Hacker News Reader

Comment by raesene9 | Hacker News Reader

I really don't understand why CEOs like this who are AI boosters try to justify the idea that this will create more developers, when that seems exceedingly unlikely, even if they are correct in their predictions of productivity increases.

The quote from the article "if you 10x a single developer, then 10 developers can do 100x." implies that companies have 100x the things to be developed productively, what if that isn't the case?

What if companies actually have about as many systems as they need, if you really can 10x your existing developers, then that would predicate a cut not an increase.

Of course the follow on from that, for suppliers who have per-seat licensing is that they'll need to find some other way to monetize if there are fewer seats to be sold, I guess they could start charging AI Agents as "developer seats"....

View on HN

This line is the difference between your thinking and people like them:

> What if companies actually have about as many systems as they need

They think ~"companies don't have as many systems as they need".

My opinion? Even without GenAI, our sector has too many cooks.

I've not valued most of the stuff that was added to any consumer facing software product or service in the last decade or so. Facebook has, if anything, regressed since 2007. YouTube (itself, no comment on the content) peaked sometime around 2016. SwiftUI (2019) still isn't an improvement over UIKit. Google Maps is very marginally worse than it was in 2018. MacOS has improved so slowly since it was named after cats (2012) that it's hard to even care about the aggregated changes since then (even if I ignore all the bugs*), and similarly for Google Translate since they bought and integrated Word Lens in 2014.

Entirely new things are sometimes (like ChatGPT or Stable Diffusion) genuinely interesting and make a novel departure from what came before, but it's very rare.

* https://benwheatley.github.io/blog/2025/06/19-15.56.44.html

macOS has very much changed for the worse since those days imo. System preferences is a cluttered mess, they've made it nigh impossible to run unsigned binaries, they've been changing CLI tools without documenting them which has broken some of my build scripts.

Though we got dark mode, so, maybe actually it is better

> Though we got dark mode, so, maybe actually it is better

One of the bugs I didn't list in that blog post, was that dark mode sometimes only partially activates — I can get some (not even all!) dark mode icons on an otherwise light mode display.

There is a mountain of code that needs to be written that can’t be due to costs. A project that needs 100 developers may be prohibitively expensive. But 10 10x developers would be within budget. Think health care, manufacturing, government, finance, etc.

There is also a mountain of code that shouldn't have been written, and increasingly it's Ai generated.

> what if that isn't the case?

Why wouldn't it be? I believe programming is very much susceptible to Jevons paradox. I'm coding things I never would have started working on a year ago because it either would have taken too long or would have been simply impossible without an LLM (like and MCP server).

I'm not saying more code won't be produced, I'm saying companies might only want and make money off a certain number of systems with a given amount of code, and therefore they'll need to employ fewer developers to create that code, if the productivity enhancements are borne out.

For example, say you're the CEO of one of the many many companies that view development costs as an overhead. Your CTO comes to you and tells you "hey we've just 10x the productivity of all our developers", do you think that conversation leads to the CEO saying "great hire more developers"?

To make it even more explicit, if the company has 100 developers, the CEO’s response will be “Great! Layoff 90 developers.”

Great, more work for us as consultants and security researchers when your vibe coded system gets its api keys on a public s3 bucket.

(speaking at least for myself) I'm not suggesting this is a good outcome, more that having seen the way many companies handle development and IT, it's quite a likely one.

FWIW I'd agree that (for now at least) it's challenging to get LLM/AI created code to avoid security vulnerabilities without careful prompting and review. Whether that's a fixable problem or whether it'll just generate mountains of new CVEs, bug bounties and ransomware attacks, remains to be seen.

> implies that companies have 100x the things to be developed productively, what if that isn't the case?

Good point. I think that's demonstrably already a problem. Software products are notoriously prone to enshittification in part because successful companies build up vast development teams that have to somehow justify their existence with continuous feature churn.

Reply on news.ycombinator.com