For this particular experiment, regardless of phrasing, I think the guys with the most appetite for risk have to be Cloudflare. They're shipping at an astonishing pace but I think there have been far more outages than there were before in jgc era. Perhaps Anthropic's application side teams are faster and more cowboy[0] but they are super AI-native so that makes sense.

0: I think this is the eras cowboys win so they're (unsurprisingly) smart about doing this

All right, I think perhaps then I'm smearing all the sci-fi books I read in my teenage into one. Thanks.

The sequels are pulpy and quite sleazy to be honest. I read them some decades ago but there are ex-beauty-queens in a tiny human colony who must have sex with everyone else to keep the population going or some such stuff. You moved from top-grade cosmic level thought to whether X or Y is sleeping with Z. It's not that the subject is not meaningful. It's just like if you were reading about WW2 in some book and the first part talks about Hitler's invasion of Poland in a strategic sense and then everything else is about the affairs among the officers' wives or something.

In my case, what I like to do is extract data into machine-readable format and then once the data is appropriately modeled, further actions can use programmatic means to analyze. As an example, I also used Claude Code on my taxes:

1. I keep all my accounts in accounting software (originally Wave, then beancount)

2. Because the machinery is all in programmatically queriable means, the data is not in token-space, only the schema and logic

I then use tax software to prep my professional and personal returns. The LLM acts as a validator, and ensures I've done my accounts right. I have `jmap` pull my mail via IMAP, my Mercury account via a read-only transactions-only token and then I let it compare against my beancount records to make sure I've accounted for things correctly.

For the most part, you want it to be handling very little arithmetic in token-space though the SOTA models can do it pretty flawlessly. I did notice that they would occasionally make arithmetic errors in numerical comparison, but when using them as an assistant you're not using them directly but as a hypothesis generator and a checker tool and if you ask it to write out the reasoning it's pretty damned good.

For me Opus 4.6 in Claude Code was remarkable for this use-case. These days, I just run `,cc accounts` and then look at the newly added accounts in fava and compare with Mercury. This is one of those tedious-to-enter trivial-to-verify use-cases that they excel at.

To be honest, I was fine using Wave, but without machine-access it's software that's dead to me.

Well, the license change sounds pretty strange, but to be honest if I were to use this software I would use it without adhering to the MIT. It's machine-created content which is not, in general, copyrightable. You can assert whatever license you want on such content, but I am not going to adhere to it. For example, I declare you may use the following under the Elastic License

    The