julvo

Fair enough. In that case small models like devstral [1] are probably your best bet

[1] https://mistral.ai/news/devstral

Do you store code on GitHub? If so, how is GH's guarantee to not use your code different from Cursor's (with privacy mode enabled)?

Looks great! What's your experience of using this for working on real world production code?

There's a pricing link at the top of the page, we should probably try to make this more visible: https://www.wondercraft.ai/pricing

wouldn't salting and hashing be enough for this use case if you keep the salt on the client?

Not saying that LLMs are particularly data efficient, but in their defence, they don't only learn what a human learns in a lifetime but throughout the whole evolution. There may be information encoded in our genes that LLMs need to learn from the training data

Yes 100%. Using the vim bindings for VS Code and that works great. Makes working without touching the mouse really enjoyable. Also being familiar with vim is a great fallback option if you're sshed into a remote machine.

> Now we're creating tools to jump directly to the final product

The same could've been said when the printing press or highly automated manufacturing was invented

Looks amazing! Interesting to see the fully integrated approach. Working on gitpaper.com, which takes the bring-your-own-static-site-generator (byossg TM) route.

Really like the idea of keeping content in git, especially for smaller projects.

Time will tell if we need symbolic representations or if continuous ones are sufficient. In the meantime, it would be more productive to present alternative methods or at least benchmarks where deep learning models are outperformed, instead of arguing about who said what first and criticising without offering quantitative evidence or alternatives

To add to the list of very handy footguns:

@reloading to hot-reload a function from source before every invocation (https://github.com/julvo/reloading)

Sorry, I realise my comment may have sounded a bit harsh, it was not meant as a criticism of your website - just describing why I don't perceive it as minimalist design

Comparing Gwern's to your website is really interesting. I'd argue, Gwern's website is a good example of minimalist design, whereas the website you posted is the opposite.

To me, minimalist design is as little design as possible (Dieter Rams). All design elements need to serve and enhance usability/functionality of the product. That doesn't mean that the total functionality of something needs to be minimal. E.g. you can still have tags, recommended articles and a comment section.

In the website you posted, I see design elements that don't serve the usability of the website, but are there just for design sake. E.g. the background image, the colour of the toolbar or the right alignment of the text

Tangential, but I wish there was open traffic data with high coverage for planning low-traffic cycling routes. Looked at openly available aerial images, but the resolution is not high enough to detect traffic

Hey author of reloading here - cool to see this mentioned. Would love to know how reloading could be more usable for you.

Since the post on HN in 2019, reloading can also be used as a decorator to load functions from source before execution. For some use cases that may work better than the reloading loop.

Wrote a similar thing [1]. It can also load loop bodies from source before each iteration. Can be helpful for long running loops like when downloading or training deep learning models.

[1] https://github.com/julvo/reloading

That's right. Here's a handy chart with common sensor sizes: https://en.wikipedia.org/wiki/Full-frame_DSLR#/media/File:Se...

E.g. on a Canon 450D (crop factor 1.6) the equivalent of a 50mm full frame lens is a 30mm lens

Sounds intriguing. What would be the primitives of such a language? Will it be similar to a formal system?

Joscha Bach was for sure one of my favourite discoveries this year. He seems like someone who has deeply thought about many disciplines including CS, philosophy and psychology

We could add Deep Learning to the list as "everything is a vector".

Love how the article can be interpreted as making statements about itself. The article states everything on this list is an implementation of 'everything is an X'. Or put differently, everything is an X, where X = 'everything is an X'.

> Why on earth use a dynamic language like Ruby or Python and then try to bolt types on top

Probably using the languages for the ecosystem (e.g. Python for scientific computing or ML and ruby for ruby on rails) but still wanting to benefit from type checking

Firstly, congrats on the launch! Active learning is a super interesting space.

You say it's possible to download the data and use Humanloop for annotation only while still benefitting from active learning. I'm curious about your experience with how much active learning depends on the model. Are the examples that the online model selects for labelling generally also the most useful ones for a different model trained offline?

Ah that's better, thanks. here a link for others who are reading this https://gobyexample.com/errors

The problem with this is, that the caller of the function which returns what you wrote (i.e. a dynamic error message) can't match the error anymore for conditional error handling. That's because errors in go are just strings. I guess a solution could be to provide an error matching function but that seems quite cumbersome compared to typed errors.

Depends on the model. E.g. image enhancement or super resolution models are sensitive to the camera model and can be trained to fix artifacts introduced by specific cameras.

Another option could be to add a search box to the landing page which when submitted takes you to the 'add' page

Thanks for the reply - that's great. I'm on mobile and looked for 'search', but now I see how to do it.

Neat idea, this could become a valuable resource for discovering interesting things.

From a UI perspective: Maybe there could be a searchbox for books, songs etc. so that one can quickly enter things one likes. With the current system of entering preferences for the 3 suggested items, the problem is that I don't know many of the items and so I can't enter a preference.

Have a look at https://github.com/julvo/reloading, it's a little library I wrote that can give you edit-code-and-continue-debugging in Python loops. A more general solution for this sort of development in Python would be neat though.

URLs are compact as in a single string they specify everything you need to locate and use the dependency and you don't need any context, like a specific dependency manager (incl some sort of dependency file). I guess I mean compact as in self-contained, not necessarily short.