claude code's ultracode feature used to kill my sub limits with just 1-2 dynamic flows
With the intelligence, speed and usage limit uplift in opus 5.5, it feels like driving a ferrari that doesn't run out of gas so quickly
for anything you want to quickly create, your prompt is now:
<simple idea> + with ultracode
then you just kick back, relax and wait for the agent swarm to give you an artifact that brings your idea to life
1
4
122
made a galactic pixel racer game with opus 5.5, can you beat my score? 👿
velven.ai/david/pixel-racer
3
113
David retweeted
建築バトルゲーム【トンテンかんてん】を公開しました！ライバルより早く、かつ頑丈な橋を作ってください！（おもろい...！と思ってくれたらオンライン対戦機能とか、追加要素も盛り込みたいと思います）こちらから遊べます↓
amix-design.com/tl/tool-g-to…
12
53
493
88,568
thanks to @poteto, three things have instantly improved code quality & performance of my shipped features
1) banned comments
2) codified more lint rules over any other context-based safeguards
3) built a custom verification cli (e.g., open routes at certain screen width, measure long tasks & LCP, etc.)
4) created a few "gardener" scripts to identify & clean up anti-patterns before they spread like a virus
here's how i shipped 2,500 PRs last month to production
this was originally supposed to be for Cursor Compile in London. i couldn't make it since i was livestreaming for Grok @Bot Galaxy so i'm making it available for free here on X! watch it on 2x speed, i talk slowly
77
inspired by @steveruizok, i ran a similar workflow where i swarmed ~500 opus / fable agents at my app and had them coordinate over an offline @tldraw board
Generally, they:
1. screenshot every screen at every size
2. score each on a rubric you define (e.g., consistency, performance, language, etc.)
3. fix whatever scores under 4/5
4. repeat until everything passes
You can tune how many rounds, scorers per round, scope of screens, modes (report or report & fix), sample sizes (if you have lots of pages), and more. Just ask Claude.
It doesn't tank usage limits as much as you would think, and results were great
I was invited to test Opus 5.5.
The differences at the high end are hard to notice unless you're being ambitious. They reward ambition though.
If you can't tell the difference here, or if you're intentionally micro-prompting a model, I think the fast models (composer, Gemini flash, etc) are much better. Whenever I have to do that type of work, I switch to those.
For models like Opus 5.5 though, I've been treating every problem as a Manhattan Project level organization of labor. There is a meteor heading towards the Earth and our only salvation is making this microcopy internally consistent.
My prompts have been like: create a series of dimensions by which to assess quality / consistency, then generate images of every screen in my app, it them on a tldraw offline board, and analyze them in random groups, identifying problems and making fixes until every screen / group passes the threshold in every dimension.
Get weird, get creative. Opus is a freak, it feels good to Opus, Opus likes it.
2
5
154
Opus 5.5 is great for cleaning up content slop from previous iterations
Try this prompt, works well for developer docs, seo deep link pages, landing pages, and more:
Edit this so it reads like a sharp human wrote it: cut filler, hedging, buzzwords, formulaic phrasing, and anything that says nothing concrete. Keep the meaning, facts, and intended tone for its audience, add nothing new, and flag doubtful claims. Use a Fable 5.1 subagent to adversarially review your changes
4
94
notes from @cursor_ai /@SpaceXAI compile (developer event) in London
1) whiteboard talks continue to be the most engaging form of knowledge sharing
2) can each team member explain your system with a diagram in less than 1 minute, and only with a small degree of variance ?
3) for each pr, review diffs in team maintained diagrams rather than in code (for relevant parts)
4) way more gtm focus on grok bot than cursor
5) lots of cursor origin feature discussion, more releases in upcoming weeks. Maybe time to try it ?
1
5
147
24/7 live lo-fi beats!
Music generated with @ElevenLabs, pictures by @OpenAI's Image 2, and painted with 420k-3m particles in @threejs
flip the channel mode on and work, study, vibe or whatever at velvenizer.com
3
191
upload an image or clip for free into scenes mode at velvenizer.com and watch the particles take shape.
tune particles up to 3 million if your device supports it
1
2
136
420k particles in three.js that listen
Turn on your mic and try, or start with velven's select music
velvenizer.com
4
32
4,511
introducing the Velvenizer: an in-browser 3d audio visualizer that takes the shape of whatever's playing (including yourself)
enable your mic, put on your favorite tune, and enjoy
velven.ai/david/velvenizer
2
1
4
211
white christmas gave us a glimpse of that future
Maybe we can't agree if a fly is conscious but if we can map out the full brain of fruit fly, theoretically it could eventually be possible to do the same for a human brain, but that raises the question is a digital brain conscious.
Most people will say modern LLMs are not but if you have 1:1 digital recreation of a human brain is it? If not why, is there something special about organic matter, is there something more to consciousness than the brain, is the body required, is there a spiritual component. These are uncomfortable questions that are going to have to be answered if we're going to start forcing digital brains to play Beat Saber
2
288