The world's fastest AI code editor is here
Zed is engineered from scratch in Rust like a video game. Not another fork, but a purpose-built editor designed for collaboration between humans and AI, delivering a lightning fast agentic editing experience.
https://t.co/dPumqwS4ls
@daylightco@savkruger@JonathanHillis Hey! Typically just strategic use of shade. If both my head and screen are in shade it's fine - https://t.co/0M5x9Vd3sN . Can be a bit of a hassle to rearrange as the sun moves. Looking forward to trying Jon's daylight computer!
@amix3k Yuup, I printed this out a couple years ago and stuck it on my wall https://t.co/1r7I9sJo0z and have filled in one square on sunday for about 2 years. A good reminder of the preciousness of time!
@runarorama Ok I actually want to try this language, with the right sugar it could be nice. To make it definitely bad, "just pick the first one" if multiple values in scope qualify for use.
@runarorama A language where locals have no names, they just get passed in wherever it's assignable. You declare what coercions are allowed within the scope to guide this routing. Sets of coercions can be given a name and parameterized.
@soychotic@supersat I've tried climbing belay glasses upside down for this, and it sorta worked but the narrowish field of view and chromatic aberration interfered. Yours look more suited to the application! Pic from an old blog post about computing in this posture - https://t.co/Dih4vIhEvA
@mwotton Ah, another reason the cleaning plugins would immediately exist - most compilers reject unicode, even unicode aware GHC with -XUnicodeSyntax is rejecting me doing `intersperse '\xE01A1'`. This is probably a good thing, better to fail when something funky is happening
What if unicode added codepoints indicating that text is machine generated (LLMs)? Invisible by default, but would be preserved on copy when supported.
@FakeUnicode Awesome, thanks!! I propose use of VS178 for this as its codepoint is U+E01A1 :)
Initial experiments are mixed.. Some things handle it perfectly, but emacs and urxvt do not.
@FakeUnicode Definitely leaning towards annotating every single character, to avoid statefulness and for well behaved copy-paste.
I only have a rudimentary knowledge of unicode stuff. Is there a way to add an annotation to every character in a way that does not affect display?
@FakeUnicode Alternatively, just annotate every char with an invisible combining codepoint. Yeah, not so efficient, but should compress well so at least wouldn't waste much resources in transit.
@FakeUnicode My rough read now of info about RTL isolate codepoints is that they are quite stateful delimiters and this is being encouraged as of unicode 6.3. My hope is that if there is already code handling copy/paste of stateful delimiters, this could be adapted to LLM provenance markers
@sand2drn @mwotton I believe that codepoints E0080..E0FFF (~3k codepoints) are both reserved for future use and ignorable (current clients skip for display)