# Pencils down: what I took from DHH's Rails World 2026 keynote
By Dima Barabash · 2026-09-27
Source: https://dbarabashh.com/thoughts-and-experience/dhh-rails-world-2026-pencils-down
---
Sooo, I'm writing this post because I watched DHH's opening keynote at Rails World 2026. I found a lot of interesting things in that talk, and some of the things he said really resonated with me.
So overall this post is a kind of compact version of what David said during the talk, but through the lens of how I took it myself.
But let me start with who DHH even is.
DHH is short for David Heinemeier Hansson. He is the creator of the very well known framework [Ruby on Rails](https://www.youtube.com/watch?v=HDKUEXBF3B4), the creator of [Omarchy](https://omarchy.org/) (a Linux distribution), co-owner of [37signals](https://37signals.com/), and the author of [bestsellers](https://dhh.dk/#rework) about how to work productively and build teams. He invests in Danish startups and sits on the Shopify board. He shows up on all kinds of podcasts very often, and people sometimes react to him in pretty interesting ways, because he has fairly unusual views on a lot of things.
Still, I admire him, because you can feel that his thoughts are not AI slop and not just a remix of other people's thoughts. He says what he thinks. At least that's how I feel it, and I like that.
That's the short intro to who he is. Now let's get to what he actually talked about in the opening keynote at Rails World 2026.
## Paintings, photography and the cost of making things
David started his talk with historical analogies about how tools changed, and how the cost of producing a result changed together with them.
First he shows some paintings.
*The Ladies Waldegrave*
*The Family of Christian IX of Denmark*
And he talks about a very high level of craft, a huge amount of manual work, months or even years of effort. One of the examples is the painting of the Danish royal family, which took about three years to paint. And he stresses it: the result is wonderful, but the process itself is incredibly expensive in time and in effort overall.
Then photography shows up, and an important shift happens.
What used to require a whole master and a huge amount of time can now just be photographed. Much faster and much cheaper. And you get a photo of your family or some other portrait. You can just see the image, and you don't have to wait years for it anymore.
And the main point here is that he is not saying painting became useless or pointless. And he is not saying photography just "replaced art".
He is talking about how the role of the person changes, the value of the skill changes, and what it's even worth spending that manual work on changes.
His take ends with this: we live in a wonderful time, when the cost of creating programs and software is dropping sharply. The same way photography once sharply dropped the cost of capturing an image of reality.
## November 24, 2025
Then David talks about November 24, 2025, the day Anthropic released Claude Opus 4.5. He literally draws a line at that moment between what was "before" and what started happening "after". After that, LLM models, as he feels it, let us look at our work in a completely different way. And that's what started changing the industry. He marks this moment as a kind of reference point. A starting point.
## Algebra and Maze
Next he has a slide with a table from a 1966 paper about individual differences in programming.
Participants got two tasks, called Algebra and Maze, and then the researchers measured how much time they spent writing code, debugging, CPU time, what the program size was, and so on.
And the key point in this table is not that algebra was some especially hard concept. It's how enormous the gap was between the best and the worst programmers.
For example:
- **Debug hours, Algebra**: 170 hours for the worst vs 6 for the best. That's about a 28x difference.
- **Code hours, Algebra**: 111 vs 7. Around 16x.
- For Maze the gaps were also around 25–26x.
David uses this as an argument that programming has historically had a very large difference in productivity between people. And then this feeds into his AI analogy.
Before, a strong programmer could be roughly 10–20x more effective than a weaker one. But now AI can shrink that gap sharply, because part of the mechanical work (writing something, finding a bug, generating an implementation, explaining code) can be delegated to an LLM.
So Algebra here is just the name of one of the test tasks. But David uses the whole thing as a central metaphor. In short: we have known for a long time that there can be a gigantic difference in productivity between programmers. AI is now changing the very nature of that difference.
And the way I see it, the value of a programmer is shifting from:
"How fast can I write code by hand?"
to:
"How well do I frame the task? How well do I check the result? How well do I make architectural decisions? How well do I understand and use my tools?"
## Pencils down
And then something really interesting happens.
He just says from the stage that at his company, 37signals, they literally made the decision that they will never write code by hand again.
He declares it as: **Pencils down.**
And when he said that, I honestly got a little bit of goosebumps. Because it's a strange thing to feel when you built your career around learning to do exactly this by hand.
But still, it's a cool time. Like he said at the very beginning: a wonderful time to be alive and watch what's happening.
After that he asked the room, straight from the stage, how many people there still mostly write code by hand.
About five people raised their hands.
And his reaction:
> "Holy fuck, it's about 5"
>
> — DHH
This is a pretty big conference. I'm more than sure his talk was just sold out, there were a lot of people.
And only about five people raised their hand.
This is just unbelievable. That's how big the change that has already happened is.
Then he goes as far as announcing that he is "retiring" from programming, because programming in the form it used to be is no longer relevant for him.
In English it sounds especially strong:
> "I am a professionally retired programmer now."
>
> — DHH
## HEY never wanted to be web
After that he moves to a concrete example from 37signals products. He talks about HEY, their email client, which they sell to users as software.
HEY is a web app. But he says that HEY actually never wanted to be web. It's just that web used to be the thing that let you build productively and fast. And native apps are, well, native apps: great latency, great performance. And he shows how they took their codebase and basically built a Windows app. At first the UI looked, well, not very okay. But then literally by the third or fourth prompt they got something that looked awesome in the presentation. A full app that runs on Windows, with a great interface.
And the cost of it, in terms of time and effort to build, is incredibly low.
## To be continued
Lol, I'd happily keep writing, but I can already see there is a lot of text here, and there is still about half of the talk ahead.
Next he goes into, for example, the topic of Rust in a really awesome way: how he felt about this language, why he thinks it's incredibly horrible, but at the same time great if you just don't look at it and hand all the work to an AI agent.
There are a lot more interesting takes in there, and in general many things I want to write about separately. So I'll probably stop at this part and leave the rest for the next post.
***
But at the end I want to say one thing.
Probably right now, from a programming point of view, I feel like I'm in the best position ever. I can build apps and systems. I can see the trade-offs. I have a tool in the form of an LLM that can suggest something to me, and at the same time I can validate its answer and understand when it suggests something wrong. And that is incredibly awesome. I'm just thrilled about all of this.
But on the other hand, there is a bit of jitters too. I'm a little worried about how to keep growing my skills in this new world, and what exactly to focus on.
But one thing I can say for sure: networking and soft skills matter now more than ever.
[Rails World 2026 Opening Keynote - DHH](https://www.youtube.com/watch?v=vDjW_dRyKXY)