I don’t like that they reversed the the comment thread order but didn’t include a native way to include previews. Now instead of having to scroll to bottom to see the merge button. I have to scroll to the bottom to see the preview.
I agree, is the 100th rest api I’ve made for my 5th employer art? I don’t think so. It’s just a way to (maybe) make my employer money.
I’d say the brainfuck language on the other hand is art. Hard to draw a fine line as to which is which.
I can say most if not all the projects I’ve ever made land in the former. I’ve made some clever solutions but nothing approaching something with meaning.
Is it not a problem of understanding and searching for complex problems to solve, and therefore understand?
If you've written 100 REST APIs, you're probably pretty familiar with them. But Brainfuck is completely alien and something you would have to work to understand. Maybe the distinction is in the effort required for understanding.
Regardless, I think meaning is something one has to self-impose. It does not exist naturally in the universe, so if you don't think it matters, it doesn't matter.
I agree that meaning is something one has to self-impose, but there is also meaning in the creation of art. I don’t find that same meaning in work that accomplishes the goal of making an employer money in the same way I’d incorporate meaning into a short story I’d write.
I think what I’m trying to say is that there are both ways to create art by coding and ways to just make a product for consumption. And I think using AI for said products doesn’t really reduce the meaning of said work.
I think I gave brainfuck as an example because it seems there was a meaning behind it beyond being profitable. The writer had something to say. Does writing something like brainfuck with AI take away its meaning? I’d personally say yes, but that’s an interpretation.
There can be beauty in even the smallest of things. You can tell alot about a person by how they choose to structure even down to the loops and methods.
Good on you for doing this and I hope you get promoted. But I can imagine a lot of people are like me and just don’t want to engage in office politics on or offline. Not that I don’t sometimes, but I’d much rather just work and be paid to do a job and live my life.
But you still have to share the undodir for both vim and neovim right? If your undo data was that precious why would you gamble the interoperability? You could just have them in different folders and it would be fine.
So the case is, you really need you undo history and want to try out neovim so you just copy paste you vimrc to the new place and accidentally delete your data? I don’t think it’s that bad personally. It’s bound to happen but it’s not automatic.
That headline would be inaccurate. Persistent undo is different than cache. Cache implies a temporary copy of data whose actual existence is elsewhere.
Persistent undo is advertised as persistent, not as temporary.
In practice the feature is as persistent as your last git checkout. There are countless events (including git checkouts, vim recoveries) that simply reset the status of the "persistent" undo. Use a real VCS if you treasure your history as this is just a best effort feature.
Do you have a link to how this happened? From what I understand you’d have to point the undodir at the same place as vim for this to happen? It’s not this way by default and you have to put a copy of the exact same thing in your neovim vimrc right without understanding that it would do this?
To be fair to the pnpm maintainers, they have been working on this well before the bun AI rewrite. The first commit in their rust experiment is in 2021.
> In this report, we argue that frontier performance can be achieved by a wide range of institutions through Continual Learning on readily available open-weight models.
> As opposed to existing limited approaches such as small-scale fine-tuning, prompt engineering, or tool-augmentation with a frozen model, our Continual Learning approach takes advantage of the effectiveness of a modern mid- & post-training stack while introducing safeguards preserving both plasticity and stability at each training stage and seeking to make the minimal number of high-impact interventions on the parameters.
For the large model, Thomson is utilizing the fine tuning stack they describe in the article, running it on Snowdon 1.0-Large, which in turn is a fine tune of Qwen3.5 397B. Same thing for the small model, but it's a fine tune of Snowdon 1.1-Small, which is a fine tune of Qwen3.6 35B.
As for the small version's run:
> The full pipeline consumed approximately 1.63 × 10²³ FLOP over 35,207 B200 GPU-hours, showing that these results are achievable with compute and personnel budgets substantially lower than commonly thought.
That would amount to around a quarter to half a million dollars of spend on that run. 100k minimum, if they got a great deal.
reply