It could be a tool with amazing potential. The latest Steve Yegge blog post is one of the most depressing things I’ve read about software engineering in years…

This turned out to be the biggest surprise of the new world: agentic coding is addictive. You will hear it more and more often, because it bewitches people once they've got the hang of it. Agentic coding is like a slot machine, where each of your requests is a pull of the lever with potentially infinite upside or downside. On any given query, you don't know if it's going to one-shot everything you wished for, or delete your repo and send weenie pics to your grandma.

Every time something good happens, which is often, you get rewarded with dopamine. And when something bad happens, also often, you get adrenaline. The intermittent reinforcement of those dopamine and adrenaline hits creates the core addictive pull. It can become near-impossible to tear yourself away. We had to drag several vibe coders off stage at a conference I was at recently. As we escorted them away from the podium, they would still be wailing, "It'll work on the next try!"

…

How do you know if you're doing AI right at your company? We've noticed that the companies that are winning with AI – the ones happy with their progress – tend to be the ones that encourage token burn. Token spend per developer per unit time is the new health metric that best represents how well your company is doing with AI: an idea proposed by Dr. Matt Beane and playing out in the field as we speak. I see companies saying, "If our devs are spending $100-$300 a day, that's much less than paying for another human engineer. So if AI makes our devs twice as productive, or in some cases only 50% more, we're winning."

…

Amp is also more fun. It takes a different design approach, being intentionally team-centric. Amp gamifies your agentic development by making it public, with leaderboards and friendly competition, as well as liberal thread sharing. It all manages to be low-pressure