Kyle Hughes

I'll burn out on talking about agentic programming soon. What is compelling me is the same thing a lot of people are beating around the bush over: Claude Opus 4.5 was a step-change for productive autonomous programming and what feels like the single most impactful event in my software-making life. Tough to say that with grace.

In November, the value of the labor of writing code was significantly reduced, whether or not you use these tools. This is what has always been on the horizon.

Dec 30 at 5:31pmWeb
Show threadKyle HughesDec 30
To say that AI is irrelevant in the making of software because coding wasn't the important or difficult part ignores that coding was an important and difficult part and is one of two reasons it has been such a well-paying career for so long. It asserts that friction is nothing—outcomes happen regardless of cost, effects happen without cause. I assert that friction is everything. The shape of any system is defined exclusively by where friction exists. Change those locations, change the system.
Show threadKyle HughesDec 30
“Can you write this” is, largely, no longer valuable. All of the value is in “can you specify this” and “can you verify this is correct.” Those are all different skills, and the latter two were always the most valuable but always bottlenecked by the former. Software engineers made their living for the last 70 years on that bottleneck.
Show threadDaniel LyonsDec 30

@kyle Are you trying to say that software engineers will be obsolete? Cause I think Jevons paradox is making a compelling case that that is probably not going to happen in the long run.

https://simonwillison.net/tags/jevons-paradox/

Simon Willison on jevons-paradox

3 posts tagged ‘jevons-paradox’.

Simon Willison’s Weblog
Show threadKyle HughesDec 30
@dandylyons No, right now I would bet on an explosion of software. But, I think the valuable skills for making software have changed and the people in those roles presently may not have the skills, or interest in acquiring them, and we are now in a strange interim where that knowledge is not widely distributed.
Show threadDaniel LyonsDec 30

@kyle Yeah I largely agree. This seems to be like the transition from human computers to machine computers.

Computation went up not down, but humans had to find new skills and roles.

Show threadHalDec 30

@kyle my optimistic take is that the value of code and the value of labor are different things.

The value of code has not been significantly reduced. The *cost* of creating it has been significantly reduced.

The value of your labor is somewhat more esoteric and out of your control, but if you can increase the quantity or speed at which you create revenue-generating code, this seems like a good thing in terms of your value as an employee?

Show threadKyle HughesDec 30

@hal Yeah, I don’t disagree with any of that. And I’m pretty optimistic right now, doing the best and most enjoyable work of my life faster and better than ever. I think I capture that angle a bit better here: https://mister.computer/@kyle/115809829164348633

But I’m still in the arc of my posts where I’m trying to convince people there’s something here worth paying attention to. And I do think the literal labor I have done for most of my career has less value. But I can do even more valuable labor now.

Kyle Hughes (@[email protected])

“Can you write this” is, largely, no longer valuable. All of the value is in “can you specify this” and “can you verify this is correct.” Those are all different skills, and the latter two were always the most valuable but always bottlenecked by the former. Software engineers made their living for the last 70 years on that bottleneck.

Mister Computer