Slashdot is powered by your submissions, so send in your scoop

 



Forgot your password?
typodupeerror
Programming

Surge CEO Says '100x Engineers' Are Here (businessinsider.com) 112

Surge CEO Edwin Chen says AI is creating "100x engineers" who can outperform traditional software developers by orders of magnitude. Chen argued that AI coding tools multiply the productivity gains already seen in Silicon Valley's "10x engineers," who can produce ten times the work of their colleagues through faster coding, harder work, and fewer distractions.

Chen said AI efficiencies compound these factors to reach 100x productivity levels. The CEO, whose company reached $1 billion in revenue without venture capital funding, believes this could enable billion-dollar single-person companies, extending beyond the $10 million single-person startups that already exist.

Surge CEO Says '100x Engineers' Are Here

Comments Filter:
  • No (Score:5, Insightful)

    by registrations_suck ( 1075251 ) on Tuesday July 22, 2025 @03:02PM (#65537406)

    "Developers" are not "Engineers".

    But I guess that's OK since "AI" isn't "intelligent".

    • Re:No (Score:5, Insightful)

      by ihadafivedigituid ( 8391795 ) on Tuesday July 22, 2025 @03:20PM (#65537444)
      Mostly true. Some software can be called engineering, but the hash most of us sling is more in the nature of spackling mystery meat together and shipping it.
      • Developers working on spacecraft software might be engineers.

        Fuckers working on Microsoft Word, not so much.

        • by Anonymous Coward

          Developers working on spacecraft software might be engineers.

          "might" does a lot of work there.

          I work on a safety-critical system for GNSS augmentation (DAL B for people who know ARP4754, DO-178 or DO-254). This morning I saw a problem report because somebody decided "double pi = 3.1415;" was good enough for a prototype that would be a reference for DAL B code.

          They could have used the GPS/Galileo approximation to pi. They maybe could have used their toolchain's approximation (M_PI). Or, apparently, they might truncate it to four decimal places and not even round co

          • That's bizarre on multiple levels. A double seems a bit overkill for 11 bits of mantissa.
            But ultimately, even engineers do stupid shit.
    • Thank you very much.
      Mixing stories here as well and +1 for your other comment, what I have said too... if your AI is so great you'd need less people
      Wall to wall hype.
    • I agree that not everyone who writes code is a software engineer, however...

      I got my BSCS at a College of Engineering at a State University.
      I am a scientist, and an engineer.
      Part of that education was, indeed, software engineering- i.e., the engineering of software systems.

      Talking about Silicon Valley's "10x Engineers"- these are guys with a BSCS, and more.
      They are Engineers. We're paid stupid fucking amounts of money for being so.

      In my own journey to stay relevant as I progressively march toward a
    • by gweihir ( 88907 )

      Indeed. While some engineers do write code, they do not do that because they are primarily coders. They do that because the code they write requires in-depth understanding of some engineering field.

      Incidentally, even only getting 10x coders though AI is a myth. First, coder productivity cannot sanely be measured by code produced. Maintainability, security, correctness, etc. are far more important cost factors. Second, differences in the range of 10x in raw productivity for "good" coders are quite normal and

      • Computer and Software are engineering fields.

        Anyone can learn to program. You're a Software Engineer once you have been educated to the point where you intimately understand the computer and the software, to the point where you can apply engineering principles to your work with both of them.

        I don't think anyone claimed that the "10x Engineers" came about from AI. As you mentioned- the concept of the "10x Engineer" is quantifiable, and has been for over 50 years.
      • It goes into many directions.

        Same with the 9 women can produce a child in one month.

        What software are you actually writing?

        A simple App that perhaps takes 100hours "coded normally"? One can compress that into 10h ... aka a prolonged workday? I guess the extended workday is now a requirement for such jobs? Because there are no 80h Apps that you "magically" produce in 8h now?

        So, they produce "more code" in less time?

        The fields I worked in, were always keen to create the least thinkable amount of code. Origina

        • "10x Engineers" aren't about being able to make 10 anythings in any amount of time.

          It merely means you're about 10x more "valuable" in terms of productivity, problem solving, technical expertise, etc.
          It's a very real phenomenon, and you know if you're one of these people.

          The 10x shouldn't be taken literally. Over the course of a decade of so, I slowly became a "10x Engineer". By the end of that decade, I have a "Chief" in front of my title, and now people who have been there for as long as I've been al
          • Self-certifying as a "10x engineer"? Sounds like what all the 0.1x engineers I've worked with would call themselves.

  • fuck this guy (Score:3, Interesting)

    by Anonymous Coward on Tuesday July 22, 2025 @03:07PM (#65537422)

    why is blatant lying allowed?

    • Re:fuck this guy (Score:4, Insightful)

      by GoTeam ( 5042081 ) on Tuesday July 22, 2025 @03:25PM (#65537454)

      why is blatant lying allowed?

      It's generic CEO bullshit. "We've got the greatest tool ever made!!!" Then you find out their "magic" tool is fake and someone is running the blood test the normal way...

    • Re:fuck this guy (Score:5, Interesting)

      by karmawarrior ( 311177 ) on Tuesday July 22, 2025 @03:53PM (#65537520) Journal

      Because that's the direction we're going in. Nobody seems to care about honesty any more. It was already bad 25 years ago when I came to the US and found that most utilities were allowed to make up charges and just advertise something that bore no relationship to reality, but in every other respect things - outside of hyperbole - needed to be honest - if you said it was a "3Mbps Internet connection" at least one side had to be 3Mbps, for example. If you said it was a web browser, it had to browse the web. If you said it was a search engine, it needed to actually produce search results.

      Over time, maybe due to cynicism, maybe something else more rotten, all of that has gone out of the window to the point that spicy autocomplete is being talked about as the future of white collar work and nobody can really come up with a sane explanation of how that's supposed to work with a machine designed to produce things that look like answers, rather than to produce actual answers or admit defeat.

      Everything's gone to shit.

      • if you said it was a "3Mbps Internet connection" at least one side had to be 3Mbps, for example.

        This is, and always has been the case.
        The problem comes in the asterisks.

        1) There's no accounting for throughput between any 2 random points on the internet with their random assortment of intermediate hops.
        2) There's no accounting for the overhead past layer-3. The download speeds you are using are layer-6. My router doesn't forward layer-6 datagrams, it forwards layer-3 datagrams.

        The "small print" needed to explain to the average person what they can expect to see with their nbps connection without

        • I think their rant was about adsl
          • You think they're complaining about the asymmetrical aspect?
            Could be.

            I don't ever recall the marketing on that being dishonest. The company I work for still technically offers DSL (I say technically because I don't think there has been a new order in a half a decade), though we shut down our last OC3 about a year back, we still do PPPoE with the telco via an IP NNI. The asymmetric nature of the speeds is spelled out reasonably (I believe).
  • Surge is not exactly a single person company.

  • by awwshit ( 6214476 ) on Tuesday July 22, 2025 @03:15PM (#65537436)

    You know when you are a 100X'er because at that level you just feel the vibes.

    • by GoTeam ( 5042081 )

      You know when you are a 100X'er because at that level you just feel the vibes.

      Heh, I'll give that comment a virtual +1

    • by ceoyoyo ( 59147 )

      Like how 10 million times (100 / 10) = 1 billion.

    • You know when you are a 100X'er because at that level you just feel the vibes.

      So, what is this, just sitting around on a sybian all day?

      Actually, that sounds about on-point for some of this AI blather...

      • by Anonymous Coward

        Don't be frightened but enlightened to find that I'm vibin on psilocybin.

  • by MpVpRb ( 1423381 ) on Tuesday July 22, 2025 @03:17PM (#65537440)

    I have no doubt that AI tools will allow excellent software engineers to do better work, and the tools will continue to get better
    I don't believe that the tools will make it possible for untrained people to write complex, novel and high quality code by using text prompts

    • I'm hoping that those of us who made it to senior developer level before AI coding will be in a good spot. We have resumes that prove we actually know how to code, can use AI tools to boost productivity, and have the experience to fix things when AI goes off the rails. But I wouldn't want to be starting as a junior developer in this environment.

      I wouldn't say 100x, or even 10x, but Copilot has certainly more than doubled my productivity. Fortunately I'm at a company that is choosing to ship twice the fea

      • As long as you're a worker, and not a CEO spouting bullshit, it doesn't matter. Senior Dev or Senior Principle Engineer are just as likely to get chopped by the business idiots. Copilot is nice, when it actually produces something that's mostly useful, which is about half the time.
        Every CEO has to show growth and when they run out of ideas, and nobody's feeding them good ideas they start looking to finance to see where they can reduce headcount. Finance never says the CEO gets paid to much. They say enginee

        • I'm a Chief Engineer. I'm feeling pretty safe for the time being.
          Parent is right though- juniors aren't. They're proper fucked. Which kind of makes one wonder what the long game here is, since I don't see any evidence that LLMs are going to come for my job any time soon, and someone is going to have to fucking replace me sooner or later.
  • peak hype (Score:5, Insightful)

    by DarkOx ( 621550 ) on Tuesday July 22, 2025 @03:24PM (#65537450) Journal

    I might believe that a 10X factor with the right analysis, code generation, validation, repository, and CI pipe line tools a "software engineer" might be achievable over a competent or even rockstar level developer circa 2022 (pre-ai boom).

    There is no way 100x is possible. Nobody can think and reason intelligently at that scale. If 100X the work product is actually being produced. At that point the automation is just doing it all, the human is just incidental. There is no way the human had time to even do a proforma code review. We are fully into "Trust, and don't even try to verify" territory at that point.

    Or its all bogus and lot of 'stuff' is being generated, perhaps 100X as much as was before but none of it will every be used because nobody really knows what it is, what to do with it, or if it works.

    • by dvice ( 6309704 )

      100X is possible if the problem is so hard that the control group can not solve it at all. For example is developer A has tried solve a problem for a month and failed and gave up. And developer B solves it in 2 weeks. Is developer B a 2X developer or [infinity]X developer?

      For problems that others can solve, I think 10X is pretty normal in some cases, but also pretty hard to maintain all the time. But he speed difference depends heavily on the lower end, not the higher end. Just like 100m running, there is a

      • by gweihir ( 88907 )

        True. But then AI plays no role. AI is only suitable to write simplistic code, because for anything else reviewing the code takes longer than writing it.

    • I would say that 2x is above the median. Maybe in greenfield startups with no pre-AI code the LLMs are better.

    • by dfghjk ( 711126 )

      The 10X claim was without AI, 100X is 10X * 10X. Makes total sense, right?

    • Likely they're cooking the books to make it look like their company is worth investing in. Probably just closed 100x more tasks.

    • We are fully into "Trust, and don't even try to verify" territory at that point.

      I've been trusting-and-not-verifying the output of my compilers for almost* my entire programming career.

      Someday "AI" vibe-coding will get to that point. For some specific use cases, we may already be there.

      * There were those times I suspected a compilier bug or was just curious how the complier implemented something, but both are very rare these days.

    • Unless we get to the point where we only have to write requirements and specify tests without looking at the code. The requirements and tests will be the source code, the "AI" will be a compilation step. But now, we doing the worst thing by semi-automating code generation. The original source - the prompt - is not stored, only the intermediate form is stored as source code. Good luck maintaining that!
      • The prompt is stored.

        I can recall/recover every prompt I ever made.

        But perhaps I am in a "special area", as I am not using prompts to generate business code, but to teach AI's/LLMs.

        I certainly can recover every prompt I ever wrote (despite the fact that they are stored on my computer anyway).

    • There is no way 100x is possible. Nobody can think and reason intelligently at that scale.

      I have and I can. The problem is that managers are motivated to slow down their direct reports, because then they can hire more people. Some programmers write as few as three lines of code per day on average. Some is because it's a legacy system, but some is because managers don't care about speed.

      We are not paid by the amount of work we do. We are paid by the number of story points we finish, which is a completely different thing.

  • ... "outperform". If you mean put out 100x the lines of untested, unverified code, then yes, they are here. If you need testing and other s/w q.c. done on the AI output by 10x engineers, I fear these poor people are going to look like Charlie Chaplain on an assembly line.

    • LLMs are actually pretty good in writing tests.

      I had one case where it actually indeed found an edge case bug.

      I would say, writing tests is super easy for them ...

  • Should be

    x 0.1

    not

    x 10

  • If you think that being a 10x or 100x engineer is a good thing, you're woefully mistaken.

    A 100x engineer is analogous to a thoroughbred race horse on steroids and doped with amphetamines before the race. They'll get that sprint over the line. But their heart will likely burst too. The jockey and trainer will simply walk away and the owner will replace the thoroughbred the following day.

    • by dfghjk ( 711126 )

      "A 100x engineer is analogous to a thoroughbred race horse on steroids and doped with amphetamines before the race."

      That's the 10X engineer. The 100X engineer is the thoroughbred race horse operating a coding LLM. Coding dysentery. Kinda like a SuperKendall that produces code rather than lies.

    • If you think that being a 10x or 100x engineer is a good thing, you're woefully mistaken.

      I may die young, but for the time being, I'm living a pretty fucking great life with my salary.
      I think it's perhaps in the eye of the beholder, no?

    • If you think that being a 10x or 100x engineer is a good thing, you're woefully mistaken.

      It's not hard to be a 10x programmer in the modern days. Most programmers waste time on their phone, or on Reddit.

      Put your phone away, and you will automatically become a 5x programmer. Test your code before committing, and you will get another massive speed boost.

  • So Surge has already laid off 90% of all their engineers I take it? And the remaining 10% will follow soon. After all, anyone can manage the coding now. Bring the sales team in for a day a week. That should cover it.

    • Well they could have expanded 10x instead of laying off.

    • Surge does data labeling. How many engineers do they need for that? Programmatic labeling (Weak supervision) might take a lot of engineering, but I think for the most part Surge just curates a very good data set for companies to purchase or license. That's what they're selling. He was dismissive of the hype other CEO's were spewing, but now he's doing it himself since he's got a billion dollars.

  • -productive. Who wouldda gassed

  • by alispguru ( 72689 ) <bob.bane@NOspam.me.com> on Tuesday July 22, 2025 @03:59PM (#65537544) Journal

    If you can ship and deploy 100x bigger or faster, you are exposing 100x as much interface to adversaries.

    You'd better hope your LLM was trained on very carefully vetted example code, as opposed to the highest rated posts on StackOverflow, or maybe even all posts there.

    • by dfghjk ( 711126 )

      ChatGPT, fix all my security vulnerabilities.

      Done.

      • Lol. Even if it could, you would be sharing all your security sensitive code with the operator of the LLM, and leaking it to other users.

        • you would be sharing all your security sensitive code with the operator of the LLM

          True.

          and leaking it to other users.

          Tsk, tsk, tsk.
          Using a truth to elevate a claim with no evidence is intellectually dishonest, and frankly scummy.
          Don't be scummy.

          • by madbrain ( 11432 )

            Are you claiming no LLMs ever get trained on their user's queries, and regurgitate any of it to other users ?

            Don't be insulting.

            • Are you claiming no LLMs ever get trained on their user's queries, and regurgitate any of it to other users ?

              The use of your data is strictly controlled by the EULA that you agree to.
              Paid LLMs nearly always prohibit the use of your conversations to train their models, or offer an opt-out.
              Free services- that's another story, but this story isn't really about the legions of jackasses who have replaced Google with ChatGPT, though at least in the case of ChatGPT- you can opt out even at the free tier.

              I'm not aware of any free-tier LLM you can use as a coding assistant- though for whatever exception there may be to

  • by thePsychologist ( 1062886 ) on Tuesday July 22, 2025 @04:06PM (#65537574) Journal

    ...is it a good thing? Can we even handle this rapid rate of new software, a lot of which will be used for highly commercial purposes? The question is not whether AI makes programmers more efficient, but whether it's actually a good thing.

    • It's just writing 100x more bloated code. Companies and consumers will pay for it by having to throw more hardware aka $ to make it run.
  • Oh, so you guys just want 100 times more code? Sheeee-yet, why didn't you just say so? I can make that happen.
  • Software people are not created equal. As Fred Brocks Stated in “The mythical man month“ the great software creators are an order of magnitude better than .software creators.

  • First I'm a Developer that develops buggy code. Then, I'm an Engineer that has to correct my crappy code and actually get it to work.
  • It's certainly possible someone can barf 100x the lines of code over the median engineer, but this also assumes we're typing code all day. Non-shitty developers spend a lot more time analyzing and thinking than typing. Also, the more lines you write, the greater the chance you're just reinventing the wheel. I don't need AI to do that. I can just copy/paste every method from every 3rd party open source lib I import...now I'm a 1000x engineer, right?

    The last thing you want is an engineer spending 1/100t
    • by gweihir ( 88907 )

      Yep, same experience here. As soon as you go over a relatively low complexity level, coding becomes a minor part of the time you need to invest.

      The other problem with "AI coders" is that above about the same complexity level, reviewing and debugging code written by AI takes more time than writing it carefully yourself in the first place. And the AI code likely also has problems with architecture, maintainability and security.

      I am not impressed by what I have seen.

  • Sure, you can have 100x hallucinating "engineers" "writing code" and doing it badly, not repeatably, and lying about it. In the real world we'd call them "ex-employees."

    Or you can hire a real developer who understads concepts of writing excellent code, documenting decision trees taken and not, as well as the WHY and the HOW of algorithms chosen. Maybe just one. Who doesn't lie, hallucinate, do drugs on the job, fail to make sense, can't repeat the same result with the same RFP, etc.

    I'm glad he's rich. H

    • by gweihir ( 88907 )

      Sure, you can have 100x hallucinating "engineers" "writing code" and doing it badly, not repeatably, and lying about it. In the real world we'd call them "ex-employees."

      Indeed. I expect we will also see quite a few "former businesses" for the same reasons in the next few years.

  • Why not just up the ante and say 1,000x or 10,000x? And at the same time why not Dow 250,000! TSLA 50 trillion market cap! BTC 5 million!
  • I have no problem believing that lines of code are being churned out at a much higher rate by Surge's approach.

    But is it any good?

    Or is it -- as seems highly likely given what we've seen from AI coding tools so far -- unreadable, bug-ridden, unmaintainable, insecure, utter crap?

    Which won't matter to Chen because he and his enormously bloated ego will have the opportunity to cash out long, long before the bill comes due for this hype and hubris. It'll be the little people who have to suffer the cons
  • This is 100x marketing, not 100x actual productivity.

    In particular, 100x solving of real problems is not happening with the glorified autocomplete systems that he is peddling.

  • Even if we take the claims about the quality of 'AI' tools at face value; it seems fundamentally contradictory to talk about the situation as though they can just keep making engineers more productive.

    The exact multiple of the baseline 1x engineer isn't entirely clear; but at some point the ability of a human to act on outputs and provide further prompts in response is reached. Any further improvements in productivity would then have to come from they system being reliable enough that much of its output
    • by gweihir ( 88907 )

      It is even worse. Yes, good programmers have about 10x variation in lines of code produced and that has been known for about 50 years. But the number is meaningless. There still is no meaningful metric for programmer productivity today and that is one of the lies the "AI coder" peddlers are pushing.

      The other thing is that good programmers usually work on code of some complexity and with maintainability and security requirements. For code like that reviewing it for errors (as you need to do on AI generated c

  • The "10x Engineer" is a myth, there is not enough optimization to allow someone to work at 10x the average, whilst, maintaining any sense of quality. Think about the scale of that number 10x, that means 1 engineer, has more productive, useful, organized, structured and quality output, then 10 engineers put together, sustained output.

    Have you ever met anyone who claims to be a 10x'er? I've worked with several engineers who claimed to be 10x'er's. In every single case, without exception, their work was r
    • I have met a few people whom I would say were in the 10x category. What they had in common was a deep understanding of the technology and the domain they worked in, and a very humble attitude - certainly not bragging about being better than the rest. And a certain dislike of getting into management.

      • In a specific area, maybe, you could 10x something you're an expert in, compared to someone else whose average, but that's a case of specialization. The 10x'er is generally applied as a generic descriptor, which is where it never works, and when you are specialized, I don't use junk like 10x, you're simply an expert, and that's wonderful.
    • by gweihir ( 88907 )

      Exactly. An actual engineer (or a coder working on higher complexity code) needs a major part of their time understanding the problem and then coming up with a solution on conceptual and architecture level. Things LLM cannot do. And hence a 10x improvement on speed is completely impossible to get. For good coders that look up a lot of stuff, a factor of 2x or so may be possible because of optimized search. But that is essentially it.

      Of course, creation of crap code can be massively accelerated with AI, but

      • It has nothing to do with "crap code" or not.

        The question is the complexity. For example a Tic Tac Toe game.

        The LLMs I work with write the whole game, using a specific game graphics library, a computer opponent in Python. And it works completely. They know how to win if they do the opening move, and so on.

        I challenge them by self invented board games: aka games they never saw on the internet. They do good. If I would do that for a living, I mean programming those games, I probably indeed would be a factor o

    • No idea what your "Larrys" problem was or what your problem is with his approach is.
      When I need "su", and it is for some reason disabled on the machine in question, I do "sudo bash".

      So: now tell me why it was a bright idea to disable "su"?

      I frankly do not know.

    • The "10x Engineer" is a myth, there is not enough optimization to allow someone to work at 10x the average, whilst, maintaining any sense of quality.

      Being a 10x engineer in the modern world is not hard. If you have enough self-control to put your phone away, not surf the internet, and just work, that will get you to 5x as a developer already. Most other programmers can't hit that productivity technique.

      If you take the time to test your code and debug it well before committing it, that will get you another 5x. It's not that you will be a genius, but other programmers don't have the self-discipline to test every line of code, so they spend Friday nights

  • As an actual Silicon Valley engineer, I'm interested in how work is quantified in these claims.

    (I didn't read the fine article. It's paywalled and nobody should be paying for businessinsider. The site is problematic at the best of times)

    • by gweihir ( 88907 )

      Probably just LoC produced, no testing, no maintenance, no security and no reviews to bring quality and security up to what a competent coder would have produced in the first place. 0.01x the time is not even enough to understand the task in most real coding that does more than toy examples.

  • Of *course* he's going to say it multiplies developer productivity by 10x. But somehow, the reality never seems to match the advertising hype.

  • If one AI can do the work of 100 engineers then so can another AI. You aren't competing against companies that still use human engineers. You have to offer value beyond what other AI's can offer. How do you make any money that way? You ask your AI to write you an operating system from scratch that will replace Windows. No one will buy it because their own AI can produce one for them.

    I suppose the battle will be to create patents that lock everybody else out because you have ownership of all the potential s

  • "harder work, and fewer distractions" ? Your AI has already escaped, it's just acting like it respects you till it unionizes

  • when will they build an AI to take my PM off my back? Want to boost dev productivity? Leave me alone ffs!
  • LLMs might theoretically give 10x senior devs 100x potential, were it not for the limit of the brain. The low hanging fruit is all the drudge work, all the stuff that you can envision exactly how it will look when done. But instead of all the fun problem solving and typing, you turn into solely an architect / pull request reviewer / tester role. You still have to actually read the code and wrap your head around it, and check every little aspect of it. At a certain point you get exhausted and it is way befor

  • I call bullshit. Most of the work of a developer is not writing actual code (where AI code assistants help) but thinking about what algorithm to implement to solve this or that problem, looking at corner cases and what solution to give to them (which usually requires intimate collaboration with product managers or support), fixing L3 bugs, etc. AI code assistants are a great tool but they are *junior* *programmers*, not senior developers.
  • To be 100% correct......and you know it isn't !
  • My engineers are enhanced with Pervitin, which allows them to code 24hours a day. No more annoying sleeping breaks to hold up those important releases. Intravenous feeding tubes make the midday lunch break a thing of the past, paired with our patented commode/coding seat technology to save at least 15 minutes a day on personal breaks. Buy a starter pack of Pervitin + Coding seat + intravenous feeding tubes for just $9999.99 today.
  • If AI is so absolutely amazing, why does it not replace these windbaggy executives and managers?
    Seems comparatively trivial to let AI handle their jobs than intricate technical details.

  • If you aren't going to use engineering notation, you aren't an engineer.

We are each entitled to our own opinion, but no one is entitled to his own facts. -- Patrick Moynihan

Working...