4,910 karma · joined March 25, 2015
GitHub: https://github.com/jasonjmcghee
The only solution I've seen on a Mac is doing it on a separate monitor.
I couldn't find a solution here and have built similar things in the past so I took a crack at it using CGVirtualDisplay.
Ended up adding a lot of productivity features and polished until it felt good.
Curious if there are similar solutions out there I just haven't seen.
I took inspiration from tiling window managers, tmux, picture-in-picture, and Starcraft.
It's a canvas that lets you create virtual desktops that you can teleport the mouse or windows into, move stuff between, save window position, size, canvas position, zoom etc. as macro / savepoints you can rapidly swap between.
There are a bunch of features.
I have used it for longer than it took to me to build to really nail the feel of using it. I really like where it is, but everything can be improved, and would love feedback - especially for performance.
There's 2800+ people in the discord.
And the community made a backup of the site https://archive.canvascommons.io/
Update: sounds like the author had some life changes and had to stop his YouTube channel which was the primary motivator for the library- but there are forks popping up (https://github.com/canvas-commons/canvas-commons).
---
This notion that an open source library is "dead" and shouldn't be used because it's not being actively updated is a bit odd. You can fork and fix issues you have. It's got years of many people's work put into it. It's a great library and widely used on YouTube and elsewhere.
But- probably reasonably simple, if you mean the "writing" style of manim. That's just animating an SVG stroke.
And then colors and typography
It can be used for interactive blog posts and embedded animations etc as well.
It also has a built in animation editor / timeline.
Most smartphones are.
I'm genuinely surprised by this comment.
For example, I thought there was universal sentiment that apple silicon / M-series computers are pretty unmatched.
I'm guessing it's fast due to 3B experts
This is very similar to what the agent is doing. You are building the appropriate environment for it to be able to complete the task most reliably etc
Not just functions/tools and documentation available (which is similar to a library), also context and critically, enforcement of behavior.
This is probably the key thing that makes it a "harness". If the agent can do whatever it wants, it's not in a harness.
Depends- how good is your signal? Mosh has a great property that it buffers everything locally so there's no lag even if your connection sucks.
On ssh, every keystroke is a roundtrip
What I posted "just works".
Could even use cc to check in on and/or "send-keys"
What wasn't working about mosh? Just install mosh and use mosh to connect
Based on my experience many people don't know this is a thing you can do.
- - -
get tailscale (free) and join on both devices
install tmux
get an ios/android terminal (echo / termius)
enable "remote login" if on mac (disable on public wifi)
mosh/ssh into computer
now you can do tmux then claude / codex / w/e on either device and reconnect freely via tmux ls and tmux attach -t <id>
- - -
You can name tmux and resume by name via tmux new -s <feature> and tmux attach -t <feature>
Nearly all of its value is facilitating your interaction with the LLM, the tools it can use, and how it uses them.
Qwen3.5 released a couple of days ago but I'm not that RAM rich
But Rust can dynamically link with dylib but I believe it's still unstable.
It can also dynamically load with libloading.
The one built in is much faster, and you only have to toggle it on.
Are these so much more accurate? I definitely have to correct stuff, but pretty good experience.
Also use speech to text on my iphone which seems to be the same accuracy.
CRDT or OT will work great but are even overkill. But so many of the edge cases you'd usually need to think about just disappear.
(I've built an agent / chat that used CRDT to represent the chat. You can have an arbitrary number of tabs, closing/opening at any time. All real time, in sync.)
Families of model sizes work great for speculative decoding. Use the 1B with the 32B or whatever.
It's a balance as you want it to be guessing correctly as much as possible but also be as fast as possible. Validation takes time and every guess needs to be validated etc
The model you're using to speculate could be anything, but if it's not guessing what the main model would predict, it's useless.
The most absurd benchmaxxing.
https://x.com/jeffdean/status/2024525132266688757?s=46&t=ZjF...
Think of it like an "IDE actions". Done right, there's no need to ever use the GUI.
As opposed to just being documentation for how to use the IDE with desktop automation software.