On Sun, Oct 6, 2024 at 08:13 Martin Blais <[email protected]> wrote: > On Sun, Oct 6, 2024 at 7:17 AM Chary Chary <[email protected]> wrote: > >> Martin, >> >> realistically, this is mind blowing! Also, the podcast was spot on about >> beancount, a very good introduction in any case. (the only thing I noticed, >> is that in the middle the female and the male characters swapped for a >> while their roles of being an banecount advocate and a sceptic). >> > > They're also a bit too enthusiastic all the time. > > > In few years machines will be creating movies, better than Hollywood >> (definitely cheaper)! >> > > I'd be surprised at the rate this is going if this isn't already going H1 > 2025. > > > Coming back to beancount, I am just wondering: is it possible to download >> there an old source code of bean-query and related documentation and then >> the new source code of beanquery and then to ask it to update the old >> documentation to match the new source code? >> > > I'd love to be able to say "translate all of this to Rust" (or C, or > whatever) and even have a crappy but working implementation (with tests) of > the entire code that I can fix up and refactor. >
I’ve been doing a lot of dev lately with Aider + Claude. The tech is moving this direction but it’s not there yet. One shortcoming is the models and frameworks have a limited scope. They’re good at reasoning about one function or file at a time. They start to struggle on multifile coordinated changes and entire codebase rewrites are well beyond scope. The other is the models are much better at python and JavaScript than rust. Part of this I think is the nature of Rust’s type, ownership, and lifetime rules which are complex, and you want the model to do some tool use with the rust analyzer rather than hallucinate what it thinks makes sense. Another is I guess less rust training data. A third is the long tail of rust crates and versions thereof — again the models need some explicit representation of which version of which crate is in use and then code to that API rather than hallucinate a vague memory of how to use libraries for the task at hand. > > Regards. >> >> On Saturday, October 5, 2024 at 4:48:08 PM UTC+2 [email protected] wrote: >> >>> Okay, not really. >>> >>> But I uploaded the Beancount intro to NotebookLM (Deep Dive) this >>> morning and generated this: >>> >>> https://notebooklm.google.com/notebook/16555495-485e-4977-9e3d-d853e0719c07/audio >>> >>> Here's one built from the cookbook documents: >>> >>> https://notebooklm.google.com/notebook/3e24b1e4-400d-4921-956e-572f73839505/audio >>> >>> FYI these aren't real people. >>> I honestly don't know what to think about it. >>> It's a bit all over the place in terms of relevance and fluffy and in >>> ways and wrong too in bits but it's also kind-of amazing at the same time. >>> Undoubtedly the least time-efficient way to consume this information >>> though. >>> But it's amazing that this is possible. >>> >>> -- >> You received this message because you are subscribed to the Google Groups >> "Beancount" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> To view this discussion on the web visit >> https://groups.google.com/d/msgid/beancount/30393d9b-e24e-4eaa-ba42-fd7807045be8n%40googlegroups.com >> <https://groups.google.com/d/msgid/beancount/30393d9b-e24e-4eaa-ba42-fd7807045be8n%40googlegroups.com?utm_medium=email&utm_source=footer> >> . >> > -- > You received this message because you are subscribed to the Google Groups > "Beancount" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To view this discussion on the web visit > https://groups.google.com/d/msgid/beancount/CAK21%2BhP81s%2BvTTtaihk-y3-n9fD9G2PQ0ozT45-NzOE37hNw_w%40mail.gmail.com > <https://groups.google.com/d/msgid/beancount/CAK21%2BhP81s%2BvTTtaihk-y3-n9fD9G2PQ0ozT45-NzOE37hNw_w%40mail.gmail.com?utm_medium=email&utm_source=footer> > . > -- You received this message because you are subscribed to the Google Groups "Beancount" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/beancount/CAFXPr0t3iePwuxymzR2pWXQrcj8HGunOWmqDiuXev1HxS4aP4Q%40mail.gmail.com.
