sanderjd

I hope this is true! But I don't think I'd make a bet that we won't have a big breach like this of a hyperscalar in the next 6 months...

Why do we think hyperscalars are likely to be immune to this?

I don't know how that utilization rate and electricity share is being calculated, but my guess is that this is because nearly all current batteries are 2-hour or less, so even if they fully discharge each day, there is a pretty low upper bound on their utilization.

But they don't shine in that sustained discharge use case. They are most useful for helping manage small fluctuations, and for time shifting energy to the daily peaks.

So I just think utilization is a less useful way to think about battery storage. Unless you're comparing them to gas peakers specifically. (That is, separating out the peakers from base load gas.)

Not particularly, relative to producing anything, including turbines and the fuel to run them. (And then you have to burn that fuel...)

How are you defining "slow" in this comment?

My understanding is that batteries are demonstrating longer useful lifetimes than was anticipated.

Ultimately I think diversity in technologies is the best policy, but difficult to set up the incentives to achieve it.

Fair. But a big part of what I disliked was the tedium, which is largely alleviated. I quite enjoy making sure things I'm building work properly, now.

I understand your reasoning but my reaction to this suggestion is very strong revulsion :) On both sides of it, I would despise this process.

But you're right about thinking in terms of group productivity and costs and incentives.

Yes, there is a genuine divide between people drawn to programming for the art and people drawn to it for the practical. AI has impacted the former much more intensely and negatively than the latter. It's really a bummer.

Yeah the review fatigue thing is a huge problem at the moment. When your work becomes 90% reviewing AI code, it is very demotivating.

I have been focusing a lot more on testing and "manual" validation than on reviewing individual lines of code. (And on setting up rules for what the lines of code should look like that are easily checked with deterministic tools or by llms.)

Yeah this tracks my own experience. AI has amplified both my ability to do "just do it fast and suffer the knowledge that it isn't great" and my default toward endless tweaking perfectionism. I can just give the AI the requirements and ignore the results, but I can also do infinite rounds of human and adversarial AI review and validation. Both things are easier now.

But would doing any of that be anything more than a neat hobby? I don't mean that in a disparaging way, I'm genuinely curious.

No, the article is about what can be accomplished now vs what could be accomplished previously. That it is much cheaper to run thousands of agents at once than to have thousands of employees is directly relevant to that question.

We didn't go off topic at all. You introduced the analogy to hiring a large number of people. That is the right analogy. But the reason that the conclusion is different in the two cases is that it is much cheaper to run a large number of agents in parallel than it is to hire a large number of people. Your analogy illustrates the article's point rather than contradicting it.

It is very cheap relative to hiring people (you introduced the analogy to a company hiring a large number of people).

Yes, but the cheap parallelism is (part of) the point he's making.