"The Structure of Engineering Revolutions" by @johnallsopp https://webdirections.org/blog/the-structure-of-engineering-revolutions/
I feel like I've been waiting half a year for someone to write this essay. This makes more sense out of everything I've been reading, thinking, and feeling for the past few months than just about anything else I've read.
The other weird part has been feeling the rift inside myself – the part that's a craftsman mourning the old ways, and the part that can't help but notice what's coming and is both terrified and curious about it.
I used to take code quality *very* seriously – I sweated every semicolon – but now I'm starting to wonder if the slop-slingers are seeing something I don't. That they're playing a completely different game, and there's no sense judging them by the rules of a game they're not playing.
@nolan What's missing in a lot of this conversation is the systemic effect. Same as we saw in the mad rush to React et. al., the assumption of people who say "oh, this makes X better" is that they are generally self-reporting, failing to measure baseline quality, or even do longitudinal analysis of how things net out over moderate time frames.
Quality *can* go down. And when code is cheap, guardrails are the compliment (and therefore become relatively more valuable).
@slightlyoff Oh absolutely, I've noticed a few things:
- Everyone is just going off vibes (there are maybe good reasons for this; the benchmarks don't tell you much IMO)
- Even among vibecoders there is disagreement about whether you should just be a slop cannon ("the code is a binary, the spec is source") or use agents to refine the output for higher quality (I lean this way of course).
I think a lot of unmaintainable junk is being written, and some folks are hoping the models will outrun it.
@nolan The very worst takes seem to be of the "tech debt is free" variety. Even if it's much cheaper, cheap !== free, and that difference matters *enormously* for what happens next.
My relatively strong suspicion is that paying "full freight" for models (when that happens) will not deter folks in domains where output quality is decent, but that the cost curves for both tail risks (quality, tech debt, etc.) and running costs will pull excesses back and/or orgs will accelerate straight into walls
That's nothing:
Look what happens when the US military plans wars with AI.
"Grok, plan a war with Iran that costs the least amount of money and the least number of troops..."
I appreciate there's software disasters, but that seems like the least of our problems.
@matt It probably depends where you work. At my small startup nearly everyone is adopting these tools from the bottom up, swapping tips on different models, etc. I'm basically a laggard.
I don't think it's a deficiency, I think it's human nature. I haven't switched from WebStorm for ~15 years because I came from a Java background and know all the keyboard shortcuts. I'm amazed I ever changed my workflow this much to adopt AI coding tools.
@nolan @johnallsopp There is a long history of using Kuhn to argue that something new is equally revolutionary.
Consider the old saying "First they ignore you. Then they laugh at you. Then they attack you. Then you win." Sagan commented "They laughed at Columbus, they laughed at Fulton, they laughed at the Wright brothers. But they also laughed at Bozo the Clown.”
Few things cross the chasm or follow the Gartner hype cycle.The wonder material asbestos needs remediation. Our plastic piles up.
@alesroubicek @nolan @johnallsopp Global thermonuclear war accelerates entropy. Anthropogenic global warming accelerates entropy. Sustainable living accelerates entropy. No matter how Panglossian things are, in the end we're all dead, so everything is fine, and there's no need to change anything or care.
I care.
@nolan @johnallsopp I'm not sure about the argument that I should hurry up and adopt. I don't rely on AI tools today, does that make it more difficult to start using Claude? Have I missed some vital experience from a year ago that makes me unable to adpot? How does being an early adopter of tabnine in 2021 help when using the tools of today, and how would making the tools of today make it easier to adopt the tools of tomorrow?
Things change so fast that the experience doesn't seem useful
“And yet there’s the interference pattern on the screen. Tickets get closed. Tests pass. Software ships.”
I haven’t seen this. At my day job, there is no AI integration with any ticket workflow. There is ALOT of nuance in regards to ticket backlogs. That require ALOT of conversation and organising. But the tickets themselves might be super simple, a one line change for example.
I would be interested to see how people have worked with the quoted scenario.
> This isn’t a lower standard. It’s a different standard.
How can you not see it as a lower standard? The standards of software quality were already plummeting before AI; I can't help but feel that the rise of AI-generated code is flooring the accelerator. I want more lovingly crafted apps like Pinafore, not more React monstrosities where a single page payload is bigger than Doom (as noted by @slightlyoff in his latest performance inequality gap post).
> The adoption is grassroots–developers dragging their organisations along, not the other way around.
This claim surprises me, because we've all heard of top-down mandates, like the infamous Shopify AI memo. It seems to me that top-down mandates are what many of us are afraid of, like Larry Garfield's "Selfish AI" post (https://www.garfieldtech.com/blog/selfish-ai) where he's afraid that he'll have to use the products, over his strong ethical objections, just to have a job.
@matt Speaking for myself, I am afraid. I am not even sure that adopting these tools today means that I won't be out of a job in 5 years.
I am being a bit paternalistic, though, because I find that a lot of people do seem to be in a state of willful ignorance. I was too up to like a year ago. https://nolanlawson.com/2026/02/15/days-of-miracle-and-wonder/
@matt I'm also getting a bit tired of these conversations, because I feel a bit like Kent Brockman with the ants, except the ants are real, and the only satisfaction I get from saying "Well guys, looks like we better accept these new ant overlords" is that a few years from now we'll be toiling in the mines and someone will turn to me and say, "Boy, you were sure right about those ants!" https://knowyourmeme.com/memes/i-for-one-welcome-our-new-insect-overlords
(To be clear: not tired of talking with you, Matt. 🙂)
@nolan @johnallsopp what’s the logical fallacy again where you only argue half the argument so it makes you look good?
This excellent essay is missing so many criticisms that people have with the current crop of GenAI and narrowly focusses on just the tech gains to fit a compelling narrative, it’s astounding.
“Yes many people suffered under the industrial revolution” is not an absolution to do it again, it is an imperative to never do it again.
Nolan Lawson
Alex Russell
Marius Gundersen - mdg 🌻
Alex Grant
ǝʌɐp
Matt Campbell
Andrew Dalke
Aleš Roubíček
Brian Douglas
Jan Lehnardt 
Steve Loughran