New technology in programming and poker

12 Feb, 2026

Work feels different now. Steve Yegge captured something important in an interview with Gergely Orosz:

There’s a vampiric effect with AI, where it gets you excited and you work really hard, and you’re capturing a ton of value. I find myself napping during the day [...] Let’s say an engineer can be a hundred times more productive; who gets to capture all that value?

Yegge is focused on the ethical-economic question of value capture, but I'm interested here in the value creation. What is it like to create that much value by working longer at draining work?

The work feels different, and you can reasonably do a lot more of it. Your B- and C-games get more valuable, because the activation energy required to do something useful is, I suspect, a lot lower.

The situation reminds me of the Internet revolution in poker. The dominant structure of the game changed from sitting around a table and mostly folding (no-limit hold'em had famously been characterized as "hours of boredom, punctuated by moments of terror") to playing many tables--4, 8, 12, 24, or more--online, simultaneously. There are many similarities:

1. The game became interrupt-driven: rather than paying continuous attention to a single game¹, you switched from game to game as they pinged you.
2. Different forms of poker, played with different stack sizes and sometimes rules, emerged.
3. Whole new categories of skills (e.g., in game selection) emerged.
4. You could play meaningful poker if you had even a few minutes.

I could go on. Many people resisted the change:

1. Some players just didn't like it and never tried it.
2. Others tried it but never really had their hearts in it.
3. Others really did try, but never let go of the idea that Internet poker would be like live poker, just on a screen. They kept the same priorities and strategies.
4. Even among those who bought in, it simply never occurred to many of them to try to play 12 or 20 tables at a time.
5. And even among the more open-minded ones, few had the discipline to drill the new skills that the new styles of play required.
6. Fewer still undertook health regimens that they needed for peak performance at this new kind of endurance skill.

I shouldn't overstate the similarities. Most obviously, agentic programmers should be trying to automate away much of the repetitive clicking.² We also need long stretches of pure planning that you don't get in poker, even if planning is itself mediated with AI.

Still, I can't get the analogy out of my head. Here's some of what happened:

1. There were large rewards to ambition, and even to audacity. It might seem obvious that achieving 70% of your maximum winrate on 24 tables is better than 90% of it on 6, but relatively few people were willing to try it.
2. The disciplines were different, and although some players succeeded at both live and online (or transitioned from live to online), others simply couldn't. There is a richness, a history, a set of equipment, and a suite of skills specific to live poker: I'm a romantic about it. None of this entailed that every good live player would make a good online player, and certainly not a good online multitabler.
3. Physical fitness and general mental training became enormously relevant to playing poker, because endurance mattered even more and because online play required a different kind of intense sustained performance. I'm not aware of any systematic studies about this, but I think it's anecdotally obvious that fitness, diet, and training mattered a lot to online poker success.

Again, things are changing so quickly in agentic programming that I can't know what peak performance in it will look like. I'm struck, however, by how important I'm finding it to just keep going, and just keep thinking hard. Logging bugs, describing features, approving requests, figuring out what should be a skill, telling something to make the skill, looking at the code that's scrolling by: these all require effort, can all be done better and worse, and can all seem less fun than getting a coffee. (And don't underrate the importance of maintaining discernment as a user-creator: this takes effort and produces essential inputs.) For now, at least, performance in programming feels a lot more like athletic performance.