If AI is so good, why don't we have an infinite supply of 10x engineers? | Hacker News Reader

If AI is so good, why don't we have an infinite supply of 10x engineers?

10 pointsYounesDz11 comments

I read here a lot about (maybe a mythical creature) called 10x engineer and today it occurred to me the above question.

Interesting question. We also heard about how "apps" were going to disappear, and we'd have an "AI" driven slurry that configured itself intelligently to whatever the task at hand required. Or in the alternative, a flood of really great and helpful "apps" would flood app stores, all vibe coded by non-technical people who had really great ideas. I myself wonder why we still have compilers, VMs and interpreters. Shouldn't "AI" produce a kind of object oriented, strictly typed, functional, array language PHP, where a developer could write whatever kind of syntax(es) in whatever mode. "AI" would just figure it out on the fly.

The answer to all of this not happening is that "AI" isn't that good, of course, and we're being sold a bill of goods.

We get a few not infinite. with a Claude 5x Max subscription I get 3-5 workable parallel agents that themselves spawn subagents to complete work in a ~3 hour window before tokens are spent. I imagine with 20x Max that could be increased to maybe 10 within a 5 hour window? so if you want infinite you're also talking infinite money. Working in parallel takes me personally to a higher plane of output than that of a 10x engineer as it pertains to fulfilling requirements.

In the must general terms, how do you verify that the code produced by agents does what you intend it to do?

How do you approach about this problem?

Modern agents can write tests that are meaningful and require the agent to pass them with any change to avoid regressions. Humans can review the code/test the downstream application to ensure it works as intended.

It rarely takes a single prompt to get something done, but the agents can figure out as long as the human is specific about what constitutes accurate.

you can't use the thing that doesn't quite work to ensure that the output of the thing that doesn't quite work works

It's counterintuitive yes, but you can. You can just look at the tests to ensure they're consistent and with the latest models that has always been the case, and it's very rare that the models try to cheat the tests.

you can convince yourself it can, by all means. But it doesn't make it true. In fact we even have this rule for apecoding: a developer cannot review their own code.

Via Doctorow a day ago:

https://singletrackworld.com/2018/01/collision-course-why-th...

Something about cycling?

did your paycheck match your new found productivity?

Hasn't there always been a significant surplus of 10x performers in excess of positions that pay 10x?

Reply on news.ycombinator.com