DeepSeek Elastic Compute (DSec) | Hacker News Reader

DeepSeek Elastic Compute (DSec)

arxiv.org

323 pointsshenli3514115 comments

It seems every DeepSeek paper/patent has a huge number of authors, and this one is no exception. They couldn't even fit everyone on the page, there are 31 others not shown. This could be an asset protection strategy (i.e., human assets). Imagine if there were only 3 authors. Those authors may get hired away by competitors. If you list every employee on every paper then competitors don't know who to lure away.

Just look for the “corresponding author”

The unfortunate intern who had to upload to arxiv?

Corresponding author is the head of the group, usually the one who coordinates the study, the one who knows everyone else. Is the "if you have any question about anything in this paper, contact me" person, even years after publication.

If it’s a Chinese institution, the first author will be the principle, not necessarily the corresponding author.

click the link, or read the PDF.

all authors are listed. there is no conspiracy to hide authorship.

arxiv simply want to keep the page short not too long.

Compare to https://arxiv.org/abs/2407.21783

and this: https://arxiv.org/abs/2503.19786

I'm available if anyone wants to use my name on an AI paper for misdirection.

I'm also available for patents and trust funds too

Lmao with the conspiracy. Large scale experimental research is always like this, many papers in experimental physics have pages of authors.

Competitors will try to touch everyone on the list.

Yeah what possible company can try to lure 100 people... Just for reference, Linkedin has 17,000+ full time employees.

Maybe aquihire like Nvidia with groq?

> there are 31 others not shown

just click the link and it will show the others, this seems to be a limitation/UI feature of arxiv. The paper itself contains the full list

I think they just meant that it's so many other people that they don't reasonably fit in the UI

Why is this the top comment? Many of the comments, as well as this one, have no relation to content and only mention a triviality

It’s top comment because it’s anti-Chinese.

> It’s top comment because it’s anti-Chinese.

You need to tighten up your tinfoil hat. When HN features all the discussions on OpenAI's shenanigans on Navier-Stokes, you weren't seeing claims this was anti-US.

It wasn't anti us, it was anti ai company. Maybe because American exceptionalism states that the us can do no harm

I don't know the original intentions, but I don't feel it's anti Chinese just by itself.

It could be pro-Chinese if you frame it like they give merit to anyone involved, or that they take care of their most valuable employees (maybe they pay them another way), or whatever.

Or just observant of how Chinese labs publish their stuff. (Not sure if it's always like that I haven't take a look at the number of authors in this type of publications. Or at least I can't remember it.)

Because if they get you to dismiss or not care about what nonAmerican AI companies are doing maybe people won't see that they are innovating in public. So they can continue smear campaigns that China is simply stealing everything and that's all they do. So the average person will be relieved when the govt moves to regulate away competition from us AI companies and labs. Horray

> Why is this the top comment? Many of the comments, as well as this one, have no relation to content and only mention a triviality

This is hardly a triviality. If you are interested in a topic and you find a paper interesting, more often than not you are interested in reaching out to the author.

If a paper features a list of authors that outnumber the paper's pages 10-to-1, it's a major red flag, and it's quite plausible and expectable that 99%of the names in that list had zero input and might even have zero contribution to give to the topic. I'd even argue it's a kin to academic fraud.

By the way, the same goes for those researchers who work on paper mills, and manage to rake in production metrics that go well beyond an article per day.

If you're confused about who to contact to follow up on the ideas of a particular paper then write to the first author, or whichever author is listed as the contact person. This is not hard and is an orthogonal consideration to the length of the author list.

Which itself not necessarily a sign of fraud: https://www.science.org/content/article/physics-paper-sets-r...

> If you're confused about who to contact to follow up on the ideas of a particular paper then write to the first author, or whichever author is listed as the contact person.

That just goes to show how much you understand the topic. The leading author is usually the institutional leader, and in paper mills it's frequently the guy who stands to benefit the most from the scam. Not the guy who actually did the work.

So now that you missed that one option, what's your plan? To walkt through the hundred names in the list?

I miss the YOLO (CNN computer vision model) days where one dude can publish a paper, completely disregard academic conventions, and yet push the field forward by leap and bounds.

Yea those days were cool. Shame he quit doing research once he realized the biggest application for his work was killing people with robots I guess. It was a nice period of time. Before everyone in industry realized the cool stuff they were building was likely only going to be used to surveil the public, target people, remove skilled labor, and probably develop new weapons (whether people call them that now or not isn't relevant). Good old days.

Or maybe it's an healthy company doing proper research? Who knows?

>this is the top comment

christ

Or without the cynical lens they just operate like CERN where they acknowledge everyone who without them could not complete the research. Hyperauthorship is common in particle physics for example.

As models get better, safe-and-secure sandboxes/environments are going to be the way forward. With the recent rise in cases where models can somehow gain access to the internet and blast past the sandbox, it's very important to have all the resources contained within the sandbox with no access to the outside world.

DSec is a good step in this direction!

380.000 concurrent sandboxes on 160 Epyc based server nodes. Crazy stuff

that's only 12 sandboxes per core

Um, how is that not impressive

if it's agentic stuff, they likely aren't hammering a core constantly and they will maybe sit idle quite often between model requests, so it makes sense. I do wonder how much memory they allocate to each one though.

it's just very efficient use of shared cores that is required to make these kinds of workloads cost efficient

If they sit mostly idle, you can swap out a lot of the memory to SSD, I guess.

A place I worked back around 2020 was running a Grafana instance per customer that got embedded on the web dashboard. We had 110 pods per GKE (kubernetes on gcp) 4 CPU node because that was a network imposed pod limit at the time. The nodes were usually idle--could have shoved a lot more on if not for the IP limit.

I think around that time Grafana changed their license tho so you couldn't host OSS Grafana as part of your service.

how is it impressive?

>"Within one scale unit, the platform spans nearly 160 CPU nodes with 30K cores and ∼250 TB of DRAM. It manages petabytes of layers and images. On a typical day, a single scale unit serves about 3 M sandbox instances, with peak concurrency reaching ∼380K and a creation rate exceeding 5,000 instances per second."

Impressive numbers!

Whoever would have thought (in prior years) that in 2026 AI Agents (not people or corporations, at least not directly) seem to be (or seem to be rapidly becoming) the biggest consumers of cloud computing resources...

Anyway, a very interesting paper and environment!

12 sandboxes per code is insane, I wonder how many of these sandboxes are idle at a time. Depending on the tasks assigned the resource requirements are different. Compare an agent doing pdf conversion and one responding to a simple question. One is cpu bound the other is mostly network wait.

This is an interesting problem from infra perspective since you cannot predict the workload. On a bigger scale you may get away with forecasts.

Im waiting for tech that elastically allocates cpu/mem without restarting a container.

Appears to be similar to what Google is building with ax https://github.com/google/ax

Is this like agent substrate?

Yes and it needs to be load bearing. It’s not a claim that its seams are.

Very similar to Google Ax, looks like it's Deepseek answer to Google

Ax/agent substrate is focused on enterprise inference. This is squarely targeting the training and reinforcement learning patterns (it's not to say these technologies don't overlap, but perhaps closer to kube with knative & kata container support (and some level of session persistence across all 3). Or nomad... Or any orchestrator that supports multiple underlying isolation techniques.

I can't work out from the paper if the session can cross all 3 types of isolation (between persistence) but I'm assuming it can if it needs to evolve tool calling and security requirements.

In any case it feels like they're just composing and mangling technologies into a platform and not bringing much to the table (from what I've read so far; ie still using well established isolation tech and not something unexpected like a microkernel) but the outcomes are impressive

I wonder if they are signalling that if they can do this for training, then they can create an style agent swarm to hack anyone with 380k concurrent agents.

Isnt the bottle neck for that inference capacity not how many sandboxes you can generate?

Is there a lab more innovative than DeepSeek? Imagine if they had the same compute resources that Anthropic and OpenAI have.

Food for thought: Constraints are the source of creativity.

Yeah if they had the resources of an OpenAI or anthropic they’d be OpenAI or Anthropic. Scrappy underdogs have to be nimble and innovative.

They are as “underdog” as Linux is to Windows.

Not really, they are underdogs in the true sense of the word.

Let's wait a bit and see how your US frontrunners will age with their operating costs above revenue, model pricing collapsing faster than compute efficiency improvements and cost pressure from DeepSeek and friends.

Possibly relevant: 突破技术壁垒, "break the technical barricade" -> overcome an obstacle through innovation (in this case, sub SOTA GPUs at least).

Native Chinese speakers to confirm....

your translation is correct, I would say Chinese labs may be able to figure out the current capability of latest frontier models in 3-6 months, but then Anthropic and OpenAI may have already been ASI in that time already. China's main problem is still lack of (good) chips, and that is a hardware issue that is unlikely to be solved for a while. and more effort for efficiency means less effort for actual capability improvements. we have already seen what anthropic can do if they focus on efficiency with opus 5.5

This isn’t a law.

Bell Labs would be a good counter example. All the resources in the world and no demands.

General Magic failed with all the talent and all the resources. Cautionary Tales podcast featured General Magic in a recent episode; the basic argument was no constraints led to no focus.

You don't need a counter example to a counter example.

It's not that I think the opposite of the original statement is true, I'm suggesting that's it's just all noise as there are both successes and failures at both extremes.

Short term they might have less compute, but long term they will most definitely have more compute. They don't have to worry about energy, they don't have to worry about people blocking them building data centers the only thing stopping them is there no Chinese manufacturer that can produce a chip as good as Nvidia but I would bet that solved in a year or so.

They don't have to worry about people blocking DC construction in the US either. All new AI datacenters are designated "dual-use" so the federal government is already letting local councils know to fuck off.

they can also put them in better places, vs trying to arbitrage expensive electricity costs and various US corruption thats built more around paying people off than getting things done

Yeah, but that brings up another problem that China doesn't have. Their federal government is always on the same page. The US federal government changes every two years and it doesn't seem like one side is going to be allowing building of AI data centers anymore and in fact might just ban AI in general.

It's hard enough to build them in the US that multiple companies are unironically spending 10s of millions of dollars to try to build them in space

Which cynics could say is marketing hype, but I tend to believe it's extremely difficult to build anything land/energy intensive in the modern United States

Or some people have more money than sense. Building data centers on earth is trivial compared to getting the same amount of compute to space.

There are places that aren’t in the US that also aren’t in space. I still don’t get why it makes any sense to put this stuff in space, but if it does, it’s not the political difficulty of building stuff on earth. So it’s hard in the US. You’re telling me Mexico wouldn’t allow it? Argentina? Mongolia? Build a data center on a French nuclear test site in the Algerian desert and it’s still easier than space.

10s of millions isn't even a rounding error in the context of AI-related spending.

Just because other companies don't write papers about what they do doesn't mean they aren't innovative...

Sure, but we’ll never know what they do or if it is innovate because we won’t know what they do.

Well we do know the results. They’ve been at the frontier constantly, including developing the entire field and set of capabilities. In the beginning every OpenAI paper was basically a landmark.

They were at the frontier early since they allegedly “stole” the internets data first. Now they’re just lame ducks trying to ensure they stay at the top by legislation and ensuring open source doesn’t get its say. Also OpenAI trying to adopt any other product (legal tooling, healthcare, latex replacement, etc…).

They were at the frontier because they invented many of the techniques. Google already had the internet’s data well beforehand. Anthropic and OpenAI are still at the frontier today, independent of legislation.

Oh no I agree, they were among the first to be popular in the llm frontier with their techniques. Today they are only holding on to the frontier because of that lead.

They want the legislature to keep their lead from other competitors.

I'm actually the most innovative, I've written thousands of papers advancing the state of human knowledge. They're just in my basement and I don't show anyone.

"Necessity is the mother of invention."

Not sure they'd be the same without the constraints.

This is literally just a scheduler over firecracker man what

So.. serverless ?

Still very impressive! Love how it's done!

Can you give a brief for what this is and why it is impressive?

It just describes the platform they built for scheduling workloads, and running those workloads. After a second thought it is not that impressive and probably doesn't deserve any kind of hype. It's the same kind of setup AWS is running for Lambda, as well as anyone else basically running SLURM clusters out there.

Still giving them credit because creating such as scheduler/platform from scratch is quite complex, and I know that they probably struggled a lot to get it right.

I'd say it's cool that they're openly writing about it and how they run their workloads. shows a nice window into how these things are actually deployed at scale.

it also shows how much density you can get easily from a single core if you wanted to replicate this

That's not quite right. A recurring trend in ML is figuring out how to get an elastic interface for the the particular quirks of ML workloads (source: I worked on an open source one years ago for inference). Recently, there's been a lot of interest in doing this for agent workloads. Google recently released something called ax that is similar in spirit. The core of it is this: https://github.com/agent-substrate/substrate

At a high level, agent sandboxes have peculiar needs. Agents are really bursty, but also long lived. You need low latency suspend/resume calls. Checkpointing and recovery have some particular considerations. And naive approaches are often really wasteful, but over optimizing without harming durability, isolation, or consistency in performance can get tricky.

This is the new hot infra topic for agent swarms, for the time being. Whether that's impressive or not is up to the reader I guess, but it's a pretty involved project nonetheless.

The person simplified the paper with such a misunderstanding that I thought 10x before rebutting. They clearly didn’t read anything of the paper. Thanks for the clarification.

I've overlooked too much the optimization in the workload that get dispatched and just looked at the scheduling part. Thanks for the explanation!

That number of authors though

The topic isn't as interesting as how 131 authors communicated to get this out.

Is it possible they're doing a research lab "socialism style" and everyone gets equal credit for just being a part of the lab, regardless of actual input into the specific papers? If they're innovating in computer, maybe they're not so afraid of innovating in social/academic structures as well?

This is common practice in Biology labs in the west I think

Maybe there are even reasonable justifications for it, instead of yelling about "socialism".

Sorry, didn't mean it like "yelling", I'm quite a fan of socialism and don't see it as any negative. I didn't think anything in my comment was negative.

It's negative in the sense it suggest some authors didn't contribute (or contribute little0 to the research, and were perhaps just warm bodies that happened to be present at a certain location

I don't see my comment suggesting that at all? Personally I like the approach, and felt it was a positive approach when I made my comment, apparently it reads very differently than that.

I understand your objection, and would like to retract my reading of your comment, it was sloppy.

This doesn't mean anything partisan to me, but I mentally auto-classified what you said as a partisan argument, and reacted in a knee-jerk way. I'm happy that the comment you are responding to still made a best-faith argument to explore possible implications of what I said, and what you said.

But I want to draw a sharp line of what I want to discuss here, and failed at doing that. Open discussion of economy and politics is interesting to me. I don't consider myself a communist or socialist, neither a fierce "anti-leftist".

But my comment did invite exactly this polarization, and I'm thankful that despite of it, there was still a constructive follow-up comment.

Sorry for the blunder on my part: I simply skimmed your comment while distracted, instead of reading it, then reacted reflexively.

I'd myself be confused by the accusation I made ("yelling about socialism") would it have been thrown at me when simply discussing some geopolitical/economy topic.

The west does it too in experimental papers. This is not a socialism thing.

I am author in two author papers, but also in 200 author papers. Nothing wrong with it. In our field they are studies among a dozen centers, where each center has to do something and the results are pooled to some central hub. If I do a relatively easy data prep before sending to the hub, I go among the authors, it is just the way it works. A huge work that is spreaded thinly, or else can't be done unless you ask people to work without attributio. That might work once or twice, but not more, if you are the person who ask for favours but returns nothing.

This is your brain on propaganda.

It's not socialism to recognize the people who actually did the work. I realize it may seem to be capitalist to actually believe that managers or ceos are actually building products at all. But anyone who has worked an engineering role is well aware that's not how it goes and it's not capitalism to do that it's just crappy human behavior. Yea like how Elon musk is actually building those rockets while designing AI systems and new cars or huang is out there giving tips to the people designing the chips ...

A nice work perk is recognizing your teams accomplishments publicly. Give it a try. I hear even capitalists like to be treated as if they are human. Some dude named deming used to talk about workers pride. Sort of a smart guy I guess...

Again, not sure what makes people believe I don't like the approach or disagree with it, I was just replying what the reason could be to parent commentator. My previous comment has no negativity or anything, and you're barking up the wrong tree here.

The thing that makes people think you are against socialism is, this very website, which means towards American exceptionalism rather hard and socialism is the antithesis to that so it's often criticized here and by other right wing places

I don't understand why half the comments here are about the author list, this is very common practice in e.g. large-scale physics experiments and biology, and every new GPT release from OpenAI equally had papers with tons of authors

Agreed, I'm very confused as well. It's like no-one here has been paying attention to research papers.

Or they are just rushing to say anything, and it's much easier to comment on that than the content of the paper.

Some people don’t want to recognize the important work that went into this.

Doesn't deepseek have like 150-200 employees?

I wonder how those few that were left out feel :)

Will admit that I haven’t read it yet, just saw the crazy number of authors and think this may compete for one of the papers with the most authors.

This one [0] about the Higgs boson is hard to beat. Pages 25-32 are just full of around 5150 „authors“, pages 33-37 are their affilliations.

[0] https://arxiv.org/abs/1207.7214

Even without that particularly special case, high-energy physics collaborations have long since broken the idea of authors. There are now many collaborations with hundreds of authors publishing regularly and quite a few that are into the thousands.

I like it, represents science's general 'standing on the shoulders of our precedents' far more realistic then the 'genius solo' PR mythos.

Where is the balance between this and citations? Authors implies said person provided additions to get the work done.

With projects like the LHC, there really are thousands of people involved in "getting the work done."

Can you cite the paper with that?

https://www.science.org/content/article/physics-paper-sets-r...

After parsing that article to get the paper link, that is insane. I am not in the field, but find it curious that each every person of that list of authors had direct significant contribution. Not my field though so just seems odd.

I think the number is ~2930 authors according to the CDS [1] which is comparable to the corresponding CMS paper which has about ~2900 authors [2]

[1] https://cds.cern.ch/record/1471031

[2] https://impact.ornl.gov/en/publications/observation-of-a-new...

That, is crazy to me, but different areas and all that I suppose.

That, is just bonkers to me. Different fields and everything, but that sounds like a whole company authored the paper.

how do you even keep up with the amount of research coming out these days

Reply on news.ycombinator.com