I find that just about everybody is leaking my data. Either on purpose or accidentally. But at the end of the day I do want to order those online exotic vegetables and have very little choice of sites to do it in my country. At this point I don't care anymore. Maybe I should but in a sense it's too late. It's been decades of leaked information, I'm certain the advertising companies know me better than my friends. And still royally fail at selling me stuff. They likely use this info for more nefarious things.
I hate that 90% of the effort on the internet is about stealing information from users and serving invasive ads.
I don't quite disagree but this comparison is typically unfair, because when you really know about a subject you tend to ask way more difficult questions than about other subjects, so of course the LLMs are gonna struggle more. If you ask really basic questions they will regurgitate well known bachelor-level knowledge and look good. What do I know about biology anyway? about silos for grain storage? any passable answer is enough to wow me on those topics. But on the topics I really know about, I never ask the basics.
Check out their YouTube channel where they show plenty of interesting features. But just to list some I can think of:
- optional reactivity (i.e. you create chain of cells where editing 5th cells in the past causes update down the stream, pretty neat when working with dataframes). Its reactivity is a very cool feature once used to but you might not want it for something like running heavy ML training task so it can be toggled off
- you can switch notebook to multi-column notebook mode
- notebook is a web app that has sidebar with a lot of menus, there cool sections like Docs, Packages (you can download new packages right away there with uv), plenty of LLM integration with their custom prompts where you can reference dataframes so that it would be able to understand schema, some SQL and other DB integrations as well, cells can even contain SQL instead python code and output query result into python variable
- thanks to reactivity it got a lot of interactive elements like sliders buttons text fields or ability to create entire own widgets, there's even mode where all code blocks get hidden and you're left with complete app
- you can make web export of notebook that will translate python to WASM and publish it as fully working static page (though publishing something heavy complex like torch probably won't go well), this fits well with previous point as you can basically build simple interface hide all the code and publish it (like imagine matplotlib with couple of sliders)
- DataFrames (pandas/polars) displayed as interactive tables where you can filter by columns, scroll through pages of rows etc
- notebook stored in a .py format, unlike .ipynb with its json like structure. So code is very Git-friendly but you don't store computation results anymore
It's not blame, it's useful feedback. For a large application you have to understand what different parts are doing and how everything is put together, otherwise no amount of tools will save you.
It's very difficult. It took ASML 20 years, and Apple has none of the core competences to make this happen, like optical lithography, EUV optics, plasma physics, vacuum, laser, sources...and then they would have to catch-up to the other tech. For example, today's top end ASML stages accelerate with >10g while still having nm position accuracy.
As I feared, they seem to be using D3.js "by hand", which is the only way to achieve such high quality but needs IMO an enormous amount of work and dedication. No wonder they are offline, that's hard to sustain. I hope a technology comes that makes it significantly easier to make such interactive charts. I know there are many out there, but there's always a "but".
I hate that 90% of the effort on the internet is about stealing information from users and serving invasive ads.
reply