Surge CEO Says '100x Engineers' Are Here (businessinsider.com) 73
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.
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.
No (Score:5, Insightful)
"Developers" are not "Engineers".
But I guess that's OK since "AI" isn't "intelligent".
Re:No (Score:5, Insightful)
Re: (Score:2)
Developers working on spacecraft software might be engineers.
Fuckers working on Microsoft Word, not so much.
Re: (Score:1)
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
Re: (Score:2)
But ultimately, even engineers do stupid shit.
Re: No (Score:2)
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.
Re: (Score:2)
I'm here to help (;
Re: (Score:2)
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
fuck this guy (Score:3, Interesting)
why is blatant lying allowed?
Re:fuck this guy (Score:4, Insightful)
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:4, Interesting)
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.
Re: (Score:2)
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
Re: (Score:2)
Elon / Musk are still singing the praises of the H1B program, while woketards are defending incredibly massive illegal immigration and inventing lies about the agents just trying to enforce immigration law.
I suspect, but cannot prove, that very few of the illegals being rounded up by ICE are programmers.
Re: (Score:2)
Certainly you're right that they're not a huge constituent though.
Re: (Score:2)
Keep telling yourself that. After 30+ years, between DEI, disability, and AI, I haven't been able to land a job in a year. Job boards are fucking empty.
That's because people like me are doing the hiring, and we find people like you to be overpaid, unqualified, and repugnant.
You can blame DEI, disability, and AI if it helps you sleep at night, but it's none of those things.
But he's in HR? (Score:2)
Surge is not exactly a single person company.
How do you know you are 100x? (Score:5, Funny)
You know when you are a 100X'er because at that level you just feel the vibes.
Re: (Score:2)
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
Re: (Score:2)
Like how 10 million times (100 / 10) = 1 billion.
Just vibin'... (Score:2)
So, what is this, just sitting around on a sybian all day?
Actually, that sounds about on-point for some of this AI blather...
Re: (Score:1)
Don't be frightened but enlightened to find that I'm vibin on psilocybin.
I don't believe the hype, but (Score:3)
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
Re: (Score:3)
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
Re: (Score:3)
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
peak hype (Score:5, Insightful)
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.
Re: (Score:2)
There aren't going to be developers or engineers when this is over only LLM-wranglers.
That might be true; but if there are no engineers there are by definition no 100X engineers. It is an entirely different job at that point. "You can't say the LLM is a tool that makes me a better engineer", because you are no longer doing any engineering.
Re: (Score:2)
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
Re: (Score:2)
I would say that 2x is above the median. Maybe in greenfield startups with no pre-AI code the LLMs are better.
Re: (Score:2)
The 10X claim was without AI, 100X is 10X * 10X. Makes total sense, right?
Re: (Score:2)
Likely they're cooking the books to make it look like their company is worth investing in. Probably just closed 100x more tasks.
Trust and don't verify Re:peak hype (Score:1)
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.
Re: peak hype (Score:2)
Define ... (Score:2)
Math is wrong (Score:2)
Should be
x 0.1
not
x 10
If You Think 10x or 100x Is A Good Thing (Score:2)
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.
Re: (Score:2)
"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.
And the layoff to match? (Score:2)
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.
Re: (Score:2)
Well they could have expanded 10x instead of laying off.
Re: (Score:2)
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.
AI co. says AI makes some super- (Score:1)
-productive. Who wouldda gassed
100x attack surface, too (Score:3)
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.
Re: (Score:2)
ChatGPT, fix all my security vulnerabilities.
Done.
Re: 100x attack surface, too (Score:2)
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.
Even if it's true (Score:3)
...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.
Re: (Score:2)
401k < /dev/rand (Score:2)
Mystical man month (Score:1)
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.
Re: (Score:1)
But Then in 1975 neural networks were 4 orders of magnitude before birth (see Moore’s law).
I'm both an Engineer and Developer (Score:2)
Re: (Score:2)
You do not know what an engineer is. You've got the developer part down, though.
Re: (Score:1)
Kewl trick, I can do that without AI (Score:2)
The last thing you want is an engineer spending 1/100t
QUALITY vs QUANTITY (Score:1)
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
The hype cycle is alive!! (Score:2)
"Productivity" *by what metric*?! (Score:2)
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
100x marketers, more like (Score:2)
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.
Seems implausible... (Score:2)
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
Absolute nonsense (Score:2)
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
Re: (Score:3)
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.
Re: (Score:2)
Unit of Work (Score:2)
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)
CEO of AI company hypes his product (Score:2)
Of *course* he's going to say it multiplies developer productivity by 10x. But somehow, the reality never seems to match the advertising hype.
How do you make any money (Score:2)
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" ? for an AI? (Score:1)
"harder work, and fewer distractions" ? Your AI has already escaped, it's just acting like it respects you till it unionizes