I think the focus should be on making JavaScript use less memory right now a simple website like Discord can use 1GB of memory for no reason that's the biggest issue in my opinion making javascript take less memory will also make it more performant at the same time (thanks to cache locality). I don't think CPU performance is really an issue right now the solution could be adding fixed layout stack allocated primitives to the language basically structs and arrays. imagine for example if React used struct values for all components props. in a typical app all those objects would not need to be tracked by garbage collection or heap allocated. they would simply be values on the stack deleted at function scope end
JavaScript was never designed for the server. If that doesn't change soon, JavaScript on the server will be replaced by Rust. The fix needs to happen in the engine. Sound types. Ahead-of-time compilation. Threads with shared objects.
6
27
3,038
tried using the same prompt and asked to make the video more colorful. opus 5.5 is really amazing one shot, it put all the code in a single index.html file and rendered it using playwright frame by frame in a headless window, then ffmpeg to generate the mp4. it used a seek function and eval to render each frame. the model seems to prefer doing everything with zero dependencies from scratch instead of using tools like remotion, egaki, or hyperframes. maybe it's because anthropic created special RL environments to create videos from code? and they probably went for the simplest possible approach. the model seems to have an internal idea of the video it wants to make. it can even reason spatially. it then outputs some spaghetti code with magic numbers to make the video actually render. the code doesn't seem to be the important part. it's just an output artifact. i wonder how good the model is at recreating the internal video representation from the code. for analyzing the audio, it used python. it used it to output timestamps of the beats and used them as magic numbers in the html page. if things continue to go this way, humans can't look at the code anymore. it's like looking at assembly. it makes zero sense to us. we need a higher-level form to describe the model's internal representation. markdown sure isn't that. traditional tools like ae aren't it either. here is the code it used gist.github.com/remorses/3d4…
opus 5.5 is f*cking cracked at motion design this entire video is code, 0 after effects im open sourcing the prompt template for these motion designs steal it to recreate these ↓ <inputs> Ask me for: 8 to 12 UI states I want the shape to become (e.g. button, loader, player, slider, toggle, tabs, chart, command palette, toast), pure black and white or one accent color, and a royalty-free song around 120 BPM (e.g. Mixkit, free for commercial use). </inputs> <direction> Dribbble-level UI motion. One shape, never cut: every state is the same element morphing its size, radius and color while its content swaps with a short blur. A cursor drives every change with real clicks and drags. Light warm-gray canvas, black and white components, one clean UI font (Geist). Springs everywhere, a tiny overshoot at most. The camera zooms so each state fills the frame. The last frame is the first frame, so it loops. Banned: bouncy easing, particle bursts, glows, gradients on UI chrome, mismatched icon strokes, dead time, anything that looks like a template. </direction> <structure> 120 BPM, 7 bars, something happens on every beat. Button → loader → check → dynamic island → music player with a play/pause morph → scrub the progress bar → it becomes a volume slider that stretches when dragged past max → a toggle flips on the beat → the knob becomes a liquid tab indicator → the tabs open into a chart that draws itself, with a tooltip on hover → it collapses into ⌘K → type to filter → enter → toast → back to the button. </structure> <build> 1. One HTML file, square 1440x1440. Every style is computed from time inside seek(t): no CSS transitions, no timers, no state carried between frames. 2. Springs are closed-form step responses. A value that changes target many times is the sum of one spring per change, so it stays a pure function of time. 3. The tab indicator's two edges ride different springs, so the leading edge stretches ahead of the trailing one. Same trick for the toggle knob. 4. Drags are direct manipulation: while the cursor is held, the value is computed from its position. On release it springs back from wherever it was. 5. Analyze the song with numpy for the beat grid and start on a downbeat. Place every UI sound by its measured peak. 6. Render with Playwright: 4 subframes per frame, blended with ffmpeg tmix for motion blur at 60fps. 7. Render one frame per beat before the full render. Fix anything off the grid, cramped or hard to read. </build> <gotchas> Never put will-change on anything the camera scales or the text renders blurry. Text that swaps inside a morphing container needs its own enter and exit timing or it overlaps. Make the last frame identical to the first, cursor position and speed included, or the loop stutters. </gotchas> <start> Ask me for the inputs, then show me the state list on the beat grid before you write any code. </start>
6
3
158
19,390
this is the golden age of SWE gpt-6-sol is amazing opus-5-5 is amazing LFG
5
38
1,187
grok 4.7 is officially cooked it writes hacks on top of hacks to do anything DO NOT use this model unless you want your codebase turned into spaghetti slop
9
1
90
10,375
very sad about this because I was the top grok 4.6 fanboy
5
599
I think grok 4.7 just got opusfived
2
19
2,760
latest release of kimaki supports agents spawning more agents and communicating between each other kimaki cli can list channels and threads, then read transcript of other sessions and send messages. this way they can communicate you can try asking to split some big work into parallel sessions and sessions can send messages between them. they can even send their own session id to ask questions to another agent, the agent then can send the reply back I wanted to replicate what openai did with Navier Stokes where they used 10k agents that could message each other. now kimaki has the same capability and it looks cooler because you can see tens of threads spawning up in Discord and messaging each other 😆 the nice thing is that this works even across machines thanks to using Discord as the orchestrator. Discord threads are the source of truth to find other agents the only drawback of this method is that messages now are just user prompts. so agents can stop doing what they were doing before sometimes I think openai used the same approach tho, so models like astra seem to continue previous work anyway and use steering prompts only as hints
3
1
9
826
Rust is pretty great when working with agents at least until you start using async Rust. complexity explodes at that point
2
10
788
what is the last good Opus model?
16% opus 5
33% opus 4.8
5% opus 4.7
46% opus 4.6
141 votes • Final results
5
1
7
1,144
astra is crazy. time to lock in
7
32
1,976
another app built with GPUIX! super cool, let’s kill Electron
I’m super excited for Superlogical, but I couldn't resist trying to build a superterminal. I asked Claude Fable 5.1 for planning and Muse Spark 1.3 for execution to build a multiplexer superterminal using Rust, gpuix, and Bun and it actually pulled it off. - Native cross-platform: Mac, Windows, & Linux - Speed: Hits that legendary "half-bounce" startup time - Layouts: Split panels + horizontal/vertical tabs - Open Source: 100%
10
8
276
29,574
I tried asking astra to recreate the pantheon using my little tool I created a few months ago, rensei it’s a simple cli that can use typescript code to create 3d models. it has a command to take screenshots from many angles to let the model iterate. it was created to help me 3d print a replacement part I didn’t expect anything. last time it barely managed to create a super simple part but then… it fucking recreated the pantheon I could not believe it. how the fuck does this thing work? it just used some typescript and 3d utils functions like exclusion and intersection this is the biggest breakthrough of the year you can now 3d print basically anything you have an image of
made a cli to let agents generate 3d printable models it can generate STL files from typescript code. then screenshot them using node.js Three.js and WebGPU I just used it to 3d print a missing part I needed from 4 images + Opus via @opencode agents can write typescript code to generate a 3d model, then do `rensei screenshot model.ts --output render.png` after reading the screenshot the model can compare to the intial image, update the code, repeat this feedback loop is perfect to clone simple parts from images into STL models. you can give it measurements and it will just update the code to match your part rensei means transmutation in Japanese. this stuff feels like alchemy in Full Metal Alchemist. materialize things from nothing github.com/remorses/rensei
1
25
2,186
this model has an incredible sense of space. it can basically think of a point in space and know the perfect coordinates for it. the code is a mess of magic numbers everywhere
194