that's confusing "subscribers" and "paid subscribers".
from Zitron's own website [0]:
> I have 84,000 subscribers and a 55-60% open rate, as well as an 8-11% clickthrough rate.
if the 84k number was all paid subscribers, then the "55-60% open rate" would mean that ~40% of his paying audience doesn't bother to read the thing they're paying for, which does not add up.
also, that's in the "Can I Advertise On Your Newsletter?" section. if there were an even higher number of non-paying subscribers, he'd have an obvious incentive to mention that, because the total number of eyeballs is what a potential advertiser cares about.
I find that completely unbelievable. There’s too much high-quality free information for me to ever consume to even think about paying for it, let alone something of mediocre quality. Are we sure the newsletter subscribership isn’t just a total fabrication? Or that it isn’t just a money laundering scheme?
Also, while we're pitching new features to openrouter, I'd like to see a "$ spent" chart, which would remove all these huge freebie spikes. It looks like it would be pretty much dominated by claude.
Of all the incompleteness-style theorems, I find the Halting problem to be the most approachable and also the most interesting. Maybe it's because I'm a software dev that dabbles in math rather than the other way around. But that makes me wonder if all of Gödel's theorems can be stated if 'software form', so to speak.
Right, if you're a software engineer, the realization that the two theorems are nearly-equivalent really takes the air out of a lot of the existential philosophizing around Gödel's incompleteness.
Gödel's argument basically says that any system of mathematics powerful enough to implement basic arithmetic is a computer. This shouldn't be surprising to software engineers because the equivalency between Boolean logic and arithmetic is easy to show. And if you have a computer, you can build algorithms whose outcome can't be programmatically decided by other algorithms.
I think that's selling the theorems a little short. A math system with arithmetic is equal to, or more powerful than, a computer. For an example, even classical logic comes with the law of excluded middle that can say (internally) if a program halts or not. Incompleteness applies to all the stronger systems as well.
There is no logic that is more expressive than a Turing machine. In fact, just about every logic you know can only expressive necessarily terminating programs. There is a bit of an issue on what exactly someone means by expressive, but if we're talking programs that compute outputs from inputs (without caring about the invariants imposed on said programs) then this holds.
It's convenient that Henry Rice lived long before the age of language cults. I don't even think Rice wrote software, he's just a mathematician, he proved this nice property in mathematics. Stuff like FORTRAN and ALGOL happens later.
Also though, just as for the Halting Problem, we are always allowed a three-way split. Rice proves that "Has property" vs "Does not have property" can't be done, but "Has property" vs "Does not have property" vs "Shrug - I dunno, seems hard" is possible, and indeed easy if you're OK with lots of machines landing in the "Shrug" pile. You can expend as much work as you like to shrink that pile, Rice just proved it would need infinite work to empty it completely.
Another way around the Rice's theorem is the Curry-Howard correspondence. A constructive proof of existence of a program that has a property can be transformed into a program that has this property. Yet another way is to have a programming language where syntactic correctness implies a range of semantic properties.
The undecidability of the halting problem yields an easy proof of Gödel's "zeroth" incompleteness theorem:
Statement: Every sound (i.e. not just consistent, but sound) recursive theory of arithmetic is incomplete.
Proof: Assume it is complete. List all its theorems by a program. Then one can decide the halting problem as follows: for any instance, look whether "the program halts" or "the program does not halt" shows up in the list of theorems (since the theory is complete, one of them must show up; and since the theory is sound, the theorem is true).
This also makes it obvious that at some point, the halting problem becomes "unprovably hard." There must be Turing Machines for which it is independent of the accepted axioms of mathematics whether or not they halt. And indeed, constructing such machines is not too difficult.
Sure but that's fairly pedantic. You can derive Godel's first incompleteness theorems strictly as a consequence of undecidability of the halting problem.
You could say they have a sort of anti-moat (drawbridge?) since you can use their product to create a competitor. But that's true of most dev tools, in a sense.
I dunno if it's that clear cut. In space with a shadowless orbit you get 5x more solar energy per day than the sunniest place on earth. And it's always on, so you don't need batteries. Also, the lack of gravity and weather means that the structures can be a lot more brittle - I imagine something like a gpu on the back of a large thin film solar panel, where the panel also acts as heatsink. Could be pretty cheap!
So you are comparing putting up slightly more solar panels and some batteries, both things that are mass produced commodity with re-engineering the whole computing ecosystem and how data-centers have been done for decades, transporting them on the most expensive known transportation system and then operating it in a way where you need specially educated experts.
So my bet would be that just ordering a bunch of sodium batteries and a bunch of solar panels (or you know using a source of energy that is constant) is cheaper then going threw all the effort of putting things into space.
I have been following everything space for decades, and not once have I thought, wow putting super complex engineering things into space so easy and you should do it for things that have an easy alternative on earth.
while there may not be atmospheric weather, low earth orbit has its own "weather". Before you even reach LEO you start getting bombarded by all forms of energetic particles. None of these are things you want your computers saturated with
Nobody (sane) is talking about putting nuclear reactors on Satellites in close Earth orbit so we don't have to worry about them generating heat. They've got solar panels that move some of the solar energy they absorb to a central location which presents problems in moving the waste heat back out so that spot doesn't get too hot. But that doesn't change the overall equilibrium temperature.
I'm not sure if you're being serious or not? Any use of power turns useful work into heat (conservation of energy and all that), which raises the temperature of the satellite, until radiative cooling can equalize within incoming heat (solar irradiation).
You’re ignoring where the power comes from. Unless you take it up there with you (eg. nuclear material) it has to come from the sun, so you can’t use or emit more energy than the sun hit you with. You can move it around in space and time, but you can’t get on average hotter than a lump of rock in the same spot.
I didn't think of it like that. Does that mean all solar radiation is heat from when it hits a solar panel? I thought it would be something like solar -> chemical -> electrical -> heat.
You're concentrating it into a very small area of compute.
If you don't spread that heat back out, it's going to find a much higher thermal equilibrium than the solar panels themselves would find just absorbing the sunlight and radiating the energy back into space.
It's like you've pointed a magnifying glass at your compute, except with electricity, which means you can reach temperatures higher than you can with a magnifying glass.
I guess I'm curious: all the comments I see about this act as if the people proposing putting data centers in space are complete idiots. Do you believe they are complete idiots?
They're hucksters who know that adding "in space!" to a sales pitch is a free booster for tech enthusiasts.
It's the same way that Sam Altman talks about the risks of AI deciding to kill humanity: because that's dramatic and attention grabbing, and also the most unlikely outcome. Talking about it keeps us from talking about the real, ground level problems like the massive, unplanned-for disruption in jobs and education.
They just need to keep the money tap flowing, and tomorrow can worry about itself. Who's going to hold them accountable for data-centres-in-space five years from now, when they don't exist? Has Musk suffered any blowback from his hyping the Hyperloop that never materialized?
How many smart people worked quietly on Zuck's metaverse for years? How many knew it was never going to work at some point on the line to $70 billion wasted, but thought "hey, maybe I'm wrong, and it's an interesting job that pays well"?
How many smart people worked quietly at Theranos, knowing that a drop sample from a thumb was incapable of carrying sufficient blood volume for a legitimate sample, but thought "hey, maybe someone will figure something miraculous out that violates a basic tenet of my professional experience"?
Are there engineering reasons why the Metaverse wouldn't work? I thought it was more about the actual reality of the product, even in its perfect engineered form, still not being that appealing. Or most charitably being ahead of its time.
True, the Metaverse was a practical product failure, rather than an impossible-in-principle failure. Regardless, a lot of smart people worked a long time trying to make it work, and it was pretty obvious it wouldn't once Zuck demo'ed it and everyone saw a creepy cartoon world, which was all the bandwidth and compute at the time could support.
Grandparent is trying to argue that a lot of smart people working quietly on something confers plausibility upon the premise of their work (i.e., "they must know something you don't"). I've rebutted with two examples showing that large numbers of smart people working on something don't make plausible a premise that is obviously flawed for other reasons [*].
[*] (ETA) and is known at the time by the smart people.
I mean, if someone wants to pay you to do a lot of very interesting R&D that will never result in an alternative to ground-based datacenters, more power to you?
It might even be useful in other circumstances. Better radiative cooling systems, hardening commercial high-end compute for space, etc etc. R&D you can feel proud of, even if your bosses are only paying you to do it to fleece rubes who think it's the next trillion dollar industry.
If they’re actually serious about this, they could simply address the points about cooling that numerous experts have raised. But they haven’t done that, at least not that I’ve seen. I have no idea whether they’re complete idiots, and I don’t really care. Maybe it’s idiocy, maybe it’s hubris, maybe it’s a grift, I have no idea. But until I see a compelling solution to this known problem, or a compelling suggestion as to why they’re not sharing a solution, I’ll continue to think they aren’t particularly smart or serious about this.
What if it's actually not that hard to cool something in space, and y'all just have these beliefs about the people talking about this that make you think there must be something obvious they aren't thinking about?
The problem is, that person is deeply underinformed. For instance:
"you don't lose that much power through the atmosphere"
Assuming you can point the orbiting panels at the sun and remain direct, you lose 80% of your power through the atmosphere (from angle and day/night). On the ground you lose 25% even if your panel is directly below the sun pointed exactly at it, which is never the case in many latitudes.
And of course it's not trivial to radiate heat. But it's also a fairly simple mechanical problem. You pump the heat to spread it out, and radiate it. You've already got the surface area shaded by the panels (which is more than enough, because the panels don't absorb 100% of solar radiation).
Sure, you need a lot of them. Starship V3 is probably about to get us past 100 tons of payload capacity - even if they blow up a few first.
The key people miss is that you don't have to spend money on ongoing cooling once the thing is in space. This isn't going to save money now, but the cost lines are going to cross.
> Assuming you can point the orbiting panels at the sun and remain direct, you lose 80% of your power through the atmosphere (from angle and day/night). On the ground you lose 25% even if your panel is directly below the sun pointed exactly at it, which is never the case in many latitudes.
When people say "you don't lose that much power on the ground", it's in the context of cost.
So a solar panel is 5x more efficient in space. You solve that in the ground by buying 5x more solar panels.
Solar panels cost -- rounding up -- $10/kg.
Lifting things into space costs -- rounding down -- $1000/kg.
For the same amount of money, you can put 100X solar panels into a ground based array as a space based array. You don't lose that much power on the ground. You aren't overcoming that difference because solar panels are more efficient in space.
It's frustrating to talk to someone about this and get "they need to address these points" (which frankly, they have addressed) and then get "this will happen of course" when I point out that it will in fact happen. It feels like moving the goalposts to avoid saying "thanks for informing me of something new."
It's frustrating to try to have a discussion when any source contrary to the desired position is met with "this person is underinformed"; it's why I don't really bother with any "provide me sources or I won't believe you" on the internet. People don't believe you anyway, so why bother?
If you gather 1kW of power from the sun then you have to reject 1kW of heat once you are done with whatever computation you are doing. There’s a bit more heat absorbed from the environment since some sunlight strikes parts of your satellite that are not solar panels, but it’s not too bad. Starlink satellites, just to pick a relevant example, do not need a radiator at all because they stay mostly edge–on to the sun and they can radiate all the heat through their own surface area. The ISS needs big radiators because they want it to be comfortable for humans, but electronics can run significantly hotter than that.
The joint solar panel + computer system will be pretty close to an ideal black body, which near earth will have an average temperature of about 10°C. And radiation is an issue, but starlink seems to work so I don't see why this wouldn't.
Of course it works, the question is how this would look like and if its financial feasable.
You make a H100, ship it to a space dock, load it onto a rocket (rocket requires fuuel, the rocket, etc.) send it up, deploy it, monitor it live 24/7, have means of adjusting its orbit, if it breaks, its immediade full loss, otherwise it will degenerate faster in space than on earth, now it needs a high speed up/downlink to do anything reasonable which also requires a base station. The base station has to track this satelite.
One H100 costs 40k, consumes 700 Watt peak and need probably at a minimum 5 square meter of area for cooling and solar.
The colossus datacenter from musk has 250.000 of these.
Now you have to track 250.000 single satelites, you have to coordinate the communication between the, up and downlink to earth.
250.000 * 5 square meter of area.
This alone increases the potential debris in space.
And this is ONE 300 MW Datacenter replacement. ONE.
It’s very easy to overestimate the difficulty of cooling things in space, unless you actually run the numbers. So please follow along as Scott Manley runs the numbers: <https://www.youtube.com/watch?v=FlQYU3m1e80>.
Basically a Starlink v3 satellite has an estimated power budget of 20kW. Add in the heat absorbed from the environment (both directly from sunlight and reflected off of the Earth) and you’ll find that it must reject about 22kW of heat. That’s a fair amount, but at 65°C it can radiate it all away just using it’s own surface area! No radiator required at all!
Of course the power density of computer racks has been going up over the years. If you want to reach 100kW per satellite then they will need a modest radiator, but nothing extravagant. It would still be smaller than the solar panels, and far smaller than the ones on the ISS. And don’t forget that because radiated heat goes up as the fourth power of temperature, raising the temperature of the system by even a small amount raises the radiation emitted by a lot. If you design the system to run hotter you can minimize the size of the radiator. Most chips these days are designed to max out at 100°C to 110°C without lasting damage, although running them at that temperature 24/7 may reduce their lifespan. There will be some sweet spot in the middle.
And it turns out that a Starlink v3 already has a volume somewhat larger than a 48U rack. You talk about launching 250k satellites in order to have 250k GPUs in orbit, but that’s ridiculous. A real compute swarm will be hundreds or thousands of satellites each equivalent to a whole rack of GPUs.
But you’re not wrong to be skeptical. The economics might not work out even if the cooling is easy enough. It’s just that rejecting the idea takes a lot more than back–of–the–envelope calculations.
What doesn't make sense to me here is that even on Earth, where we have an atmosphere to disperse heat into, we find that closed-loop cooling is too expensive and so use evaporative cooling.
If the economics make it too expensive not to use freshwater on Earth, I don't see how closed-loop cooling suddenly becomes affordable in space where dispersing heat is already more difficult.
Here in datacenters we use cooling systems to move heat away from the computers and out of the building. If each rack in the data center were outdoors and 100 yards away from every other then almost no cooling would be required. Just some fans to suck in air at the bottom and eject it out the top. Even less would be required if the individual GPUs were somehow separated from each other. Then we would truly be dispersing heat into the atmosphere with no cooling system at all.
Similarly, a satellite only needs a cooling system so that it can move the heat from the internal components outward towards the hull. A satellite containing a rack’s worth of GPUs might literally have heat spreaders that touch the chips on one side and the outer hull of the satellite on the other. Combine that with some heat pipes or something to spread the heat out efficiently and you hardly need anything else. A satellite the size and shape of a Starlink v3 already has enough surface area to dissipate something like 28kW at 80°C, and more if you run it hotter. If you want more than ~30–40 GPUs per satellite then you might need a small radiator to increase the surface area, or you might just make the thing thinner and wider instead. You’ll need more solar panel area anyway, so making the bus wider to match the wider solar panels is fine. The “closed–loop cooling system” you say is so unaffordable might be no more than a bunch of heat pipes. Or it might be an aquarium motor that pumps a few kilos of ammonia through some pipes or channels in the hull of the satellite.
I'm not rejecting the basic idea in itself. There is nothing in this idea which we as humans can't do today. No issues here. Its just so much more expensive than just doing it in a dessert and putting fibre and solar panels and batteries there.
The Starlink v3 doesn't exist yet in space, it also needs Starship apparently and Musk said it will have the size of a Boeing 737 fully deployed. So it will not be small and its not proofen yet.
A rack with 48u will either have 12 or 24 GPUs which equals to 9kW or 17kW. Than its not 250k satellites for a 'small' 300MW DC but only 25k. Still a very crazy number.
I would love to see all of this scifi stuff happening. Spaceship in space, travel gates, dyson sphere but there is just no current breakthrough in our society which would indicate that this makes sense.
In my opinion, we as a society will have to get rid of capitalism first before we will do the next step and just because Musk needs a story to sell to keep his construct alive, doesn't mean its the right time.
> The Starlink v3 doesn't exist yet in space, it also needs Starship apparently
True but not really relevant. All prior versions of Starlink worked well enough, so there’s no reason to suspect that v3 won’t.
> … it will have the size of a Boeing 737 fully deployed. So it will not be small…
Irrelevant. Size is a weird measure here. You should ignore it because it is a marketing thing. All that is meant by it is that once the solar panels are extended they span a distance larger than the wingspan of a 737, nothing more. A Starlink v3 satellite has far less mass, interior volume, or complexity than a 737. I could get a long 30m rope and tell you that it was “larger than a 737”, but you wouldn’t be very impressed.
> A rack with 48u will either have 12 or 24 GPUs which equals to 9kW or 17kW. Than its not 250k satellites for a 'small' 300MW DC but only 25k. Still a very crazy number.
Are you sure? That fits within the estimated power budget of a Starlink v3, but I was assuming that GPUs were denser than that these days. I’m not an expert though. I figured a rack would hold somewhere between 96 and 128 GPUs depending on whether they had to be in 3u or 4u servers. They would need between 60kw and 90kW of power, and would need a modest radiator. The solar panels would be far larger than the radiator.
> I would love to see all of this scifi stuff happening. Spaceship in space, travel gates, dyson sphere but there is just no current breakthrough in our society which would indicate that this makes sense.
Ok, see, your problem is that you haven’t properly distinguished between different types of fiction. You put those three things in one category as if they were all equally fictional, but that cannot be true.
The first one is ambiguous, since the Space Shuttle was definitely a “spaceship in space”, so perhaps you just mean FTL travel like in Star Trek or Star Wars. By “travel gates” I assume you mean something like the eponymous gates from Stargate SG–1. Those are both ruled out by the laws of physics, and we can only tell stories about them because people willingly suspend their disbelief. Campbell said that a true science fiction story can only include one thing that requires the reader to suspend their disbelief. If it includes more than that then it is a fantasy instead.
But a Dyson sphere, or more accurately a Dyson swarm, is not an impossibility at all. If I can put one solar–powered satellite in orbit around the sun then if I am really industrious I could put ten, or a hundred, or a trillion. As many as I wanted and could afford, right? The laws of physics don’t say that a sun can only have 8 satellites around it, or any other number. Dyson knew that enough such satellites would eventually blot out the sun. They would absorb all the sunlight it could emit, and then the satellites would all emit infrared waste heat. If all of those satellites were doing something useful then whoever put them there would have a lot of useful work being done at their command. Even if it’s just trillions of GPUs making AI–powered cat memes, or simulating the minds of a bunch of human uploads, or even if they’re all just mirrors to redirect that light somewhere else, that’s a lot of power at our command. It is fictional only because it is an idea that nobody has actually gotten around to implementing yet. Once we’ve done it then it won't be fictional anymore.
> Its just so much more expensive than just doing it in a dessert and putting fibre and solar panels and batteries there.
No one here is arguing that it isn’t. It might still be cheaper to build datacenters on Earth. But most of the people who say that it’s _definitely_ still cheaper to build them on Earth are overestimating the difficulties, and therefore the costs, of doing it in orbit instead. We’re getting to the point where launch costs are low enough that it is no longer a given that it is cheaper.
For one thing, actually building things in the desert is expensive. You have to build all the necessary infrastructure yourself. Roads, power, water, fuel, etc, etc. You might as well be launching everything into space! No, if you want it to be cheaper you need to build somewhere closer to home, like Ohio.
Everything you wrote is some definition of hard, but all doable. None of this is purely in the territory of 'known' impossible(like FTL travel).
Now different people have different points where they quit when things get hard.
This is true for even everyday things in life. Quitting triggers exist for people at various points in the ladder. The end of ladder and path both exist, its upto you to decide if you wish to continue climbing, or give up and quit.
My problem is not the doing thing but the economy of it.
We are nowere near any resource limitation on planet earth for AI Datacenters.
Musk sells this story because he has Starship which needs payload to make financial sense. The payload doesn't exist so he inventes DC in Space.
Its the same thing as SpaceX buying Tesla Cybertrucks.
His old colossus datacenter is a 300MW Datacenter he now rents out to Anthropic because he doesn't even need his own compute. Colossus DC is probably 10x cheaper than his whole Space AI DC Story and will be for a long time.
> Everything you wrote is some definition of hard, but all doable
The first line of the post that you are supposedly replying to is:
> Of course it works, the question is how this would look like and if its financial feasable.
Unless is cost-comparable to a data centre on Earth, and I am told that it very much is not, then there is no financial feasibility for space datacentres.
More energy will be required than radiation absorbed by a spherical (ish) data center. You'll have massive solar panels piping energy in, and so the temperature would by higher than thermal equilibrium at that distance.
Well I was talking about heat. But regarding radiation, there is a long history of transistors in space dealing with radiation. But ... there is also a whole science how to deal with making it reliable: answer, expensive redundancy.
And about starlink .. as far as I know the fail quite often but work, because of redundancy. So they get replaced.
If you want to ship GPU's to the orbit, then this surely works somehow, if you are willing to replace them often, which is expensive. Or you shield them, but then you will need to get up heavy shields. In general, of course computers work in space, but it is not cheap.
Its not always on. Its only 'always' on if you would orbit the sun which starlink can't do, it has to orbit the earth. This only works in a certain constelation which would create a halo around our planet, without clear understanding what even would do.
The more power you consume, the more power you need to dissipate. These constelations wouldn't be small at all. It would also take a interesting solution to be able to move this heat from very small very intense areas to very big cooling areas. How?
And space is not easy. Space is very very cold which puts a lot of stress on materials. It has radiation. And it has A LOT of microasteroids. Stuff in Space breaks down due to this. You would need to replace all of this stuff regularly with resources from the planet earth.
You would basically just spend a lot of resources throwing a lot of resources out into space. You can't even recycle all of this.
Its still lunatic at our current state of our current system. There is so so much space on our planet. Its ridicoulous
The only reason Musk is saying stuff like this is because he knows there is no market and he needs to keep his system alive
The always on orbit exists and is called a dawn-dusk Sun synchronous orbit. It is an orbit that is always above the terminator (line between night and day) where it can face the Sun 100% of the time.
This orbit has to rotate about a degree every day to follow the terminator as the earth orbits the Sun. It uses the equatorial bulge of the earth to achieve that rotation without have to spend rocket fuel. It is really quite interesting.
Yes they are crowded. There are engineering solutions to that, including formation flying based on absolute positioning and space tethers. It is a great time to be an engineer.
A polar low earth orbit can be always-on (no earth shadow). Each satellite will be in thermal equilibrium, around 10°C. Catastrophic destruction from micrometeoroids is rare. I'm not saying it's a good idea, but I don't see any dealbreakers in the math/science.
Kessler Syndrome is the biggest dealbreaker. We're already fairly far advanced in that scenario from Starlink, and competitors/scaleouts to Starlink promise to be worse.
If you plug eleventy trillion dollars of hope that the aristos can finally replace the working class into the issue, Earth loses access to low orbit from orbital debris almost immediately.
Their entire mindset cannot deal with this. Low orbit is a physically-enforced type of commons, inextricably tied to tragedy if overpopulated. You cannot privatize it and scale indefinitely. There is no defense, and any pissed off individual actor who gets malicious can burn it to the ground.
Starlinks are in low enough orbit to passively decay in less than 5 years, that really can't meaningfully contribute to a Kessler syndrome.
Chinese mega constellations on higher orbits & their spent stages left in space are a bigger issues.
Still in case it got going & made higher orbits unusable, starlink would likely still work just fine on the lower self-cleaning orbits, not to mention using a partial (and hopefully soon full) RLV for replenishment.
A recent paper came out calculating that it would take two days of lights out at SpaceX headquarters for the whole constellation to shred itself, it was already so reliant on avoidance maneuvers.
SpaceX immediately responded by lowering its target orbits by 70km, the maximum it could legally do without renegotiating formally.
When a high orbit develops Kessler Syndrome, the billions of pieces of debris rain down on lower orbits and cause cascading collisions there, and they keep doing it for centuries.
Not understanding how any of this works, the scientists not being capable of convincing the politicos, or the leaders not being able to escape their local maxima of public stances to recognize a real threat, is a massive, civilizational level hubris. This is pass/fail - the math does not care about our level of understanding or maturity.
If COVID taught us anything, it's that logarithmic growth functions don't get no respect. They are so alien to so much of the population that you can't get real phenomena taken seriously, no matter how many grains of rice you pile on chessboards.
Every collision increases the chance of additional collisions.
"Weeks"? SpaceX is aiming to be within the guideline of 25 years for the intact spacecraft, with solar panels unfurled. The want at minimum five years of business operations out of it. Half a spacecraft is still going to be decades (if not centuries if they lose the panels), and small chunks of spacecraft years.
5 years is still 5 years and Musk needs A LOT of payload for SpaceX to justify 1 Trillion dollars.
This 1 Trillion Dollar has to be translated to either sending up A LOT of foreign payload OR his payload; All of this payload = new Satelites. Its not like we are sending earth resource up in space to build a dyson sphere.
They are large language models. Not automated development machines. They hallucinate.
The goal post has not shifted since 2023 or so. Make an LLM that doesn't blatantly disregard knowledge it has, instructions it has been giving, over and over, and you win. If trillions of USD of investment can't do it, I'd be curious to see what can.
There are definitely automated dev systems, of which an LLM is a part. The remaining part may be called a 'harness' or whatever. The quality of the generated software is another matter.
If the AI is not good enough, then don't fire the devs. If/when the devs are no longer needed, I don't see why the need would return later, that was my point.
A harness like Claude Code does not turn an LLM into a software developer.
If that was the case companies could just have their project managers managing Claude Code instead of developers, and they would immediately realize that using Claude Code to develop software is just as complex and geeky as it ever was - nothing changed in that regard.
A harness and a bunch of skills is just the new "think step by step" prompting technique. Don't just let the LLM rip and write a bunch of code, but try to get it to think before coding, avoid things like churning the code base for no reason, and generally try to prompt it to behave more like a developer not an LLM. Except it still is an LLM.
A coding agent is really not much different to a chat "agent" in this regard. You've got the base LLM then a system prompt trying to steer it to behave in a certain way, always suggest "next step", keep to a consistent persona, etc. None of this actually makes the LLM any smarter or turns it into a brilliant conversationalist, anymore than the coding agent giving the LLM a system prompt magically turns it into a software developer.
If you prefer staying in denial, be my guest. But I've seen multiple instances of fully functioning software created by people who don't even know what code is. Maybe these creators are now developers, in a sense. But no SWE's were needed.
Sure, and I could buy a model rocket engine, strap it to a stick and launch it hundreds of feet into the air. Would that make me a rocket scientist? Next step Mars?
If you don't appreciate the difference between what an LLM or a coding agent can do, vs what a human can do, then I can't help you.
80k * $7/month * 12 months/year = $6.7M/year
[1] https://www.theguardian.com/technology/2026/jan/19/ed-zitron...
reply