptive to wildlife.
Finally, the light pollution caused by the LEDs is much worse for seeing the stars, which is maybe not as important as the other harms, but still kind of sad.
> It's not especially close to the peak sensitivity
The sensitivity at sodium light is above 75% of the peak human vision (photopic) sensitivity.
This is a very small difference in light sensitivity. For example in the case of many sources of red light or blue light the sensitivity can be 5 to 10 times lower than the peak sensitivity.
Moreover, a perfect source of white light cannot achieve a better sensitivity than around 37%, i.e. less than half of the efficiency of an ideal source of monochromatic light at the sodium emission line.
Therefore the fact that currently LED lamps and low-pressure sodium lamps have about the same energy efficiency is caused by the LED lamps having a higher photonic efficiency and a lower threshold voltage (caused by a P-N junction voltage instead of the ionization potential of sodium), which compensate the disadvantage of using white light. A monochromatic LED lamp with the same color as the sodium lamps could have an energy efficiency at least double over the white LED lamps.
Red light would be even better for affecting the dark adaptation, but it has other disadvantages, like much worse energetic efficiency and lower visual resolution.
Yellow light a.k.a. amber light around the sodium emission line is a good compromise between energy efficiency, visual resolution and dark adaptation.
That's not necessarily a downside for traffic safety, though. Though I imagine someone must have studied the effects of various wavelengths on drivers...
There are 2 kinds of sodium lamps, low-pressure and high-pressure.
The low-pressure lamps emit monochromatic light and they have not only the advantage of long life but they are also the only other source of light that matches the energy efficiency of converting electrical energy to light of the LED lamps.
So replacing low-pressure lamps with LED lamps does not produce any significant economic effects, it was justified only by the supposed advantage of enabling color vision.
However in many places high-pressure sodium lamps have been preferred, which have a wider spectrum, so they allow some very poor color discrimination. The high-pressure lamps have a lower efficiency than LED lamps, so replacing them was justified by energy savings.
Outdoors at night, I prefer the monochromatic low-pressure sodium lamps, but sadly LED lamps have replaced them in most places.
In my area and esp. in the countryside they have green led lighting on various roads as an innovation, with the reasoning that is both least disturbing to wildlife, and best for human vision to see sharply. The light color takes some getting used to, but I am quite a fan of it. Esp. when cycling at home at night through the fields it makes things seem extra serene and peaceful.
Sodium lamps were deemed dangerous for driving” because they made it difficult for drivers to distinguish shapes, since they were different from day shapes. A kid in bright 1980ies colors (Little Red Hood) would look black under those lights.
LED was presented as a sharp improvement because of the large spectrum of white light.
The sodium lamps are in fact safer for driving, because they preserve drivers' night vision, which improves visibility into the shadows, and because they cause less glare.
What they aren't good for is LED manufacturers' bottom line, and the lighting industry spent a lot of lobbying money to entice friendly politicians to heavily subsidize them with public infrastructure budgets, with those subsidies then misleadingly sold to the public as "efficient" and "environmentally friendly".
They're also not very good for reading the newspaper or doing critical color analysis. Thankfully such tasks do not need to be done at night in the middle of the street.
That would make sense. Otherwise I have no idea how people wouldn't have noticed how much more difficult it makes seeing anything outside of the sharp cutoff of the light cone (or, of course, for the person being dazzled on the other side).
The power savings are minor btweeen LED and low presssure sodium lamps. The LED streetlights emit light along the full spectrum, the sodium lamps only at 589 nm. The LEDs are more controllable so smart dimming ( when there are no cars) is a perceived advantage.
Sodium (and mercury) vapor lamps may be the color they are due to physics, but you don't have to put Na or Hg in those tubes. I don't know but Li, K, Rb, Cs should, Mg, Ca, Sr, Ba probably also work, but nobody make lamps with those elements.
There are 4 important properties for the substance used in a gas-discharge lamp.
1. For a sufficient gas pressure in the lamp, the substance must be either a gas or a metal with low boiling temperature, so that it will be vaporized by an electrical discharge.
2. The gas must not react chemically with the enclosure and with the electrodes, which prevents the use of most gases except noble gases and metallic vapors. Except for noble gases and metallic vapors, the lamps using other substances must not have electrodes, so they need a more complex and less efficient electronic system for producing a high-frequency AC discharge, e.g. using a magnetron from microwave ovens.
3. The ionization potential must be low for a good energy efficiency. Alkaline metals have low ionization potentials and low boiling temperatures, so they are better than noble gases and other metals.
4. The color of the light must be one where the sensitivity and the visual acuity are high. This narrows the choice to yellow light, i.e. to sodium, between the alkaline metals.
The lamps that use alkaline metal vapor instead of a noble gas have better energy efficiency, because of a lower ionization potential, which leads to a lower voltage drop on the lamp. Therefore they have been preferred for lighting instead of neon lamps and the like.
Among alkaline metals, sodium is the cheapest, so it was a logical choice.
However, the fact that it produces light of a suitable color was a happy coincidence. If sodium had produced violet light, like potassium, and potassium had produced yellow light, potassium would have been chosen for lamps.
So among the criteria for choosing sodium for lamps, the color of the light was as important as cost, ionization potential and vapor pressure.
Two things can be true. And often that's precisely when we lose out with modern engineering that is much more single-minded.
> Their only advantage...
How are you coming to this conclusion?! Their warmer has very meaningful effects on processing, attention and other visual effects as is the point of the discussion in the first place. It's not clear what makes you so sure that color differentiation is essential and the other effects are irrelevant.
No I absolutely don't know what matters. But it seems neither do you.
...And the old Engineer was just saying that that was the area on the spectrum they aimed for, so they found a light that emitted in that wavelength that could be technically implemented and scaled.
Way better work than whoever it is handling this LED nonsense. Why we can't find a diode that we can use to simulate the old spectra would be a fun research project.
We of course can make LEDs of more or less any color. The current white LEDs are high-power blue LEDs that are covered by various phosphors to give a mix of colors for "full spectrum" illumination. Different color temperatures are produced by different mixes of phosphors. This is pretty similar to how the traditional luminescent (mercury vapor-based) lamps worked.
But different phosphors have different efficiency and price. LED lamps were first introduced for interior lighting, where sun-like spectrum is welcome. Such LEDs were produced en masse and relatively cheaply. So street lighting naturally used them, because municipalities usually look for the cheapest viable option.
We likely could produce high-power narrow-spectrum orange LEDs if there was a large market for the economies of scale to kick in. You can buy deep orange LED lamps today (look for color temperature 1800K or 1600K, "amber"), but they are more expensive, because they are niche.
> Different color temperatures are produced by different mixes of phosphors.
We can make LED light appear to be any given colour by mixing multiple LEDs. But mixed colour isn't the same as pure colour, made from a single spectra of light. Nor is it the same as true broad spectrum light - like we get from black-body radiation like the sun, or a tungsten bulb.
Its hard to tell the difference just by looking at a light. But different kinds of lights - even lights which look the same colour - will change what objects actually look like. And they probably have different effects on our sleep cycle and our low light vision. I was in a room once lit only by sodium vapour lights. The lights were yellow, but everything in the room (including me) appeared to be in greyscale. It was uncanny.
This is part of the reason why LED lights are still looked down on by a lot of old school photographers and film makers. Skin doesn't look as good under cheap LED lights.
For light with a narrow spectrum, it is possible to make LEDs that emit that light with high-efficiency, for any color inside 2 ranges, one from near infrared to yellow (corresponding to semiconductor phosphides and arsenides) and one from blue to near ultraviolet (corresponding to semiconductor nitrides).
Only green LEDs have worse efficiency, because they must be made with semiconductors for which optimum efficiency is attained at either lower or higher light frequencies.
Lamps using high-efficiency amber LEDs with about the same color with sodium lamps could be made at an energetic efficiency at least double to that of white LED lamps.
The double factor comes from the visual sensitivity being double for the light at sodium color than for ideal white light.
In reality the energetic efficiency of such LED lamps should be more than double, because they do not have losses caused by conversion through fluorescence.
seafoam green choice was also influenced by eye rest studies... since our eyes are most sensitive to middle wavelengths. just keep the room dimmer without losing detail. It reduces fatigue for operators on long shifts.
When I was an medical intern back in the day and worked 24 hour shifts every third day, I bought a roll of thick black vinyl and taped it to the window frame. 0.0 light got through.
If running twice is good, then is running N times even better? I wonder if you could even loop until some kind of convergence, say hitting a fixed point (input equals output). I wonder if there's even a sort of bifurcation property where it sometimes loops A->A->A, but other times A->B->A, or more, rather like the logistic map fractal.
I explored that, again with Devstral, but the execution with 4 times the same circuit lead to less score on the tests.
I chat with the model to see if the thing was still working and seemed coherent to me, I didn't notice anything off.
I need to automate testing like that, where you pick the local maxima and then iterate over that picking layers to see if it's actually better, and then leave the thing running overnight
Can Karpathy's autoresearch be used on this to explore what works and what does not? That is supposed to automate research like this from what I understand.
Is it two, or is it infinite? The quaternions have three imaginary units, i, j, and k. They're distinct, and yet each of them could be used for the complex numbers and they'd work the same way. How would I know that "my" imaginary unit i is the same as some other person's i? Maybe theirs is j, or k, or something else entirely.
Getting an EKG seems very prudent. I had one done for a non-heart related procedure, and afterwards was basically asked:
- Ever have any heart events? Heart racing, palpitations, that kind of thing?
- Yes, a few times a year I've noticed events like that. Resolves in a few minutes, though.
- Well, your EKG shows a slurred delta wave. Sign of Wolff-Parkinson-White syndrome. Might want to get that checked out.
I did, and it was. Fixed with ablation. No issues since. Other types of supraventricular tachycardia can also be cured with ablation.
If you do both (use flipped parentheses around the operators), it makes even more sense, and makes the parsing trivial to boot: just surround the entire expression with parentheses and parse normally. For instance:
1 + 2 )( 3
Becomes
(1 + 2 )( 3)
Which is actually just what the author wants. You might even want multiple, or an arbitrary numbers of external parentheses. Say we want to give the divide the least precedence, the multiply the middle, and the add the most. We could do that like:
1 + 2 )/( 3 ))(( 4
Surround it with two sets of parens and you have:
((1 + 2 )/( 3 ))(( 4))
I haven't just proved to myself this always does what you expect, though...
Many people think of driving in time rather than distance. I'd say it's actually more common to say a city is 3 hours away rather than 200 miles.
What makes kW less useful is really just that most EVs don't advertise their capacity very prominently. But if you knew you had an 80 kWh battery and the car uses 20 kW at freeway speeds, then it's easy to see that it'll drive for 4 hours.
The problem with this is that destinations are a fixed distance away, whereas their time distance is not fixed. In most journeys people want to reach a specific place rather than drive for a given amount of time.
I understand all this but the most important question for me is definitely still "how much distance can I cover on a charge"? That's why I prefer kWh/100km.
We know the upper bound for most of those numbers. SpaceX already achieves internal marginal launch costs of ~$1000/kg, for instance. We know their rough costs per satellite. In contrast, we know little to nothing about the inputs to the Drake equation.
The numbers don't quite work out in favor of orbital datacenters at the current values. But we can tell from analyses like this what has to change to get there.
I wouldn't downplay the opportunity cost of that much human capital. It really is quite a lot, given the obvious talents of the physicists.
I'm not saying I fully agree with the position, but one way of looking at it is that thousands of incredibly smart people got nerd-sniped into working on a problem that actually has no solution. I sometimes wonder if there will ever be a point where people give up on it, as opposed to pursuing a field that bears some mathematical fruit, always with some future promise, but contributes nothing to physics.
There is almost no opportunity cost: The academic pyramid swaps out the lower parts of the hierarchy at a high pace. You might lose a few smart people who become professors but the average sting theory phd goes to finance or whatever field requires absurd amounts of math at the moment.
You do get people who are happy for a few years since they can live their childhood dream of being a physicist before the turn to actual jobs.
Having people work on things that are at the limit of human understanding is an essential part of a modern educational system.
For every professional string theorist, you get hundreds of people who were brought up in an academic system that values rigor and depth of scientific thinking.
That's literally what a modern technological economy is built on.
Getting useful novel results out of this is almost a lucky side effect.
Starship v3 flying will be a significant leap, though. It's the first with the Raptor v3 engines and has many other improvements as well, such as updated grid fins and hot staging ring. It will be the first that achieves close to the intended capacity of ~100 tons.
Propellant transfer is indeed a significant challenge. They have already demonstrated internal transfers between tanks, but not between spacecraft.
Very exciting times ahead!