Hacker Newsnew | past | comments | ask | show | jobs | submit | httpteapot's commentslogin

What do you think of the DeepSeek OCR approach where they say that vision tokens might better compress a document than its pure text representation?

https://news.ycombinator.com/item?id=45640594

I've spent some time feeding llm with scrapped web pages and I've found that retaining some style information (text size, visibility, decoration image content) is non trivial.


Keeping some kind of style information is definitely important to understand the semantics of the webpage.


I recently purchased a MacBook, and while I'm thoroughly impressed with the hardware, I find the keyboard navigation in the desktop environment somewhat lacking in comparison to my previous experience.

Having used Linux Pop OS for many years, I've grown accustomed to the intuitive and powerful tiling window manager and keyboard navigation shortcuts it offers. I'm struggling to find a comparable solution on the macOS platform.


It's a sore-spot for sure. I've found that Rectangle[0] does a good enough job.

[0] https://rectangleapp.com/


Rectangle is so good - it's the first thing I recommend anyone new to MacOS download. I don't know why apple hasn't implemented this natively.


Magnet is also pretty good


Hammerspoon is by far the best option for devs, it’s more powerful than pretty much every solution out there, including better touch tool, and the only additional software I need is karabiner elements and alttab. Those three pieces of software will solve every single problem you have with Mac.


Does Rectangle has a shortcut to switch the focused window in a split screen?



Hammerspoon can do it for I have it in my config. I also have it so I can switch to open windows on the same Space (I can throw windows into a tiled view and jump around to them). It's extremely flexible if you are willing to play around with it.


I’ve kind of solved this. You can snap windows but you need to add a shortcut in systems to do it. It’s not perfect, but it does work and overtime you miss it less and less


I’m a very happy user of Rectangle, as well.


As a more general suggestion, macOS likely isn’t good at imitating what you previously had, try to understand the workflow they’re proposing, adapt and learn it, it generally gets second nature and pretty decent soon enough, getting to work the way you envision is not apple’s strong suit.


Classic blame the user kind of argument, what is the alternative workflow to "snap windows with a hotkey" then?


The point is that when switching OSs you are going to run into lots of places where the new OS does things differently that what you are used to. Your choice is to fight the new OS and try to keep doing things the way you are used to or embrace the new OS and learn some new workflows.

That doesn’t mean adopting everything or abandoning everything, just try to be flexible and try some different things out. Just because it is different doesn’t mean it is broken.

For window management there are dozens (at least) third party tools that add features like snap window. Try some out to see what you like. Rectangle, Magnet, Moom, BetterSnapTool come to mind but here are others.


so "3rd apps exist" that's your argument? I thought it was "alternative workflow"?


My point was a suggestion to entertain the idea of adopting alternate workflows native to the new OS when you switch. When that doesn’t work there are often 3rd party tools to address the difference.


The better general suggestion is to find good tools to make your OS life more comfortable rather than admitting defeat right away and imitating the bad old stuck ways of the OS

(it's not Apple's strong suit, but there is an app for tha™)


Sometimes that works. Sometimes you're just stuck.

For example, AFAIK, there's no way to eliminate the animation that's triggered when switching spaces/virtual desktops.


this is not an example of a "workflow they’re proposing"

(also, afaik https://github.com/koekeishiya/yabai allows you to eliminate the animation completely)


Do you know some good ressources to learn how to use macOS the productive way, using mostly keyboard shortcuts instead of the trackpad?


Apple maintains a list of macOS keyboard shortcuts that might be useful: https://support.apple.com/en-us/HT201236


You can use hammerspoon to solve almost every issue you have.


I don’t really tile windows much. The main case where I do is with a bunch of terminal windows. Most of what I get out of a tiling window manager is full screen by default and easy keys to switch workspaces.

On macOS the thing that drives me insane is that the many-finger swipe to switch desktops won’t focus the target window until the animation is totally done (like 0.7s after starting). I wish it works like cmd+tab which changes focus instantly. Apart from that I guess I’m not that bothered because I mostly just full screen things. Emacs and iterm2 can do their own tiling of windows.


I am not sure whether it applies to this case, but you might try turning on ‘Reduce motion’ under the Accessibility settings, which eliminates some animation delays.


Reduce motion changes the animation to fade-in fade-out but updating focus is still delayed


Just a heads up — you can solve that one annoyance; there are terminal commands that will reduce that animation to instantaneous.


I tried changing various defaults but lots of them seemed to no longer work. Maybe I only tried the options to increase the speed or I didn’t restart enough things or I spelled something wrong. I also tried looking at the binaries for relevant-looking names of defaults to change and didn’t find any (including the options that old internet advice recommended changing)


The keyboard shortcuts to do this animate much quicker than the swipe does. In the past I've set up BetterTouchTool to press the corresponding key on left and right swipes.

The only disadvantage is that there is no way to "peek" the previous desktop like you can with the built in swipes.


I suspect the constant emphasis on mouse use, ever since the first Macintosh, has created an attitude of "keyboard doesn't matter"; I've noticed that even early Windows is very usable with only a keyboard (the Alt, underlined letters, and arrow keys method is particularly well-designed), whereas e.g. classic MacOS is basically unusable without a mouse.

In later versions they added keyboard access, but it still feels like it was done as a bare-minimum concession and not originally planned.


I had the opposite experience with MacOS. I hold every OS I've used since the early aughts up against MacOS 9 and they are all lacking in terms of keyboard navigation. Maybe it was because I had the previous 10 years to practice, but I felt I could do almost anything in pre-OSX MacOS with the keyboard, only relying on the mouse for application-specific stuff like photo editing. Navigating, filtering, opening files and folders were all incredibly easy.

In KDE it's pretty much a joke every time I have to save-as. Can't even get through that filesystem menu without a mouse unless it supports <ctrl>-l. Dolphin is slightly better, especially if you enable the console pane to make it easier to switch to the command line, but it's still way behind Apple's finder from 1999.


but I felt I could do almost anything in pre-OSX MacOS with the keyboard

How do you open the menus and browse through them to find and select the option you want?


IIRC for the most part option-f would bring me to the file menu. I could down-arrow to scroll through it or hit the right arrow to move to the next menu. In that way I could navigate pretty well.


> has created an attitude of "keyboard doesn't matter";

I think you would be surprised at just how powerful the keyboard subsystem in osx is and how malleable it is with some programs or even just editing the plist shortcuts. every menu item in any application can be given a per app or universal shortcut in `system preferences | keyboard | shortcuts | app shortcuts`

You can also use a thing like hammerspoon to do whatever you can imagine basically too... i have a few things for window manipulations via hammerspoon.

You can set up very complex keyboard re-mappings/shortcuts using something like karabiner elements.

you can also change the keyboard access to be navigable via tab in keyboard settings too.

and you can access the menu via ctrl-f2 by default. those settings are changeable too via the keyboard preference pane.

i think you'd be surprised at how much apple cares about accessibility so keyboard nav is not just a power user thing.

At the end of the day I think apple put a lot of thought into the UI and i'd feel pretty stymied going back to windows.(though now the power utilities finally let you remap the windows key to something more useful like ctrl.


Except you can't bind every menu item: some menu names are dynamic, so you can't bind to them, and not all the keys are supported, and there is no left/right modifier differentiation. Also, you have to retype the full command path by hand, which is insane, so you can't do that either (you should be able to just hover over a command and press keybind to rebind)

Even with the awesomeness of Karabiner Elements there are some unfortunate limitations

But yeah, I'd be surprised if it were a powerful system, that's so rare in the keyboard land...


Can’t you handle a lot of that with Automator? Honestly it’s funny to see so many people saying Mac is terrible at this when I can’t imagine a worse OS for shortcuts than windows and Linux you have to script everything. Mac managed to make it user friendly and scriptable at the same time.


A lot of what? What's your Automator handler that I can select a dynamic menu item and change those two keybinds with my arbitrary key combo?

Also, how is it user friendly to require users to retype the full menu taxonomy???


>What's your Automator handler that I can select a dynamic menu item and change those two keybinds with my arbitrary key combo?

no clue, I don't use automator, but from what I understand of applescript and automator this should be completely possible. For what it's worth I use Hammerspoon like the other commentor and it's by far the best system I've seen for any OS. Also the more people that use it the better it gets so I highly recommend it.

>Also, how is it user friendly to require users to retype the full menu taxonomy???

I mean... you're able to change any arbitrary menu item in any program that exists. That's extremely powerful. I don't really know how you would make it make it any more user friendly.


I've already tried Hammerspoon a while ago

> any arbitrary menu item in any program that exists

That's false

> I don't really know how you would make it make it any more user friendly.

Just read my comment, it tells you exactly how to make it user friendly. And it's easy to come up with a few more ideas


> That's false

In what way?

> Just read my comment, it tells you exactly how to make it user friendly.

I’ve reread all of your comments in this thread and find no such thing.


use hammerspoon[0] if you are so dead set on making this edgecase work. on the whole the system is quite flexible.

there's also applescript that can do a lot of other things too. I dislike the language but it's fairly powerful too.

[0] http://www.hammerspoon.org/docs/hs.application.html#getMenuI...


I'm not dead set, that's why I wouldn't waste time on writing a "this can take a little while to complete, because quite a large number of round trips are required to the source application, to get the information" in another bad language when it's much simpler to use Karabiner to send the offending keybind

By the way, what would I do with this menu info, is there another function to permanently rebind a given menu item?

I was just pointing a few fundamental flaws in this "surprisingly powerful" system


you can write a hammerspoon script that will intercept some hotkey you want to use for that menu item and select it yes.


I don't get it, this is much worse vs a direct rebind, which is instant

Also, this brings confusion between the key shown in the menu and the key that is used to activate it, is the Mac ecosystem powerful enough to fix this little usability snag?


I was talking about pre-OSX.


My observation too.

It is kind of evolution, just for computers.

Macs early developed good pointing devices and as a result many keyboard related aspects can afford to be somewhere between weird and crazy.

Bonus for Mac people insisting everything is fine.

And I still consider getting a MacBook Pro next month, Windows PCs are that bad even with WSL :-/

Edit: at least these days, fn and ctrl can finally be remapped and CMD-tab can be fixed so it works consistently between two Firefox (or two Safari) windows, an IDE and Finder. It used to be that I would have to CMD-tab to the Firefox group, then CMD-| to get to the correct browser window and it was one of the things that truly messed up my workdays the last time I used Mac back in 2012. (No dedicated home/end buttons and every app seemingly being free to choose what shortcut they would use for it was probably the most painful one though.)


I’ll just throw Magnet in the ring for a tiling solution.

I have keyboard shortcuts well in muscle memory now for left/right third/half/two-thirds, and all four corners at a quarter of screen, and full screen. I find that meets 90% of my needs.

Left/right two-thirds is my go-to when coding on 14”. VSCode on the left, Safari on the right. Both big enough to work well, but both leave enough of the other visible to be useful.

Nice little app, costs some reasonable amount of money as a one-off purchase.


I use Magnet, but settings not syncing across devices is a bit annoying.


Hammerspoon can handle that and much more, way more customizable and you can do amazing things like grid layouts. Completely open source, totally free, and the more people that use it the better it gets so I highly recommend it.


You can mimic that in part with yabai (https://github.com/koekeishiya/yabai). Disabling SIP gets you instant Space switching and creating without animation just like Pop_OS.

But you can get very far to an automated tiling WM and keyboard navigation even without disabling SIP.


Just because this often comes up; I don't want to disable SIP. However I'm very happy with yabai.


I didn't realize it was possible to use it without disabling SIP, excited to give it another shot! AltTab with a few modifications is currently my preferred way to manage windows.


To be honest I always wrote off Yabai until someone here commented that it works fine without disabling SIP. Only a small subset of features need that.


As another suggestion, Better Touch Tool does great for tiling window manager and adding shortcuts.

It's one of the first things I always need. It started coming from Linux with a similar experience to you.

----

A big shortcut for me is cmd+shift+/

It allows searching on all submenus. So when I'm working on an IDE for example, it's how I find those obscure options I know are there but I don't use frequently enough to remember the shortcut.


I use NixOS+GNOME+pop-shell for tiling windows on Linux, and I love it!

I am quite frequently on MacOS, and I use Yabai[0] and skhd[1], managed with Nix-Darwin[2] for tiling windows and custom keyboard shortcuts. With how I make my Linux and MacOS builds look and feel identical it's pretty easy for me to forget when I'm on one vs the other.

For anyone curious, here's my repository for deploying my configs[3]. It's awesome to have one source of truth for managing NixOS servers and workstations, MacOS workstations, and other Linux workstations with Nix installed.

[0] https://github.com/koekeishiya/yabai

[1] https://github.com/koekeishiya/skhd

[2] http://daiderd.com/nix-darwin/

[3] https://github.com/heywoodlh/nixos-configs


I've been using https://www.hammerspoon.org/ for more than 5 years and it's perfect.


You are probably better off ignoring Apples built in Window management controls unless you really like the way “Spaces” work. I find them too restrictive. Same with the green icon “full screen” mode. As others have mentioned, there are tones of third part tools that manage windows in various ways and you’ll probably find one that works for you.


If you're okay with a non-open source app (App store purchase) -- "Magnet" is the way to go.

Otherwise go with whats been suggested by others


Divvy for window management (assign your own shortcut).

For almost everything else there's a standardized system-wide shortcut (including the ancient ones like Ctrl+a/e for text).

If there isn't a shortcut, but the app exposes the action in a menu, you can assign a custom shortcut from keyboard settings.

MacOS has traditionally been very friendly to keyboard navigation. Well, until somewhat recently when mobile-only/mobile-first devs started just slapping Catalyst on and calling it a day. Even Apple's own apps suffer from this.


> If there isn't a shortcut, but the app exposes the action in a menu, you can assign a custom shortcut from keyboard settings.

The standardized menu system is underrated, IMO.

Setting aside preferences for the menubar being global vs. being attached to app windows, the fact that it's system-owned makes features like this possible. Apple was able to implement custom keybinds for menu items in arbitrary apps because they know that 99.9% of apps use the system menubar. This would be next to impossible on Windows where menubar implementations are disparate and numerous (even within first party apps), and while something similar could be patched together on Linux, it'd only work on some (mostly Qt) apps since GTK has mostly abandoned menus altogether.

This also enables the menu search in macOS Help menus to work, which is underused despite being basically a command palette for every app.


Menu bar search is underused for sure. One of my top commands after I learned about it.


Yeah I find some of the shortcuts confusing, like if I want to enter a directory in Finder I push ENTER and that's ... rename? Instead I'm meant to push PLACE OF INTEREST SIGN + O


Oh yeah, that's a stupid keybind, made worse by the fact you can't reliably change it since you can't differentiate when you're in a text field

But then you can mostly get by with a finder alternative


Cmd-down would also work.



though PR should explain Why more than What, ie I can figure out what changed looking at files, but without context and design details an autogenerated description doesnt make a review any easier


Thats a good project. My PR Comments are often less than good. I could use something like that.


About using svg images I wonder if there are some studies about inlining versus using a reference to an external URL.

I have a web page with the same icons repeated 20+ times, and I my intuition is that it it could be optimized by using a shared reference instead of repeated inlined svg sources in the html documention — which is font-awesome default behavior.


SVG can handle this situation. You can define the SVGs somewhere in the HTML document and reference them anywhere you want. https://css-tricks.com/lodge/svg/13-svg-icon-system-use-elem...

The syntax is like this:

<!-- Define the SVGs. I usually put them in an element with `display:none` --> <svg id='svgID12345'><path d="..."/></svg>

<!-- Use the SVG via `<use xlink>` anywhere you want to see it. --> <svg><use xlink:href='#svgID12345'></svg>

You can reference external SVGs in external files as well. Just use `<use xlink:href="defs.svg#icon-1"></use>`. https://css-tricks.com/svg-use-with-external-reference-take-...


<use xlink> will accomplish what you want to do.


About the method used:

Q: Are you emulating apps through an HTML5 webpage?

A: Yes and no. These are not like the "web apps" offered through safari. HTML5 is used to render the visuals but the backend is in C. Using protocols intended for enterprise (that they can't take away) we are able to provide access to the hardware, we also have a working version of iCloud. Geolocation, gyroscope, notifications all work as well.

Q: You are violating the terms of the enterprise license and they can invalidate your key, though for some reason the keys still work if you change the phone date to before the key was invalidated.

A: we reverse engineered our end of the protocol without signing an agreement with them and use self signed certificates. Also we are not using the method mentioned int hat article. No one has done this before.


* Bitcoin donations


Bitcoin pledges would be great, but that option isn't available. They only have PayPal and wire transfer, with credit cards coming "in a few days".


Does it implement RFC 2324? https://tools.ietf.org/html/rfc2324


You would only get error 418 I'm a teapot. For obvious reasons !


Meteor UI does it as well.


cd, cat, ls, head, tail, grep


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: