cl3misch

I payed 1$ in July 2024 to reserve a launch-day discount. I never received any emails from Gouach after the receipt. You moved from Kickstarter to you own website.

Are you actually shipping to costumers?

I was super excited back then about your product and company, but the rebrandings and lack of communication made me wary.

But why would you use a motor to spin a rotor instead of a submerged propeller?

Not OP. I prefer JAX for non-AI tasks in scientific computing because of the different mental model than PyTorch. In JAX, you think about functions and gradients of functions. In PyTorch you think about tensors which accumulate a gradient while being manipulated through functions. JAX just suits my way of thinking much better.

I also like that jax.jit forces you to write "functional" functions free of side effects or inplace array updates. It might feel weird at first (and not every algorithm is suited for this style) but ultimately it leads to clearer and faster code.

I am surprised that JIT in PyTorch gets so little attention. Maybe it's less impactful for PyTorch's usual usecase of large networks, as opposed to general scientific computing?

This advice could be translated 1:1 to ADHD (at least in my experience). You have to be very mindful with your attention and energy levels.

I'm wondering if some underlying mechanism in the brain is similar between having ADHD and having suffered a stroke. Or maybe it's just the conscious effort how to handle the symptoms that's similar.

> This is a great way to [...] work on server-side files without laggy typing or manual copying

This is the usecase mentioned in the article and it wouldn't work with a bare repo. But if the server your SSH'ing to is just a central point to sync code across machines, then you're right: multiple hoops mentioned in the article are solved by having the central repo bare.

Do you have a guide or reference to follow?

How do you get the feeling this is AI-generated? Because of the many bullet points?

nit: normal PDFs may embed fonts, but PDF/A must embed them, and PDF/A is not the default in most programs.

In practice however most programs seem to include fonts in exported PDFs?

Thank you for sending me into a Habbo Hotel rabbit hole of nostalgia! They have a classic mode apparently, which is even on Steam now.

https://store.steampowered.com/app/3809900/Habbo_Hotel_Origi...

It's not usable because the Forth interpreter doesn't work for the interactive examples.

> The workplace of the coworker I built this for is truly offline, with no potential for LAN or wifi, so to download new models and update the system periodically I need to go pick it up from him and take it home.

I'm surprised that a "truly offline" workplace allows servers to be taken home and being connected to the internet.

For me it's working very well, now even using my own git remote in tailscale.

This app is keeping me on iOS as there is no single-app replacement on Android afaik.

> it's difficult to cool the whole house or even a single room by leaving the freezer door open

I think cooling with an open freezer is impossible in general? Or is that your point and I don't understand the argument?

Maybe the old farmers and rednecks don't have blogs which you can read online. They might well tell their friends and family!

That's almost exactly the beginning of Logic's story album "Everybody" which I have listened to so often that I can almost recite it. Despite years on HN I have never seen The Egg though. That blew my mind a bit, thanks!

I read the blog post. Then I thought "surely the eels in my local southern German lakes can't be from the sea". But sure enough, the European eel hatches close to the Bahamas.

I audibly wtf'ed multiple times while going down this rabbit hole. Thanks!

https://en.wikipedia.org/wiki/European_eel

A tmpfs itself is basically a ramdisk by definition. I assume you mean /tmp when you say tmpfs?

The previous comment highlights an example where backprop is confused with "a supervised learning algorithm".

My comment was about "confusing backpropagation with gradient descent (or any optimizer)."

For me the connection is pretty clear? The core issue is confusing backprop with minimization. The cited article mentioning supervised learning specifically doesn't take away from that.

I actually see this a lot: confusing backpropagation with gradient descent (or any optimizer). Backprop is just a way to compute the gradients of the weights with respect to the cost function, not an algorithm to minimize the cost function wrt. the weights.

I guess giving the (mathematically) simple principle of computing a gradient with the chain rule the fancy name "backpropagation" comes from the early days of AI where the computers were much less powerful and this seemed less obvious?

In the entropy implementation:

    return -np.sum(p * np.log(p, where=p > 0))

Using `where` in ufuncs like log results in the output being uninitialized (undefined) at the locations where the condition is not met. Summing over that array will return incorrect results for sure.

Better would be e.g.

    return -np.sum((p * np.log(p))[p > 0])

Also, the cross entropy code doesn't match the equation. And, as explained in the comment below the post, Ax+b is not a linear operation but affine (because of the +b).

Overall it seems like an imprecise post to me. Not bad, but not stringent enough to serve as a reference.

HN hug of death?

Not that I suspect maliciousness in the case of digipaws or OP, but does the app's code being open-source actually guarantee any security? Is there anything forcing the app I download to be consistent with the repo on Github?

Thanks for the elaborate answer. Reading about Schwind AMARIS it seems to be using SmartSurf which claims faster healing than conventional TransPRK. I can't assess how much of this is marketing speech or if it's basically the same as TransPRK.

In any case, I agree that a second opinion would be most helpful.

To my knowledge LASIK can correct astigmatism. A quick Google search confirms this pretty clearly. Do you have a source that claims otherwise?

> As the epithelium does regrow naturally however, [PRK] is less likely (both in theory and in medical literature) to lead to dry eye

This is exactly opposite to how I understood and experienced healing after Trans-PRK. My eyes are still very dry 6 months post surgery. Being at the upper limit of Trans-PRK yourself, did you actually go through with the surgery? If yes, how was your healing process? I would be very interested in chatting about it, since I am not particularly happy with mine, and wondering what can be done (and whom to blame).

I wondered the same. Similarly I wish the author had provided examples for statements like

> Numpy also needs to be paired with a JIT compiler to make python a real array language

I think none of the languages in the linked article are, and that's not what it's about as sibling comments point out.

> You cannot export /import from google authenticator

You can, at least on iOS. You screenshot the QR code(s) in GAuth and read them into Proton Authenticator.

In desktop Firefox it maxes out one core while OP's link is the active tab. When the tab is in the background, CPU usage looks negligible.

> In NumPy, AB works if A and B are both scalar. Or if A is 5×1×6 and B is 5×1×6×1. But not if A is 1×5×6 and B is 1×5×6×1. Huh?

> So, I removed it. In DumPy you can only do AB if one of A or B is scalar or A and B have exactly the same shape. That’s it, anything else raises an error. Instead, use indices, so it’s clear what you’re doing.

I also dislike the seemingly inconsistent adding of last axes with length 1.

I do like the capability that existing axes of length 1 will be broadcasted to any length of the other array in the operation, and use that all the time.

If I understand the text correctly, the author removed that as well.